找回密码
 立即注册→加入我们

QQ登录

只需一步,快速开始

搜索
热搜: 下载 VB C 实现 编写
查看: 7193|回复: 0

【书籍分享】NASM 2.11 全英文文档 可在线查看

[复制链接]

1112

主题

1653

回帖

7万

积分

用户组: 管理员

一只技术宅

UID
1
精华
245
威望
744 点
宅币
24254 个
贡献
46222 次
宅之契约
0 份
在线时间
2298 小时
注册时间
2014-1-26
发表于 2014-6-17 23:54:51 | 显示全部楼层 |阅读模式

欢迎访问技术宅的结界,请注册或者登录吧。

您需要 登录 才可以下载或查看,没有账号?立即注册→加入我们

×
                        The Netwide Assembler: NASM
                        ===========================

Chapter 1: Introduction
-----------------------

   1.1 What Is NASM?

       The Netwide Assembler, NASM, is an 80x86 and x86-64 assembler
       designed for portability and modularity. It supports a range of
       object file formats, including Linux and `*BSD' `a.out', `ELF',
       `COFF', `Mach-O', Microsoft 16-bit `OBJ', `Win32' and `Win64'. It
       will also output plain binary files. Its syntax is designed to be
       simple and easy to understand, similar to Intel's but less complex.
       It supports all currently known x86 architectural extensions, and
       has strong support for macros.

1.1.1 Why Yet Another Assembler?

       The Netwide Assembler grew out of an idea on `comp.lang.asm.x86' (or
       possibly `alt.lang.asm' - I forget which), which was essentially
       that there didn't seem to be a good _free_ x86-series assembler
       around, and that maybe someone ought to write one.

       (*) `a86' is good, but not free, and in particular you don't get any
           32-bit capability until you pay. It's DOS only, too.

       (*) `gas' is free, and ports over to DOS and Unix, but it's not very
           good, since it's designed to be a back end to `gcc', which
           always feeds it correct code. So its error checking is minimal.
           Also, its syntax is horrible, from the point of view of anyone
           trying to actually _write_ anything in it. Plus you can't write
           16-bit code in it (properly.)

       (*) `as86' is specific to Minix and Linux, and (my version at least)
           doesn't seem to have much (or any) documentation.

       (*) `MASM' isn't very good, and it's (was) expensive, and it runs
           only under DOS.

       (*) `TASM' is better, but still strives for MASM compatibility,
           which means millions of directives and tons of red tape. And its
           syntax is essentially MASM's, with the contradictions and quirks
           that entails (although it sorts out some of those by means of
           Ideal mode.) It's expensive too. And it's DOS-only.

       So here, for your coding pleasure, is NASM. At present it's still in
       prototype stage - we don't promise that it can outperform any of
       these assemblers. But please, _please_ send us bug reports, fixes,
       helpful information, and anything else you can get your hands on
       (and thanks to the many people who've done this already! You all
       know who you are), and we'll improve it out of all recognition.
       Again.

1.1.2 License Conditions

       Please see the file `LICENSE', supplied as part of any NASM
       distribution archive, for the license conditions under which you may
       use NASM. NASM is now under the so-called 2-clause BSD license, also
       known as the simplified BSD license.

       Copyright 1996-2011 the NASM Authors - All rights reserved.

       Redistribution and use in source and binary forms, with or without
       modification, are permitted provided that the following conditions
       are met:

       (*) Redistributions of source code must retain the above copyright
           notice, this list of conditions and the following disclaimer.

       (*) Redistributions in binary form must reproduce the above
           copyright notice, this list of conditions and the following
           disclaimer in the documentation and/or other materials provided
           with the distribution.

       THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
       "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
       LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
       FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
       COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
       INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
       BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
       LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
       CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
       LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
       ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
       POSSIBILITY OF SUCH DAMAGE.

   1.2 Contact Information

       The current version of NASM (since about 0.98.08) is maintained by a
       team of developers, accessible through the `nasm-devel' mailing list
       (see below for the link). If you want to report a bug, please read
       section 12.2 first.

       NASM has a website at `http://www.nasm.us/'. If it's not there,
       google for us!

       New releases, release candidates, and daily development snapshots of
       NASM are available from the official web site.

       Announcements are posted to `comp.lang.asm.x86', and to the web site
       `http://www.freshmeat.net/'.

       If you want information about the current development status, please
       subscribe to the `nasm-devel' email list; see link from the website.

   1.3 Installation

1.3.1 Installing NASM under MS-DOS or Windows

       Once you've obtained the appropriate archive for NASM,
       `nasm-XXX-dos.zip' or `nasm-XXX-win32.zip' (where `XXX' denotes the
       version number of NASM contained in the archive), unpack it into its
       own directory (for example `c:\nasm').

       The archive will contain a set of executable files: the NASM
       executable file `nasm.exe', the NDISASM executable file
       `ndisasm.exe', and possibly additional utilities to handle the RDOFF
       file format.

       The only file NASM needs to run is its own executable, so copy
       `nasm.exe' to a directory on your PATH, or alternatively edit
       `autoexec.bat' to add the `nasm' directory to your `PATH' (to do
       that under Windows XP, go to Start > Control Panel > System >
       Advanced > Environment Variables; these instructions may work under
       other versions of Windows as well.)

       That's it - NASM is installed. You don't need the nasm directory to
       be present to run NASM (unless you've added it to your `PATH'), so
       you can delete it if you need to save space; however, you may want
       to keep the documentation or test programs.

       If you've downloaded the DOS source archive, `nasm-XXX.zip', the
       `nasm' directory will also contain the full NASM source code, and a
       selection of Makefiles you can (hopefully) use to rebuild your copy
       of NASM from scratch. See the file `INSTALL' in the source archive.

       Note that a number of files are generated from other files by Perl
       scripts. Although the NASM source distribution includes these
       generated files, you will need to rebuild them (and hence, will need
       a Perl interpreter) if you change insns.dat, standard.mac or the
       documentation. It is possible future source distributions may not
       include these files at all. Ports of Perl for a variety of
       platforms, including DOS and Windows, are available from
       www.cpan.org.

1.3.2 Installing NASM under Unix

       Once you've obtained the Unix source archive for NASM,
       `nasm-XXX.tar.gz' (where `XXX' denotes the version number of NASM
       contained in the archive), unpack it into a directory such as
       `/usr/local/src'. The archive, when unpacked, will create its own
       subdirectory `nasm-XXX'.

       NASM is an auto-configuring package: once you've unpacked it, `cd'
       to the directory it's been unpacked into and type `./configure'.
       This shell script will find the best C compiler to use for building
       NASM and set up Makefiles accordingly.

       Once NASM has auto-configured, you can type `make' to build the
       `nasm' and `ndisasm' binaries, and then `make install' to install
       them in `/usr/local/bin' and install the man pages `nasm.1' and
       `ndisasm.1' in `/usr/local/man/man1'. Alternatively, you can give
       options such as `--prefix' to the configure script (see the file
       `INSTALL' for more details), or install the programs yourself.

       NASM also comes with a set of utilities for handling the `RDOFF'
       custom object-file format, which are in the `rdoff' subdirectory of
       the NASM archive. You can build these with `make rdf' and install
       them with `make rdf_install', if you want them.

Chapter 2: Running NASM
-----------------------

   2.1 NASM Command-Line Syntax

       To assemble a file, you issue a command of the form

       nasm -f <format> <filename> [-o <output>]

       For example,

       nasm -f elf myfile.asm

       will assemble `myfile.asm' into an `ELF' object file `myfile.o'. And

       nasm -f bin myfile.asm -o myfile.com

       will assemble `myfile.asm' into a raw binary file `myfile.com'.

       To produce a listing file, with the hex codes output from NASM
       displayed on the left of the original sources, use the `-l' option
       to give a listing file name, for example:

       nasm -f coff myfile.asm -l myfile.lst

       To get further usage instructions from NASM, try typing

       nasm -h

       As `-hf', this will also list the available output file formats, and
       what they are.

       If you use Linux but aren't sure whether your system is `a.out' or
       `ELF', type

       file nasm

       (in the directory in which you put the NASM binary when you
       installed it). If it says something like

       nasm: ELF 32-bit LSB executable i386 (386 and up) Version 1

       then your system is `ELF', and you should use the option `-f elf'
       when you want NASM to produce Linux object files. If it says

       nasm: Linux/i386 demand-paged executable (QMAGIC)

       or something similar, your system is `a.out', and you should use
       `-f aout' instead (Linux `a.out' systems have long been obsolete,
       and are rare these days.)

       Like Unix compilers and assemblers, NASM is silent unless it goes
       wrong: you won't see any output at all, unless it gives error
       messages.

2.1.1 The `-o' Option: Specifying the Output File Name

       NASM will normally choose the name of your output file for you;
       precisely how it does this is dependent on the object file format.
       For Microsoft object file formats (`obj', `win32' and `win64'), it
       will remove the `.asm' extension (or whatever extension you like to
       use - NASM doesn't care) from your source file name and substitute
       `.obj'. For Unix object file formats (`aout', `as86', `coff',
       `elf32', `elf64', `elfx32', `ieee', `macho32' and `macho64') it will
       substitute `.o'. For `dbg', `rdf', `ith' and `srec', it will use
       `.dbg', `.rdf', `.ith' and `.srec', respectively, and for the `bin'
       format it will simply remove the extension, so that `myfile.asm'
       produces the output file `myfile'.

       If the output file already exists, NASM will overwrite it, unless it
       has the same name as the input file, in which case it will give a
       warning and use `nasm.out' as the output file name instead.

       For situations in which this behaviour is unacceptable, NASM
       provides the `-o' command-line option, which allows you to specify
       your desired output file name. You invoke `-o' by following it with
       the name you wish for the output file, either with or without an
       intervening space. For example:

       nasm -f bin program.asm -o program.com
       nasm -f bin driver.asm -odriver.sys

       Note that this is a small o, and is different from a capital O ,
       which is used to specify the number of optimisation passes required.
       See section 2.1.22.

2.1.2 The `-f' Option: Specifying the Output File Format

       If you do not supply the `-f' option to NASM, it will choose an
       output file format for you itself. In the distribution versions of
       NASM, the default is always `bin'; if you've compiled your own copy
       of NASM, you can redefine `OF_DEFAULT' at compile time and choose
       what you want the default to be.

       Like `-o', the intervening space between `-f' and the output file
       format is optional; so `-f elf' and `-felf' are both valid.

       A complete list of the available output file formats can be given by
       issuing the command `nasm -hf'.

2.1.3 The `-l' Option: Generating a Listing File

       If you supply the `-l' option to NASM, followed (with the usual
       optional space) by a file name, NASM will generate a source-listing
       file for you, in which addresses and generated code are listed on
       the left, and the actual source code, with expansions of multi-line
       macros (except those which specifically request no expansion in
       source listings: see section 4.3.11) on the right. For example:

       nasm -f elf myfile.asm -l myfile.lst

       If a list file is selected, you may turn off listing for a section
       of your source with `[list -]', and turn it back on with `[list +]',
       (the default, obviously). There is no "user form" (without the
       brackets). This can be used to list only sections of interest,
       avoiding excessively long listings.

2.1.4 The `-M' Option: Generate Makefile Dependencies

       This option can be used to generate makefile dependencies on stdout.
       This can be redirected to a file for further processing. For
       example:

       nasm -M myfile.asm > myfile.dep

2.1.5 The `-MG' Option: Generate Makefile Dependencies

       This option can be used to generate makefile dependencies on stdout.
       This differs from the `-M' option in that if a nonexisting file is
       encountered, it is assumed to be a generated file and is added to
       the dependency list without a prefix.

2.1.6 The `-MF' Option: Set Makefile Dependency File

       This option can be used with the `-M' or `-MG' options to send the
       output to a file, rather than to stdout. For example:

       nasm -M -MF myfile.dep myfile.asm

2.1.7 The `-MD' Option: Assemble and Generate Dependencies

       The `-MD' option acts as the combination of the `-M' and `-MF'
       options (i.e. a filename has to be specified.) However, unlike the
       `-M' or `-MG' options, `-MD' does _not_ inhibit the normal operation
       of the assembler. Use this to automatically generate updated
       dependencies with every assembly session. For example:

       nasm -f elf -o myfile.o -MD myfile.dep myfile.asm

2.1.8 The `-MT' Option: Dependency Target Name

       The `-MT' option can be used to override the default name of the
       dependency target. This is normally the same as the output filename,
       specified by the `-o' option.

2.1.9 The `-MQ' Option: Dependency Target Name (Quoted)

       The `-MQ' option acts as the `-MT' option, except it tries to quote
       characters that have special meaning in Makefile syntax. This is not
       foolproof, as not all characters with special meaning are quotable
       in Make. The default output (if no `-MT' or `-MQ' option is
       specified) is automatically quoted.

2.1.10 The `-MP' Option: Emit phony targets

       When used with any of the dependency generation options, the `-MP'
       option causes NASM to emit a phony target without dependencies for
       each header file. This prevents Make from complaining if a header
       file has been removed.

2.1.11 The `-F' Option: Selecting a Debug Information Format

       This option is used to select the format of the debug information
       emitted into the output file, to be used by a debugger (or _will_
       be). Prior to version 2.03.01, the use of this switch did _not_
       enable output of the selected debug info format. Use `-g', see
       section 2.1.12, to enable output. Versions 2.03.01 and later
       automatically enable `-g' if `-F' is specified.

       A complete list of the available debug file formats for an output
       format can be seen by issuing the command `nasm -f <format> -y'. Not
       all output formats currently support debugging output. See section
       2.1.26.

       This should not be confused with the `-f dbg' output format option
       which is not built into NASM by default. For information on how to
       enable it when building from the sources, see section 7.14.

2.1.12 The `-g' Option: Enabling Debug Information.

       This option can be used to generate debugging information in the
       specified format. See section 2.1.11. Using `-g' without `-F'
       results in emitting debug info in the default format, if any, for
       the selected output format. If no debug information is currently
       implemented in the selected output format, `-g' is _silently
       ignored_.

2.1.13 The `-X' Option: Selecting an Error Reporting Format

       This option can be used to select an error reporting format for any
       error messages that might be produced by NASM.

       Currently, two error reporting formats may be selected. They are the
       `-Xvc' option and the `-Xgnu' option. The GNU format is the default
       and looks like this:

       filename.asm:65: error: specific error message

       where `filename.asm' is the name of the source file in which the
       error was detected, `65' is the source file line number on which the
       error was detected, `error' is the severity of the error (this could
       be `warning'), and `specific error message' is a more detailed text
       message which should help pinpoint the exact problem.

       The other format, specified by `-Xvc' is the style used by Microsoft
       Visual C++ and some other programs. It looks like this:

       filename.asm(65) : error: specific error message

       where the only difference is that the line number is in parentheses
       instead of being delimited by colons.

       See also the `Visual C++' output format, section 7.5.

2.1.14 The `-Z' Option: Send Errors to a File

       Under `MS-DOS' it can be difficult (though there are ways) to
       redirect the standard-error output of a program to a file. Since
       NASM usually produces its warning and error messages on `stderr',
       this can make it hard to capture the errors if (for example) you
       want to load them into an editor.

       NASM therefore provides the `-Z' option, taking a filename argument
       which causes errors to be sent to the specified files rather than
       standard error. Therefore you can redirect the errors into a file by
       typing

       nasm -Z myfile.err -f obj myfile.asm

       In earlier versions of NASM, this option was called `-E', but it was
       changed since `-E' is an option conventionally used for
       preprocessing only, with disastrous results. See section 2.1.20.

2.1.15 The `-s' Option: Send Errors to `stdout'

       The `-s' option redirects error messages to `stdout' rather than
       `stderr', so it can be redirected under `MS-DOS'. To assemble the
       file `myfile.asm' and pipe its output to the `more' program, you can
       type:

       nasm -s -f obj myfile.asm | more

       See also the `-Z' option, section 2.1.14.

2.1.16 The `-i' Option: Include File Search Directories

       When NASM sees the `%include' or `%pathsearch' directive in a source
       file (see section 4.6.1, section 4.6.2 or section 3.2.3), it will
       search for the given file not only in the current directory, but
       also in any directories specified on the command line by the use of
       the `-i' option. Therefore you can include files from a macro
       library, for example, by typing

       nasm -ic:\macrolib\ -f obj myfile.asm

       (As usual, a space between `-i' and the path name is allowed, and
       optional).

       NASM, in the interests of complete source-code portability, does not
       understand the file naming conventions of the OS it is running on;
       the string you provide as an argument to the `-i' option will be
       prepended exactly as written to the name of the include file.
       Therefore the trailing backslash in the above example is necessary.
       Under Unix, a trailing forward slash is similarly necessary.

       (You can use this to your advantage, if you're really perverse, by
       noting that the option `-ifoo' will cause `%include "bar.i"' to
       search for the file `foobar.i'...)

       If you want to define a _standard_ include search path, similar to
       `/usr/include' on Unix systems, you should place one or more `-i'
       directives in the `NASMENV' environment variable (see section
       2.1.28).

       For Makefile compatibility with many C compilers, this option can
       also be specified as `-I'.

2.1.17 The `-p' Option: Pre-Include a File

       NASM allows you to specify files to be _pre-included_ into your
       source file, by the use of the `-p' option. So running

       nasm myfile.asm -p myinc.inc

       is equivalent to running `nasm myfile.asm' and placing the directive
       `%include "myinc.inc"' at the start of the file.

       For consistency with the `-I', `-D' and `-U' options, this option
       can also be specified as `-P'.

2.1.18 The `-d' Option: Pre-Define a Macro

       Just as the `-p' option gives an alternative to placing `%include'
       directives at the start of a source file, the `-d' option gives an
       alternative to placing a `%define' directive. You could code

       nasm myfile.asm -dFOO=100

       as an alternative to placing the directive

       %define FOO 100

       at the start of the file. You can miss off the macro value, as well:
       the option `-dFOO' is equivalent to coding `%define FOO'. This form
       of the directive may be useful for selecting assembly-time options
       which are then tested using `%ifdef', for example `-dDEBUG'.

       For Makefile compatibility with many C compilers, this option can
       also be specified as `-D'.

2.1.19 The `-u' Option: Undefine a Macro

       The `-u' option undefines a macro that would otherwise have been
       pre-defined, either automatically or by a `-p' or `-d' option
       specified earlier on the command lines.

       For example, the following command line:

       nasm myfile.asm -dFOO=100 -uFOO

       would result in `FOO' _not_ being a predefined macro in the program.
       This is useful to override options specified at a different point in
       a Makefile.

       For Makefile compatibility with many C compilers, this option can
       also be specified as `-U'.

2.1.20 The `-E' Option: Preprocess Only

       NASM allows the preprocessor to be run on its own, up to a point.
       Using the `-E' option (which requires no arguments) will cause NASM
       to preprocess its input file, expand all the macro references,
       remove all the comments and preprocessor directives, and print the
       resulting file on standard output (or save it to a file, if the `-o'
       option is also used).

       This option cannot be applied to programs which require the
       preprocessor to evaluate expressions which depend on the values of
       symbols: so code such as

       %assign tablesize ($-tablestart)

       will cause an error in preprocess-only mode.

       For compatiblity with older version of NASM, this option can also be
       written `-e'. `-E' in older versions of NASM was the equivalent of
       the current `-Z' option, section 2.1.14.

2.1.21 The `-a' Option: Don't Preprocess At All

       If NASM is being used as the back end to a compiler, it might be
       desirable to suppress preprocessing completely and assume the
       compiler has already done it, to save time and increase compilation
       speeds. The `-a' option, requiring no argument, instructs NASM to
       replace its powerful preprocessor with a stub preprocessor which
       does nothing.

2.1.22 The `-O' Option: Specifying Multipass Optimization

       Using the `-O' option, you can tell NASM to carry out different
       levels of optimization. The syntax is:

       (*) `-O0': No optimization. All operands take their long forms, if a
           short form is not specified, except conditional jumps. This is
           intended to match NASM 0.98 behavior.

       (*) `-O1': Minimal optimization. As above, but immediate operands
           which will fit in a signed byte are optimized, unless the long
           form is specified. Conditional jumps default to the long form
           unless otherwise specified.

       (*) `-Ox' (where `x' is the actual letter `x'): Multipass
           optimization. Minimize branch offsets and signed immediate
           bytes, overriding size specification unless the `strict' keyword
           has been used (see section 3.7). For compatibility with earlier
           releases, the letter `x' may also be any number greater than
           one. This number has no effect on the actual number of passes.

       The `-Ox' mode is recommended for most uses, and is the default
       since NASM 2.09.

       Note that this is a capital `O', and is different from a small `o',
       which is used to specify the output file name. See section 2.1.1.

2.1.23 The `-t' Option: Enable TASM Compatibility Mode

       NASM includes a limited form of compatibility with Borland's `TASM'.
       When NASM's `-t' option is used, the following changes are made:

       (*) local labels may be prefixed with `@@' instead of `.'

       (*) size override is supported within brackets. In TASM compatible
           mode, a size override inside square brackets changes the size of
           the operand, and not the address type of the operand as it does
           in NASM syntax. E.g. `mov eax,[DWORD val]' is valid syntax in
           TASM compatibility mode. Note that you lose the ability to
           override the default address type for the instruction.

       (*) unprefixed forms of some directives supported (`arg', `elif',
           `else', `endif', `if', `ifdef', `ifdifi', `ifndef', `include',
           `local')

2.1.24 The `-w' and `-W' Options: Enable or Disable Assembly Warnings

       NASM can observe many conditions during the course of assembly which
       are worth mentioning to the user, but not a sufficiently severe
       error to justify NASM refusing to generate an output file. These
       conditions are reported like errors, but come up with the word
       `warning' before the message. Warnings do not prevent NASM from
       generating an output file and returning a success status to the
       operating system.

       Some conditions are even less severe than that: they are only
       sometimes worth mentioning to the user. Therefore NASM supports the
       `-w' command-line option, which enables or disables certain classes
       of assembly warning. Such warning classes are described by a name,
       for example `orphan-labels'; you can enable warnings of this class
       by the command-line option `-w+orphan-labels' and disable it by
       `-w-orphan-labels'.

       The suppressible warning classes are:

       (*) `macro-params' covers warnings about multi-line macros being
           invoked with the wrong number of parameters. This warning class
           is enabled by default; see section 4.3.1 for an example of why
           you might want to disable it.

       (*) `macro-selfref' warns if a macro references itself. This warning
           class is disabled by default.

       (*) `macro-defaults' warns when a macro has more default parameters
           than optional parameters. This warning class is enabled by
           default; see section 4.3.5 for why you might want to disable it.

       (*) `orphan-labels' covers warnings about source lines which contain
           no instruction but define a label without a trailing colon. NASM
           warns about this somewhat obscure condition by default; see
           section 3.1 for more information.

       (*) `number-overflow' covers warnings about numeric constants which
           don't fit in 64 bits. This warning class is enabled by default.

       (*) `gnu-elf-extensions' warns if 8-bit or 16-bit relocations are
           used in `-f elf' format. The GNU extensions allow this. This
           warning class is disabled by default.

       (*) `float-overflow' warns about floating point overflow. Enabled by
           default.

       (*) `float-denorm' warns about floating point denormals. Disabled by
           default.

       (*) `float-underflow' warns about floating point underflow. Disabled
           by default.

       (*) `float-toolong' warns about too many digits in floating-point
           numbers. Enabled by default.

       (*) `user' controls `%warning' directives (see section 4.9). Enabled
           by default.

       (*) `lock' warns about `LOCK' prefixes on unlockable instructions.
           Enabled by default.

       (*) `hle' warns about invalid use of the HLE `XACQUIRE' or
           `XRELEASE' prefixes. Enabled by default.

       (*) `bnd' warns about ineffective use of the `BND' prefix when a
           relaxed form of jmp instruction becomes jmp short form. Enabled
           by default.

       (*) `error' causes warnings to be treated as errors. Disabled by
           default.

       (*) `all' is an alias for _all_ suppressible warning classes (not
           including `error'). Thus, `-w+all' enables all available
           warnings.

       In addition, you can set warning classes across sections. Warning
       classes may be enabled with `[warning +warning-name]', disabled with
       `[warning -warning-name]' or reset to their original value with
       `[warning *warning-name]'. No "user form" (without the brackets)
       exists.

       Since version 2.00, NASM has also supported the gcc-like syntax
       `-Wwarning' and `-Wno-warning' instead of `-w+warning' and
       `-w-warning', respectively.

2.1.25 The `-v' Option: Display Version Info

       Typing `NASM -v' will display the version of NASM which you are
       using, and the date on which it was compiled.

       You will need the version number if you report a bug.

2.1.26 The `-y' Option: Display Available Debug Info Formats

       Typing `nasm -f <option> -y' will display a list of the available
       debug info formats for the given output format. The default format
       is indicated by an asterisk. For example:

       nasm -f elf -y

       valid debug formats for 'elf32' output format are
         ('*' denotes default):
         * stabs     ELF32 (i386) stabs debug format for Linux
           dwarf     elf32 (i386) dwarf debug format for Linux

2.1.27 The `--prefix' and `--postfix' Options.

       The `--prefix' and `--postfix' options prepend or append
       (respectively) the given argument to all `global' or `extern'
       variables. E.g. `--prefix _' will prepend the underscore to all
       global and external variables, as C sometimes (but not always) likes
       it.

2.1.28 The `NASMENV' Environment Variable

       If you define an environment variable called `NASMENV', the program
       will interpret it as a list of extra command-line options, which are
       processed before the real command line. You can use this to define
       standard search directories for include files, by putting `-i'
       options in the `NASMENV' variable.

       The value of the variable is split up at white space, so that the
       value `-s -ic:\nasmlib\' will be treated as two separate options.
       However, that means that the value `-dNAME="my name"' won't do what
       you might want, because it will be split at the space and the NASM
       command-line processing will get confused by the two nonsensical
       words `-dNAME="my' and `name"'.

       To get round this, NASM provides a feature whereby, if you begin the
       `NASMENV' environment variable with some character that isn't a
       minus sign, then NASM will treat this character as the separator
       character for options. So setting the `NASMENV' variable to the
       value `!-s!-ic:\nasmlib\' is equivalent to setting it to
       `-s -ic:\nasmlib\', but `!-dNAME="my name"' will work.

       This environment variable was previously called `NASM'. This was
       changed with version 0.98.31.

   2.2 Quick Start for MASM Users

       If you're used to writing programs with MASM, or with TASM in MASM-
       compatible (non-Ideal) mode, or with `a86', this section attempts to
       outline the major differences between MASM's syntax and NASM's. If
       you're not already used to MASM, it's probably worth skipping this
       section.

2.2.1 NASM Is Case-Sensitive

       One simple difference is that NASM is case-sensitive. It makes a
       difference whether you call your label `foo', `Foo' or `FOO'. If
       you're assembling to `DOS' or `OS/2' `.OBJ' files, you can invoke
       the `UPPERCASE' directive (documented in section 7.4) to ensure that
       all symbols exported to other code modules are forced to be upper
       case; but even then, _within_ a single module, NASM will distinguish
       between labels differing only in case.

2.2.2 NASM Requires Square Brackets For Memory References

       NASM was designed with simplicity of syntax in mind. One of the
       design goals of NASM is that it should be possible, as far as is
       practical, for the user to look at a single line of NASM code and
       tell what opcode is generated by it. You can't do this in MASM: if
       you declare, for example,

       foo     equ     1
       bar     dw      2

       then the two lines of code

               mov     ax,foo
               mov     ax,bar

       generate completely different opcodes, despite having identical-
       looking syntaxes.

       NASM avoids this undesirable situation by having a much simpler
       syntax for memory references. The rule is simply that any access to
       the _contents_ of a memory location requires square brackets around
       the address, and any access to the _address_ of a variable doesn't.
       So an instruction of the form `mov ax,foo' will _always_ refer to a
       compile-time constant, whether it's an `EQU' or the address of a
       variable; and to access the _contents_ of the variable `bar', you
       must code `mov ax,[bar]'.

       This also means that NASM has no need for MASM's `OFFSET' keyword,
       since the MASM code `mov ax,offset bar' means exactly the same thing
       as NASM's `mov ax,bar'. If you're trying to get large amounts of
       MASM code to assemble sensibly under NASM, you can always code
       `%idefine offset' to make the preprocessor treat the `OFFSET'
       keyword as a no-op.

       This issue is even more confusing in `a86', where declaring a label
       with a trailing colon defines it to be a `label' as opposed to a
       `variable' and causes `a86' to adopt NASM-style semantics; so in
       `a86', `mov ax,var' has different behaviour depending on whether
       `var' was declared as `var: dw 0' (a label) or `var dw 0' (a word-
       size variable). NASM is very simple by comparison: _everything_ is a
       label.

       NASM, in the interests of simplicity, also does not support the
       hybrid syntaxes supported by MASM and its clones, such as
       `mov ax,table[bx]', where a memory reference is denoted by one
       portion outside square brackets and another portion inside. The
       correct syntax for the above is `mov ax,[table+bx]'. Likewise,
       `mov ax,es:[di]' is wrong and `mov ax,[es:di]' is right.

2.2.3 NASM Doesn't Store Variable Types

       NASM, by design, chooses not to remember the types of variables you
       declare. Whereas MASM will remember, on seeing `var dw 0', that you
       declared `var' as a word-size variable, and will then be able to
       fill in the ambiguity in the size of the instruction `mov var,2',
       NASM will deliberately remember nothing about the symbol `var'
       except where it begins, and so you must explicitly code
       `mov word [var],2'.

       For this reason, NASM doesn't support the `LODS', `MOVS', `STOS',
       `SCAS', `CMPS', `INS', or `OUTS' instructions, but only supports the
       forms such as `LODSB', `MOVSW', and `SCASD', which explicitly
       specify the size of the components of the strings being manipulated.

2.2.4 NASM Doesn't `ASSUME'

       As part of NASM's drive for simplicity, it also does not support the
       `ASSUME' directive. NASM will not keep track of what values you
       choose to put in your segment registers, and will never
       _automatically_ generate a segment override prefix.

2.2.5 NASM Doesn't Support Memory Models

       NASM also does not have any directives to support different 16-bit
       memory models. The programmer has to keep track of which functions
       are supposed to be called with a far call and which with a near
       call, and is responsible for putting the correct form of `RET'
       instruction (`RETN' or `RETF'; NASM accepts `RET' itself as an
       alternate form for `RETN'); in addition, the programmer is
       responsible for coding CALL FAR instructions where necessary when
       calling _external_ functions, and must also keep track of which
       external variable definitions are far and which are near.

2.2.6 Floating-Point Differences

       NASM uses different names to refer to floating-point registers from
       MASM: where MASM would call them `ST(0)', `ST(1)' and so on, and
       `a86' would call them simply `0', `1' and so on, NASM chooses to
       call them `st0', `st1' etc.

       As of version 0.96, NASM now treats the instructions with `nowait'
       forms in the same way as MASM-compatible assemblers. The
       idiosyncratic treatment employed by 0.95 and earlier was based on a
       misunderstanding by the authors.

2.2.7 Other Differences

       For historical reasons, NASM uses the keyword `TWORD' where MASM and
       compatible assemblers use `TBYTE'.

       NASM does not declare uninitialized storage in the same way as MASM:
       where a MASM programmer might use `stack db 64 dup (?)', NASM
       requires `stack resb 64', intended to be read as `reserve 64 bytes'.
       For a limited amount of compatibility, since NASM treats `?' as a
       valid character in symbol names, you can code `? equ 0' and then
       writing `dw ?' will at least do something vaguely useful. `DUP' is
       still not a supported syntax, however.

       In addition to all of this, macros and directives work completely
       differently to MASM. See chapter 4 and chapter 6 for further
       details.

Chapter 3: The NASM Language
----------------------------

   3.1 Layout of a NASM Source Line

       Like most assemblers, each NASM source line contains (unless it is a
       macro, a preprocessor directive or an assembler directive: see
       chapter 4 and chapter 6) some combination of the four fields

       label:    instruction operands        ; comment

       As usual, most of these fields are optional; the presence or absence
       of any combination of a label, an instruction and a comment is
       allowed. Of course, the operand field is either required or
       forbidden by the presence and nature of the instruction field.

       NASM uses backslash (\) as the line continuation character; if a
       line ends with backslash, the next line is considered to be a part
       of the backslash-ended line.

       NASM places no restrictions on white space within a line: labels may
       have white space before them, or instructions may have no space
       before them, or anything. The colon after a label is also optional.
       (Note that this means that if you intend to code `lodsb' alone on a
       line, and type `lodab' by accident, then that's still a valid source
       line which does nothing but define a label. Running NASM with the
       command-line option `-w+orphan-labels' will cause it to warn you if
       you define a label alone on a line without a trailing colon.)

       Valid characters in labels are letters, numbers, `_', `$', `#', `@',
       `~', `.', and `?'. The only characters which may be used as the
       _first_ character of an identifier are letters, `.' (with special
       meaning: see section 3.9), `_' and `?'. An identifier may also be
       prefixed with a `$' to indicate that it is intended to be read as an
       identifier and not a reserved word; thus, if some other module you
       are linking with defines a symbol called `eax', you can refer to
       `$eax' in NASM code to distinguish the symbol from the register.
       Maximum length of an identifier is 4095 characters.

       The instruction field may contain any machine instruction: Pentium
       and P6 instructions, FPU instructions, MMX instructions and even
       undocumented instructions are all supported. The instruction may be
       prefixed by `LOCK', `REP', `REPE'/`REPZ', `REPNE'/`REPNZ',
       `XACQUIRE'/`XRELEASE' or `BND'/`NOBND', in the usual way. Explicit
       address-size and operand-size prefixes `A16', `A32', `A64', `O16'
       and `O32', `O64' are provided - one example of their use is given in
       chapter 10. You can also use the name of a segment register as an
       instruction prefix: coding `es mov [bx],ax' is equivalent to coding
       `mov [es:bx],ax'. We recommend the latter syntax, since it is
       consistent with other syntactic features of the language, but for
       instructions such as `LODSB', which has no operands and yet can
       require a segment override, there is no clean syntactic way to
       proceed apart from `es lodsb'.

       An instruction is not required to use a prefix: prefixes such as
       `CS', `A32', `LOCK' or `REPE' can appear on a line by themselves,
       and NASM will just generate the prefix bytes.

       In addition to actual machine instructions, NASM also supports a
       number of pseudo-instructions, described in section 3.2.

       Instruction operands may take a number of forms: they can be
       registers, described simply by the register name (e.g. `ax', `bp',
       `ebx', `cr0': NASM does not use the `gas'-style syntax in which
       register names must be prefixed by a `%' sign), or they can be
       effective addresses (see section 3.3), constants (section 3.4) or
       expressions (section 3.5).

       For x87 floating-point instructions, NASM accepts a wide range of
       syntaxes: you can use two-operand forms like MASM supports, or you
       can use NASM's native single-operand forms in most cases. For
       example, you can code:

               fadd    st1             ; this sets st0 := st0 + st1
               fadd    st0,st1         ; so does this
      
               fadd    st1,st0         ; this sets st1 := st1 + st0
               fadd    to st1          ; so does this

       Almost any x87 floating-point instruction that references memory
       must use one of the prefixes `DWORD', `QWORD' or `TWORD' to indicate
       what size of memory operand it refers to.

   3.2 Pseudo-Instructions

       Pseudo-instructions are things which, though not real x86 machine
       instructions, are used in the instruction field anyway because
       that's the most convenient place to put them. The current pseudo-
       instructions are `DB', `DW', `DD', `DQ', `DT', `DO', `DY' and `DZ';
       their uninitialized counterparts `RESB', `RESW', `RESD', `RESQ',
       `REST', `RESO', `RESY' and `RESZ'; the `INCBIN' command, the `EQU'
       command, and the `TIMES' prefix.

3.2.1 `DB' and Friends: Declaring Initialized Data

       `DB', `DW', `DD', `DQ', `DT', `DO', `DY' and `DZ' are used, much as
       in MASM, to declare initialized data in the output file. They can be
       invoked in a wide range of ways:

             db    0x55                ; just the byte 0x55
             db    0x55,0x56,0x57      ; three bytes in succession
             db    'a',0x55            ; character constants are OK
             db    'hello',13,10,'$'   ; so are string constants
             dw    0x1234              ; 0x34 0x12
             dw    'a'                 ; 0x61 0x00 (it's just a number)
             dw    'ab'                ; 0x61 0x62 (character constant)
             dw    'abc'               ; 0x61 0x62 0x63 0x00 (string)
             dd    0x12345678          ; 0x78 0x56 0x34 0x12
             dd    1.234567e20         ; floating-point constant
             dq    0x123456789abcdef0  ; eight byte constant
             dq    1.234567e20         ; double-precision float
             dt    1.234567e20         ; extended-precision float

       `DT', `DO', `DY' and `DZ' do not accept numeric constants as
       operands.

3.2.2 `RESB' and Friends: Declaring Uninitialized Data

       `RESB', `RESW', `RESD', `RESQ', `REST', `RESO', `RESY' and `RESZ'
       are designed to be used in the BSS section of a module: they declare
       _uninitialized_ storage space. Each takes a single operand, which is
       the number of bytes, words, doublewords or whatever to reserve. As
       stated in section 2.2.7, NASM does not support the MASM/TASM syntax
       of reserving uninitialized space by writing `DW ?' or similar
       things: this is what it does instead. The operand to a `RESB'-type
       pseudo-instruction is a _critical expression_: see section 3.8.

       For example:

       buffer:         resb    64              ; reserve 64 bytes
       wordvar:        resw    1               ; reserve a word
       realarray       resq    10              ; array of ten reals
       ymmval:         resy    1               ; one YMM register
       zmmvals:        resz    32              ; 32 ZMM registers

3.2.3 `INCBIN': Including External Binary Files

       `INCBIN' is borrowed from the old Amiga assembler DevPac: it
       includes a binary file verbatim into the output file. This can be
       handy for (for example) including graphics and sound data directly
       into a game executable file. It can be called in one of these three
       ways:

           incbin  "file.dat"             ; include the whole file
           incbin  "file.dat",1024        ; skip the first 1024 bytes
           incbin  "file.dat",1024,512    ; skip the first 1024, and
                                          ; actually include at most 512

       `INCBIN' is both a directive and a standard macro; the standard
       macro version searches for the file in the include file search path
       and adds the file to the dependency lists. This macro can be
       overridden if desired.

3.2.4 `EQU': Defining Constants

       `EQU' defines a symbol to a given constant value: when `EQU' is
       used, the source line must contain a label. The action of `EQU' is
       to define the given label name to the value of its (only) operand.
       This definition is absolute, and cannot change later. So, for
       example,

       message         db      'hello, world'
       msglen          equ     $-message

       defines `msglen' to be the constant 12. `msglen' may not then be
       redefined later. This is not a preprocessor definition either: the
       value of `msglen' is evaluated _once_, using the value of `$' (see
       section 3.5 for an explanation of `$') at the point of definition,
       rather than being evaluated wherever it is referenced and using the
       value of `$' at the point of reference.

3.2.5 `TIMES': Repeating Instructions or Data

       The `TIMES' prefix causes the instruction to be assembled multiple
       times. This is partly present as NASM's equivalent of the `DUP'
       syntax supported by MASM-compatible assemblers, in that you can code

       zerobuf:        times 64 db 0

       or similar things; but `TIMES' is more versatile than that. The
       argument to `TIMES' is not just a numeric constant, but a numeric
       _expression_, so you can do things like

       buffer: db      'hello, world'
               times 64-$+buffer db ' '

       which will store exactly enough spaces to make the total length of
       `buffer' up to 64. Finally, `TIMES' can be applied to ordinary
       instructions, so you can code trivial unrolled loops in it:

               times 100 movsb

       Note that there is no effective difference between
       `times 100 resb 1' and `resb 100', except that the latter will be
       assembled about 100 times faster due to the internal structure of
       the assembler.

       The operand to `TIMES' is a critical expression (section 3.8).

       Note also that `TIMES' can't be applied to macros: the reason for
       this is that `TIMES' is processed after the macro phase, which
       allows the argument to `TIMES' to contain expressions such as
       `64-$+buffer' as above. To repeat more than one line of code, or a
       complex macro, use the preprocessor `%rep' directive.

   3.3 Effective Addresses

       An effective address is any operand to an instruction which
       references memory. Effective addresses, in NASM, have a very simple
       syntax: they consist of an expression evaluating to the desired
       address, enclosed in square brackets. For example:

       wordvar dw      123
               mov     ax,[wordvar]
               mov     ax,[wordvar+1]
               mov     ax,[es:wordvar+bx]

       Anything not conforming to this simple system is not a valid memory
       reference in NASM, for example `es:wordvar[bx]'.

       More complicated effective addresses, such as those involving more
       than one register, work in exactly the same way:

               mov     eax,[ebx*2+ecx+offset]
               mov     ax,[bp+di+8]

       NASM is capable of doing algebra on these effective addresses, so
       that things which don't necessarily _look_ legal are perfectly all
       right:

           mov     eax,[ebx*5]             ; assembles as [ebx*4+ebx]
           mov     eax,[label1*2-label2]   ; ie [label1+(label1-label2)]

       Some forms of effective address have more than one assembled form;
       in most such cases NASM will generate the smallest form it can. For
       example, there are distinct assembled forms for the 32-bit effective
       addresses `[eax*2+0]' and `[eax+eax]', and NASM will generally
       generate the latter on the grounds that the former requires four
       bytes to store a zero offset.

       NASM has a hinting mechanism which will cause `[eax+ebx]' and
       `[ebx+eax]' to generate different opcodes; this is occasionally
       useful because `[esi+ebp]' and `[ebp+esi]' have different default
       segment registers.

       However, you can force NASM to generate an effective address in a
       particular form by the use of the keywords `BYTE', `WORD', `DWORD'
       and `NOSPLIT'. If you need `[eax+3]' to be assembled using a double-
       word offset field instead of the one byte NASM will normally
       generate, you can code `[dword eax+3]'. Similarly, you can force
       NASM to use a byte offset for a small value which it hasn't seen on
       the first pass (see section 3.8 for an example of such a code
       fragment) by using `[byte eax+offset]'. As special cases,
       `[byte eax]' will code `[eax+0]' with a byte offset of zero, and
       `[dword eax]' will code it with a double-word offset of zero. The
       normal form, `[eax]', will be coded with no offset field.

       The form described in the previous paragraph is also useful if you
       are trying to access data in a 32-bit segment from within 16 bit
       code. For more information on this see the section on mixed-size
       addressing (section 10.2). In particular, if you need to access data
       with a known offset that is larger than will fit in a 16-bit value,
       if you don't specify that it is a dword offset, nasm will cause the
       high word of the offset to be lost.

       Similarly, NASM will split `[eax*2]' into `[eax+eax]' because that
       allows the offset field to be absent and space to be saved; in fact,
       it will also split `[eax*2+offset]' into `[eax+eax+offset]'. You can
       combat this behaviour by the use of the `NOSPLIT' keyword:
       `[nosplit eax*2]' will force `[eax*2+0]' to be generated literally.
       `[nosplit eax*1]' also has the same effect. In another way, a split
       EA form `[0, eax*2]' can be used, too. However, `NOSPLIT' in
       `[nosplit eax+eax]' will be ignored because user's intention here is
       considered as `[eax+eax]'.

       In 64-bit mode, NASM will by default generate absolute addresses.
       The `REL' keyword makes it produce `RIP'-relative addresses. Since
       this is frequently the normally desired behaviour, see the `DEFAULT'
       directive (section 6.2). The keyword `ABS' overrides `REL'.

       A new form of split effective addres syntax is also supported. This
       is mainly intended for mib operands as used by MPX instructions, but
       can be used for any memory reference. The basic concept of this form
       is splitting base and index.

            mov eax,[ebx+8,ecx*4]   ; ebx=base, ecx=index, 4=scale, 8=disp

       For mib operands, there are several ways of writing effective
       address depending on the tools. NASM supports all currently possible
       ways of mib syntax:

            ; bndstx
            ; next 5 lines are parsed same
            ; base=rax, index=rbx, scale=1, displacement=3
            bndstx [rax+0x3,rbx], bnd0      ; NASM - split EA
            bndstx [rbx*1+rax+0x3], bnd0    ; GAS - '*1' indecates an index reg
            bndstx [rax+rbx+3], bnd0        ; GAS - without hints
            bndstx [rax+0x3], bnd0, rbx     ; ICC-1
            bndstx [rax+0x3], rbx, bnd0     ; ICC-2

       When broadcasting decorator is used, the opsize keyword should match
       the size of each element.

            VDIVPS zmm4, zmm5, dword [rbx]{1to16}   ; single-precision float
            VDIVPS zmm4, zmm5, zword [rbx]          ; packed 512 bit memory

   3.4 Constants

       NASM understands four different types of constant: numeric,
       character, string and floating-point.

3.4.1 Numeric Constants

       A numeric constant is simply a number. NASM allows you to specify
       numbers in a variety of number bases, in a variety of ways: you can
       suffix `H' or `X', `D' or `T', `Q' or `O', and `B' or `Y' for
       hexadecimal, decimal, octal and binary respectively, or you can
       prefix `0x', for hexadecimal in the style of C, or you can prefix
       `$' for hexadecimal in the style of Borland Pascal or Motorola
       Assemblers. Note, though, that the `$' prefix does double duty as a
       prefix on identifiers (see section 3.1), so a hex number prefixed
       with a `$' sign must have a digit after the `$' rather than a
       letter. In addition, current versions of NASM accept the prefix `0h'
       for hexadecimal, `0d' or `0t' for decimal, `0o' or `0q' for octal,
       and `0b' or `0y' for binary. Please note that unlike C, a `0' prefix
       by itself does _not_ imply an octal constant!

       Numeric constants can have underscores (`_') interspersed to break
       up long strings.

       Some examples (all producing exactly the same code):

               mov     ax,200          ; decimal
               mov     ax,0200         ; still decimal
               mov     ax,0200d        ; explicitly decimal
               mov     ax,0d200        ; also decimal
               mov     ax,0c8h         ; hex
               mov     ax,$0c8         ; hex again: the 0 is required
               mov     ax,0xc8         ; hex yet again
               mov     ax,0hc8         ; still hex
               mov     ax,310q         ; octal
               mov     ax,310o         ; octal again
               mov     ax,0o310        ; octal yet again
               mov     ax,0q310        ; octal yet again
               mov     ax,11001000b    ; binary
               mov     ax,1100_1000b   ; same binary constant
               mov     ax,1100_1000y   ; same binary constant once more
               mov     ax,0b1100_1000  ; same binary constant yet again
               mov     ax,0y1100_1000  ; same binary constant yet again

3.4.2 Character Strings

       A character string consists of up to eight characters enclosed in
       either single quotes (`'...''), double quotes (`"..."') or
       backquotes (``...`'). Single or double quotes are equivalent to NASM
       (except of course that surrounding the constant with single quotes
       allows double quotes to appear within it and vice versa); the
       contents of those are represented verbatim. Strings enclosed in
       backquotes support C-style `\'-escapes for special characters.

       The following escape sequences are recognized by backquoted strings:

             \'          single quote (')
             \"          double quote (")
             \`          backquote (`)
             \\          backslash (\)
             \?          question mark (?)
             \a          BEL (ASCII 7)
             \b          BS  (ASCII 8)
             \t          TAB (ASCII 9)
             \n          LF  (ASCII 10)
             \v          VT  (ASCII 11)
             \f          FF  (ASCII 12)
             \r          CR  (ASCII 13)
             \e          ESC (ASCII 27)
             \377        Up to 3 octal digits - literal byte
             \xFF        Up to 2 hexadecimal digits - literal byte
             \u1234      4 hexadecimal digits - Unicode character
             \U12345678  8 hexadecimal digits - Unicode character

       All other escape sequences are reserved. Note that `\0', meaning a
       `NUL' character (ASCII 0), is a special case of the octal escape
       sequence.

       Unicode characters specified with `\u' or `\U' are converted to
       UTF-8. For example, the following lines are all equivalent:

             db `\u263a`            ; UTF-8 smiley face
             db `\xe2\x98\xba`      ; UTF-8 smiley face
             db 0E2h, 098h, 0BAh    ; UTF-8 smiley face

3.4.3 Character Constants

       A character constant consists of a string up to eight bytes long,
       used in an expression context. It is treated as if it was an
       integer.

       A character constant with more than one byte will be arranged with
       little-endian order in mind: if you code

                 mov eax,'abcd'

       then the constant generated is not `0x61626364', but `0x64636261',
       so that if you were then to store the value into memory, it would
       read `abcd' rather than `dcba'. This is also the sense of character
       constants understood by the Pentium's `CPUID' instruction.

3.4.4 String Constants

       String constants are character strings used in the context of some
       pseudo-instructions, namely the `DB' family and `INCBIN' (where it
       represents a filename.) They are also used in certain preprocessor
       directives.

       A string constant looks like a character constant, only longer. It
       is treated as a concatenation of maximum-size character constants
       for the conditions. So the following are equivalent:

             db    'hello'               ; string constant
             db    'h','e','l','l','o'   ; equivalent character constants

       And the following are also equivalent:

             dd    'ninechars'           ; doubleword string constant
             dd    'nine','char','s'     ; becomes three doublewords
             db    'ninechars',0,0,0     ; and really looks like this

       Note that when used in a string-supporting context, quoted strings
       are treated as a string constants even if they are short enough to
       be a character constant, because otherwise `db 'ab'' would have the
       same effect as `db 'a'', which would be silly. Similarly, three-
       character or four-character constants are treated as strings when
       they are operands to `DW', and so forth.

3.4.5 Unicode Strings

       The special operators `__utf16__', `__utf16le__', `__utf16be__',
       `__utf32__', `__utf32le__' and `__utf32be__' allows definition of
       Unicode strings. They take a string in UTF-8 format and converts it
       to UTF-16 or UTF-32, respectively. Unless the `be' forms are
       specified, the output is littleendian.

       For example:

       %define u(x) __utf16__(x)
       %define w(x) __utf32__(x)
      
             dw u('C:\WINDOWS'), 0       ; Pathname in UTF-16
             dd w(`A + B = \u206a`), 0   ; String in UTF-32

       The UTF operators can be applied either to strings passed to the
       `DB' family instructions, or to character constants in an expression
       context.

3.4.6 Floating-Point Constants

       Floating-point constants are acceptable only as arguments to `DB',
       `DW', `DD', `DQ', `DT', and `DO', or as arguments to the special
       operators `__float8__', `__float16__', `__float32__', `__float64__',
       `__float80m__', `__float80e__', `__float128l__', and
       `__float128h__'.

       Floating-point constants are expressed in the traditional form:
       digits, then a period, then optionally more digits, then optionally
       an `E' followed by an exponent. The period is mandatory, so that
       NASM can distinguish between `dd 1', which declares an integer
       constant, and `dd 1.0' which declares a floating-point constant.

       NASM also support C99-style hexadecimal floating-point: `0x',
       hexadecimal digits, period, optionally more hexadeximal digits, then
       optionally a `P' followed by a _binary_ (not hexadecimal) exponent
       in decimal notation. As an extension, NASM additionally supports the
       `0h' and `$' prefixes for hexadecimal, as well binary and octal
       floating-point, using the `0b' or `0y' and `0o' or `0q' prefixes,
       respectively.

       Underscores to break up groups of digits are permitted in floating-
       point constants as well.

       Some examples:

             db    -0.2                    ; "Quarter precision"
             dw    -0.5                    ; IEEE 754r/SSE5 half precision
             dd    1.2                     ; an easy one
             dd    1.222_222_222           ; underscores are permitted
             dd    0x1p+2                  ; 1.0x2^2 = 4.0
             dq    0x1p+32                 ; 1.0x2^32 = 4 294 967 296.0
             dq    1.e10                   ; 10 000 000 000.0
             dq    1.e+10                  ; synonymous with 1.e10
             dq    1.e-10                  ; 0.000 000 000 1
             dt    3.141592653589793238462 ; pi
             do    1.e+4000                ; IEEE 754r quad precision

       The 8-bit "quarter-precision" floating-point format is
       sign:exponent:mantissa = 1:4:3 with an exponent bias of 7. This
       appears to be the most frequently used 8-bit floating-point format,
       although it is not covered by any formal standard. This is sometimes
       called a "minifloat."

       The special operators are used to produce floating-point numbers in
       other contexts. They produce the binary representation of a specific
       floating-point number as an integer, and can use anywhere integer
       constants are used in an expression. `__float80m__' and
       `__float80e__' produce the 64-bit mantissa and 16-bit exponent of an
       80-bit floating-point number, and `__float128l__' and
       `__float128h__' produce the lower and upper 64-bit halves of a 128-
       bit floating-point number, respectively.

       For example:

             mov    rax,__float64__(3.141592653589793238462)

       ... would assign the binary representation of pi as a 64-bit
       floating point number into `RAX'. This is exactly equivalent to:

             mov    rax,0x400921fb54442d18

       NASM cannot do compile-time arithmetic on floating-point constants.
       This is because NASM is designed to be portable - although it always
       generates code to run on x86 processors, the assembler itself can
       run on any system with an ANSI C compiler. Therefore, the assembler
       cannot guarantee the presence of a floating-point unit capable of
       handling the Intel number formats, and so for NASM to be able to do
       floating arithmetic it would have to include its own complete set of
       floating-point routines, which would significantly increase the size
       of the assembler for very little benefit.

       The special tokens `__Infinity__', `__QNaN__' (or `__NaN__') and
       `__SNaN__' can be used to generate infinities, quiet NaNs, and
       signalling NaNs, respectively. These are normally used as macros:

       %define Inf __Infinity__
       %define NaN __QNaN__
      
             dq    +1.5, -Inf, NaN         ; Double-precision constants

       The `%use fp' standard macro package contains a set of convenience
       macros. See section 5.3.

3.4.7 Packed BCD Constants

       x87-style packed BCD constants can be used in the same contexts as
       80-bit floating-point numbers. They are suffixed with `p' or
       prefixed with `0p', and can include up to 18 decimal digits.

       As with other numeric constants, underscores can be used to separate
       digits.

       For example:

             dt 12_345_678_901_245_678p
             dt -12_345_678_901_245_678p
             dt +0p33
             dt 33p

   3.5 Expressions

       Expressions in NASM are similar in syntax to those in C. Expressions
       are evaluated as 64-bit integers which are then adjusted to the
       appropriate size.

       NASM supports two special tokens in expressions, allowing
       calculations to involve the current assembly position: the `$' and
       `$$' tokens. `$' evaluates to the assembly position at the beginning
       of the line containing the expression; so you can code an infinite
       loop using `JMP $'. `$$' evaluates to the beginning of the current
       section; so you can tell how far into the section you are by using
       `($-$$)'.

       The arithmetic operators provided by NASM are listed here, in
       increasing order of precedence.

3.5.1 `|': Bitwise OR Operator

       The `|' operator gives a bitwise OR, exactly as performed by the
       `OR' machine instruction. Bitwise OR is the lowest-priority
       arithmetic operator supported by NASM.

3.5.2 `^': Bitwise XOR Operator

       `^' provides the bitwise XOR operation.

3.5.3 `&': Bitwise AND Operator

       `&' provides the bitwise AND operation.

3.5.4 `<<' and `>>': Bit Shift Operators

       `<<' gives a bit-shift to the left, just as it does in C. So `5<<3'
       evaluates to 5 times 8, or 40. `>>' gives a bit-shift to the right;
       in NASM, such a shift is _always_ unsigned, so that the bits shifted
       in from the left-hand end are filled with zero rather than a sign-
       extension of the previous highest bit.

3.5.5 `+' and `-': Addition and Subtraction Operators

       The `+' and `-' operators do perfectly ordinary addition and
       subtraction.

3.5.6 `*', `/', `//', `%' and `%%': Multiplication and Division

       `*' is the multiplication operator. `/' and `//' are both division
       operators: `/' is unsigned division and `//' is signed division.
       Similarly, `%' and `%%' provide unsigned and signed modulo operators
       respectively.

       NASM, like ANSI C, provides no guarantees about the sensible
       operation of the signed modulo operator.

       Since the `%' character is used extensively by the macro
       preprocessor, you should ensure that both the signed and unsigned
       modulo operators are followed by white space wherever they appear.

3.5.7 Unary Operators

       The highest-priority operators in NASM's expression grammar are
       those which only apply to one argument. These are `+', `-', `~',
       `!', `SEG', and the integer functions operators.

       `-' negates its operand, `+' does nothing (it's provided for
       symmetry with `-'), `~' computes the one's complement of its
       operand, `!' is the logical negation operator.

       `SEG' provides the segment address of its operand (explained in more
       detail in section 3.6).

       A set of additional operators with leading and trailing double
       underscores are used to implement the integer functions of the
       `ifunc' macro package, see section 5.4.

   3.6 `SEG' and `WRT'

       When writing large 16-bit programs, which must be split into
       multiple segments, it is often necessary to be able to refer to the
       segment part of the address of a symbol. NASM supports the `SEG'
       operator to perform this function.

       The `SEG' operator returns the _preferred_ segment base of a symbol,
       defined as the segment base relative to which the offset of the
       symbol makes sense. So the code

               mov     ax,seg symbol
               mov     es,ax
               mov     bx,symbol

       will load `ES:BX' with a valid pointer to the symbol `symbol'.

       Things can be more complex than this: since 16-bit segments and
       groups may overlap, you might occasionally want to refer to some
       symbol using a different segment base from the preferred one. NASM
       lets you do this, by the use of the `WRT' (With Reference To)
       keyword. So you can do things like

               mov     ax,weird_seg        ; weird_seg is a segment base
               mov     es,ax
               mov     bx,symbol wrt weird_seg

       to load `ES:BX' with a different, but functionally equivalent,
       pointer to the symbol `symbol'.

       NASM supports far (inter-segment) calls and jumps by means of the
       syntax `call segment:offset', where `segment' and `offset' both
       represent immediate values. So to call a far procedure, you could
       code either of

               call    (seg procedure):procedure
               call    weird_seg:(procedure wrt weird_seg)

       (The parentheses are included for clarity, to show the intended
       parsing of the above instructions. They are not necessary in
       practice.)

       NASM supports the syntax `call far procedure' as a synonym for the
       first of the above usages. `JMP' works identically to `CALL' in
       these examples.

       To declare a far pointer to a data item in a data segment, you must
       code

               dw      symbol, seg symbol

       NASM supports no convenient synonym for this, though you can always
       invent one using the macro processor.

   3.7 `STRICT': Inhibiting Optimization

       When assembling with the optimizer set to level 2 or higher (see
       section 2.1.22), NASM will use size specifiers (`BYTE', `WORD',
       `DWORD', `QWORD', `TWORD', `OWORD', `YWORD' or `ZWORD'), but will
       give them the smallest possible size. The keyword `STRICT' can be
       used to inhibit optimization and force a particular operand to be
       emitted in the specified size. For example, with the optimizer on,
       and in `BITS 16' mode,

               push dword 33

       is encoded in three bytes `66 6A 21', whereas

               push strict dword 33

       is encoded in six bytes, with a full dword immediate operand
       `66 68 21 00 00 00'.

       With the optimizer off, the same code (six bytes) is generated
       whether the `STRICT' keyword was used or not.

   3.8 Critical Expressions

       Although NASM has an optional multi-pass optimizer, there are some
       expressions which must be resolvable on the first pass. These are
       called _Critical Expressions_.

       The first pass is used to determine the size of all the assembled
       code and data, so that the second pass, when generating all the
       code, knows all the symbol addresses the code refers to. So one
       thing NASM can't handle is code whose size depends on the value of a
       symbol declared after the code in question. For example,

               times (label-$) db 0
       label:  db      'Where am I?'

       The argument to `TIMES' in this case could equally legally evaluate
       to anything at all; NASM will reject this example because it cannot
       tell the size of the `TIMES' line when it first sees it. It will
       just as firmly reject the slightly paradoxical code

               times (label-$+1) db 0
       label:  db      'NOW where am I?'

       in which _any_ value for the `TIMES' argument is by definition
       wrong!

       NASM rejects these examples by means of a concept called a _critical
       expression_, which is defined to be an expression whose value is
       required to be computable in the first pass, and which must
       therefore depend only on symbols defined before it. The argument to
       the `TIMES' prefix is a critical expression.

   3.9 Local Labels

       NASM gives special treatment to symbols beginning with a period. A
       label beginning with a single period is treated as a _local_ label,
       which means that it is associated with the previous non-local label.
       So, for example:

       label1  ; some code
      
       .loop
               ; some more code
      
               jne     .loop
               ret
      
       label2  ; some code
      
       .loop
               ; some more code
      
               jne     .loop
               ret

       In the above code fragment, each `JNE' instruction jumps to the line
       immediately before it, because the two definitions of `.loop' are
       kept separate by virtue of each being associated with the previous
       non-local label.

       This form of local label handling is borrowed from the old Amiga
       assembler DevPac; however, NASM goes one step further, in allowing
       access to local labels from other parts of the code. This is
       achieved by means of _defining_ a local label in terms of the
       previous non-local label: the first definition of `.loop' above is
       really defining a symbol called `label1.loop', and the second
       defines a symbol called `label2.loop'. So, if you really needed to,
       you could write

       label3  ; some more code
               ; and some more
      
               jmp label1.loop

       Sometimes it is useful - in a macro, for instance - to be able to
       define a label which can be referenced from anywhere but which
       doesn't interfere with the normal local-label mechanism. Such a
       label can't be non-local because it would interfere with subsequent
       definitions of, and references to, local labels; and it can't be
       local because the macro that defined it wouldn't know the label's
       full name. NASM therefore introduces a third type of label, which is
       probably only useful in macro definitions: if a label begins with
       the special prefix `..@', then it does nothing to the local label
       mechanism. So you could code

       label1:                         ; a non-local label
       .local:                         ; this is really label1.local
       ..@foo:                         ; this is a special symbol
       label2:                         ; another non-local label
       .local:                         ; this is really label2.local
      
               jmp     ..@foo          ; this will jump three lines up

       NASM has the capacity to define other special symbols beginning with
       a double period: for example, `..start' is used to specify the entry
       point in the `obj' output format (see section 7.4.6), `..imagebase'
       is used to find out the offset from a base address of the current
       image in the `win64' output format (see section 7.6.1). So just keep
       in mind that symbols beginning with a double period are special.

Chapter 4: The NASM Preprocessor
--------------------------------

       NASM contains a powerful macro processor, which supports conditional
       assembly, multi-level file inclusion, two forms of macro (single-
       line and multi-line), and a `context stack' mechanism for extra
       macro power. Preprocessor directives all begin with a `%' sign.

       The preprocessor collapses all lines which end with a backslash (\)
       character into a single line. Thus:

       %define THIS_VERY_LONG_MACRO_NAME_IS_DEFINED_TO \
               THIS_VALUE

       will work like a single-line macro without the backslash-newline
       sequence.

   4.1 Single-Line Macros

4.1.1 The Normal Way: `%define'

       Single-line macros are defined using the `%define' preprocessor
       directive. The definitions work in a similar way to C; so you can do
       things like

       %define ctrl    0x1F &
       %define param(a,b) ((a)+(a)*(b))
      
               mov     byte [param(2,ebx)], ctrl 'D'

       which will expand to

               mov     byte [(2)+(2)*(ebx)], 0x1F & 'D'

       When the expansion of a single-line macro contains tokens which
       invoke another macro, the expansion is performed at invocation time,
       not at definition time. Thus the code

       %define a(x)    1+b(x)
       %define b(x)    2*x
      
               mov     ax,a(8)

       will evaluate in the expected way to `mov ax,1+2*8', even though the
       macro `b' wasn't defined at the time of definition of `a'.

       Macros defined with `%define' are case sensitive: after
       `%define foo bar', only `foo' will expand to `bar': `Foo' or `FOO'
       will not. By using `%idefine' instead of `%define' (the `i' stands
       for `insensitive') you can define all the case variants of a macro
       at once, so that `%idefine foo bar' would cause `foo', `Foo', `FOO',
       `fOO' and so on all to expand to `bar'.

       There is a mechanism which detects when a macro call has occurred as
       a result of a previous expansion of the same macro, to guard against
       circular references and infinite loops. If this happens, the
       preprocessor will only expand the first occurrence of the macro.
       Hence, if you code

       %define a(x)    1+a(x)
      
               mov     ax,a(3)

       the macro `a(3)' will expand once, becoming `1+a(3)', and will then
       expand no further. This behaviour can be useful: see section 9.1 for
       an example of its use.

       You can overload single-line macros: if you write

       %define foo(x)   1+x
       %define foo(x,y) 1+x*y

       the preprocessor will be able to handle both types of macro call, by
       counting the parameters you pass; so `foo(3)' will become `1+3'
       whereas `foo(ebx,2)' will become `1+ebx*2'. However, if you define

       %define foo bar

       then no other definition of `foo' will be accepted: a macro with no
       parameters prohibits the definition of the same name as a macro
       _with_ parameters, and vice versa.

       This doesn't prevent single-line macros being _redefined_: you can
       perfectly well define a macro with

       %define foo bar

       and then re-define it later in the same source file with

       %define foo baz

       Then everywhere the macro `foo' is invoked, it will be expanded
       according to the most recent definition. This is particularly useful
       when defining single-line macros with `%assign' (see section 4.1.7).

       You can pre-define single-line macros using the `-d' option on the
       NASM command line: see section 2.1.18.

4.1.2 Resolving `%define': `%xdefine'

       To have a reference to an embedded single-line macro resolved at the
       time that the embedding macro is _defined_, as opposed to when the
       embedding macro is _expanded_, you need a different mechanism to the
       one offered by `%define'. The solution is to use `%xdefine', or it's
       case-insensitive counterpart `%ixdefine'.

       Suppose you have the following code:

       %define  isTrue  1
       %define  isFalse isTrue
       %define  isTrue  0
      
       val1:    db      isFalse
      
       %define  isTrue  1
      
       val2:    db      isFalse

       In this case, `val1' is equal to 0, and `val2' is equal to 1. This
       is because, when a single-line macro is defined using `%define', it
       is expanded only when it is called. As `isFalse' expands to
       `isTrue', the expansion will be the current value of `isTrue'. The
       first time it is called that is 0, and the second time it is 1.

       If you wanted `isFalse' to expand to the value assigned to the
       embedded macro `isTrue' at the time that `isFalse' was defined, you
       need to change the above code to use `%xdefine'.

       %xdefine isTrue  1
       %xdefine isFalse isTrue
       %xdefine isTrue  0
      
       val1:    db      isFalse
      
       %xdefine isTrue  1
      
       val2:    db      isFalse

       Now, each time that `isFalse' is called, it expands to 1, as that is
       what the embedded macro `isTrue' expanded to at the time that
       `isFalse' was defined.

4.1.3 Macro Indirection: `%[...]'

       The `%[...]' construct can be used to expand macros in contexts
       where macro expansion would otherwise not occur, including in the
       names other macros. For example, if you have a set of macros named
       `Foo16', `Foo32' and `Foo64', you could write:

            mov ax,Foo%[__BITS__]   ; The Foo value

       to use the builtin macro `__BITS__' (see section 4.12.5) to
       automatically select between them. Similarly, the two statements:

       %xdefine Bar         Quux    ; Expands due to %xdefine
       %define  Bar         %[Quux] ; Expands due to %[...]

       have, in fact, exactly the same effect.

       `%[...]' concatenates to adjacent tokens in the same way that multi-
       line macro parameters do, see section 4.3.9 for details.

4.1.4 Concatenating Single Line Macro Tokens: `%+'

       Individual tokens in single line macros can be concatenated, to
       produce longer tokens for later processing. This can be useful if
       there are several similar macros that perform similar functions.

       Please note that a space is required after `%+', in order to
       disambiguate it from the syntax `%+1' used in multiline macros.

       As an example, consider the following:

       %define BDASTART 400h                ; Start of BIOS data area

       struc   tBIOSDA                      ; its structure
               .COM1addr       RESW    1
               .COM2addr       RESW    1
               ; ..and so on
       endstruc

       Now, if we need to access the elements of tBIOSDA in different
       places, we can end up with:

               mov     ax,BDASTART + tBIOSDA.COM1addr
               mov     bx,BDASTART + tBIOSDA.COM2addr

       This will become pretty ugly (and tedious) if used in many places,
       and can be reduced in size significantly by using the following
       macro:

       ; Macro to access BIOS variables by their names (from tBDA):

       %define BDA(x)  BDASTART + tBIOSDA. %+ x

       Now the above code can be written as:

               mov     ax,BDA(COM1addr)
               mov     bx,BDA(COM2addr)

       Using this feature, we can simplify references to a lot of macros
       (and, in turn, reduce typing errors).

4.1.5 The Macro Name Itself: `%?' and `%??'

       The special symbols `%?' and `%??' can be used to reference the
       macro name itself inside a macro expansion, this is supported for
       both single-and multi-line macros. `%?' refers to the macro name as
       _invoked_, whereas `%??' refers to the macro name as _declared_. The
       two are always the same for case-sensitive macros, but for case-
       insensitive macros, they can differ.

       For example:

       %idefine Foo mov %?,%??
      
               foo
               FOO

       will expand to:

               mov foo,Foo
               mov FOO,Foo

       The sequence:

       %idefine keyword $%?

       can be used to make a keyword "disappear", for example in case a new
       instruction has been used as a label in older code. For example:

       %idefine pause $%?                  ; Hide the PAUSE instruction

4.1.6 Undefining Single-Line Macros: `%undef'

       Single-line macros can be removed with the `%undef' directive. For
       example, the following sequence:

       %define foo bar
       %undef  foo
      
               mov     eax, foo

       will expand to the instruction `mov eax, foo', since after `%undef'
       the macro `foo' is no longer defined.

       Macros that would otherwise be pre-defined can be undefined on the
       command-line using the `-u' option on the NASM command line: see
       section 2.1.19.

4.1.7 Preprocessor Variables: `%assign'

       An alternative way to define single-line macros is by means of the
       `%assign' command (and its case-insensitive counterpart `%iassign',
       which differs from `%assign' in exactly the same way that `%idefine'
       differs from `%define').

       `%assign' is used to define single-line macros which take no
       parameters and have a numeric value. This value can be specified in
       the form of an expression, and it will be evaluated once, when the
       `%assign' directive is processed.

       Like `%define', macros defined using `%assign' can be re-defined
       later, so you can do things like

       %assign i i+1

       to increment the numeric value of a macro.

       `%assign' is useful for controlling the termination of `%rep'
       preprocessor loops: see section 4.5 for an example of this. Another
       use for `%assign' is given in section 8.4 and section 9.1.

       The expression passed to `%assign' is a critical expression (see
       section 3.8), and must also evaluate to a pure number (rather than a
       relocatable reference such as a code or data address, or anything
       involving a register).

4.1.8 Defining Strings: `%defstr'

       `%defstr', and its case-insensitive counterpart `%idefstr', define
       or redefine a single-line macro without parameters but converts the
       entire right-hand side, after macro expansion, to a quoted string
       before definition.

       For example:

       %defstr test TEST

       is equivalent to

       %define test 'TEST'

       This can be used, for example, with the `%!' construct (see section
       4.10.2):

       %defstr PATH %!PATH          ; The operating system PATH variable

4.1.9 Defining Tokens: `%deftok'

       `%deftok', and its case-insensitive counterpart `%ideftok', define
       or redefine a single-line macro without parameters but converts the
       second parameter, after string conversion, to a sequence of tokens.

       For example:

       %deftok test 'TEST'

       is equivalent to

       %define test TEST

   4.2 String Manipulation in Macros

       It's often useful to be able to handle strings in macros. NASM
       supports a few simple string handling macro operators from which
       more complex operations can be constructed.

       All the string operators define or redefine a value (either a string
       or a numeric value) to a single-line macro. When producing a string
       value, it may change the style of quoting of the input string or
       strings, and possibly use `\'-escapes inside ``'-quoted strings.

4.2.1 Concatenating Strings: `%strcat'

       The `%strcat' operator concatenates quoted strings and assign them
       to a single-line macro.

       For example:

       %strcat alpha "Alpha: ", '12" screen'

       ... would assign the value `'Alpha: 12" screen'' to `alpha'.
       Similarly:

       %strcat beta '"foo"\', "'bar'"

       ... would assign the value ``"foo"\\'bar'`' to `beta'.

       The use of commas to separate strings is permitted but optional.

4.2.2 String Length: `%strlen'

       The `%strlen' operator assigns the length of a string to a macro.
       For example:

       %strlen charcnt 'my string'

       In this example, `charcnt' would receive the value 9, just as if an
       `%assign' had been used. In this example, `'my string'' was a
       literal string but it could also have been a single-line macro that
       expands to a string, as in the following example:

       %define sometext 'my string'
       %strlen charcnt sometext

       As in the first case, this would result in `charcnt' being assigned
       the value of 9.

4.2.3 Extracting Substrings: `%substr'

       Individual letters or substrings in strings can be extracted using
       the `%substr' operator. An example of its use is probably more
       useful than the description:

       %substr mychar 'xyzw' 1       ; equivalent to %define mychar 'x'
       %substr mychar 'xyzw' 2       ; equivalent to %define mychar 'y'
       %substr mychar 'xyzw' 3       ; equivalent to %define mychar 'z'
       %substr mychar 'xyzw' 2,2     ; equivalent to %define mychar 'yz'
       %substr mychar 'xyzw' 2,-1    ; equivalent to %define mychar 'yzw'
       %substr mychar 'xyzw' 2,-2    ; equivalent to %define mychar 'yz'

       As with `%strlen' (see section 4.2.2), the first parameter is the
       single-line macro to be created and the second is the string. The
       third parameter specifies the first character to be selected, and
       the optional fourth parameter preceeded by comma) is the length.
       Note that the first index is 1, not 0 and the last index is equal to
       the value that `%strlen' would assign given the same string. Index
       values out of range result in an empty string. A negative length
       means "until N-1 characters before the end of string", i.e. `-1'
       means until end of string, `-2' until one character before, etc.

   4.3 Multi-Line Macros: `%macro'

       Multi-line macros are much more like the type of macro seen in MASM
       and TASM: a multi-line macro definition in NASM looks something like
       this.

       %macro  prologue 1
      
               push    ebp
               mov     ebp,esp
               sub     esp,%1
      
       %endmacro

       This defines a C-like function prologue as a macro: so you would
       invoke the macro with a call such as

       myfunc:   prologue 12

       which would expand to the three lines of code

       myfunc: push    ebp
               mov     ebp,esp
               sub     esp,12

       The number `1' after the macro name in the `%macro' line defines the
       number of parameters the macro `prologue' expects to receive. The
       use of `%1' inside the macro definition refers to the first
       parameter to the macro call. With a macro taking more than one
       parameter, subsequent parameters would be referred to as `%2', `%3'
       and so on.

       Multi-line macros, like single-line macros, are case-sensitive,
       unless you define them using the alternative directive `%imacro'.

       If you need to pass a comma as _part_ of a parameter to a multi-line
       macro, you can do that by enclosing the entire parameter in braces.
       So you could code things like

       %macro  silly 2
      
           %2: db      %1
      
       %endmacro
      
               silly 'a', letter_a             ; letter_a:  db 'a'
               silly 'ab', string_ab           ; string_ab: db 'ab'
               silly {13,10}, crlf             ; crlf:      db 13,10

4.3.1 Overloading Multi-Line Macros

       As with single-line macros, multi-line macros can be overloaded by
       defining the same macro name several times with different numbers of
       parameters. This time, no exception is made for macros with no
       parameters at all. So you could define

       %macro  prologue 0
      
               push    ebp
               mov     ebp,esp
      
       %endmacro

       to define an alternative form of the function prologue which
       allocates no local stack space.

       Sometimes, however, you might want to `overload' a machine
       instruction; for example, you might want to define

       %macro  push 2
      
               push    %1
               push    %2
      
       %endmacro

       so that you could code

               push    ebx             ; this line is not a macro call
               push    eax,ecx         ; but this one is

       Ordinarily, NASM will give a warning for the first of the above two
       lines, since `push' is now defined to be a macro, and is being
       invoked with a number of parameters for which no definition has been
       given. The correct code will still be generated, but the assembler
       will give a warning. This warning can be disabled by the use of the
       `-w-macro-params' command-line option (see section 2.1.24).

4.3.2 Macro-Local Labels

       NASM allows you to define labels within a multi-line macro
       definition in such a way as to make them local to the macro call: so
       calling the same macro multiple times will use a different label
       each time. You do this by prefixing `%%' to the label name. So you
       can invent an instruction which executes a `RET' if the `Z' flag is
       set by doing this:

       %macro  retz 0
      
               jnz     %%skip
               ret
           %%skip:
      
       %endmacro

       You can call this macro as many times as you want, and every time
       you call it NASM will make up a different `real' name to substitute
       for the label `%%skip'. The names NASM invents are of the form
       `..@2345.skip', where the number 2345 changes with every macro call.
       The `..@' prefix prevents macro-local labels from interfering with
       the local label mechanism, as described in section 3.9. You should
       avoid defining your own labels in this form (the `..@' prefix, then
       a number, then another period) in case they interfere with macro-
       local labels.

4.3.3 Greedy Macro Parameters

       Occasionally it is useful to define a macro which lumps its entire
       command line into one parameter definition, possibly after
       extracting one or two smaller parameters from the front. An example
       might be a macro to write a text string to a file in MS-DOS, where
       you might want to be able to write

               writefile [filehandle],"hello, world",13,10

       NASM allows you to define the last parameter of a macro to be
       _greedy_, meaning that if you invoke the macro with more parameters
       than it expects, all the spare parameters get lumped into the last
       defined one along with the separating commas. So if you code:

       %macro  writefile 2+
      
               jmp     %%endstr
         %%str:        db      %2
         %%endstr:
               mov     dx,%%str
               mov     cx,%%endstr-%%str
               mov     bx,%1
               mov     ah,0x40
               int     0x21
      
       %endmacro

       then the example call to `writefile' above will work as expected:
       the text before the first comma, `[filehandle]', is used as the
       first macro parameter and expanded when `%1' is referred to, and all
       the subsequent text is lumped into `%2' and placed after the `db'.

       The greedy nature of the macro is indicated to NASM by the use of
       the `+' sign after the parameter count on the `%macro' line.

       If you define a greedy macro, you are effectively telling NASM how
       it should expand the macro given _any_ number of parameters from the
       actual number specified up to infinity; in this case, for example,
       NASM now knows what to do when it sees a call to `writefile' with 2,
       3, 4 or more parameters. NASM will take this into account when
       overloading macros, and will not allow you to define another form of
       `writefile' taking 4 parameters (for example).

       Of course, the above macro could have been implemented as a non-
       greedy macro, in which case the call to it would have had to look
       like

                 writefile [filehandle], {"hello, world",13,10}

       NASM provides both mechanisms for putting commas in macro
       parameters, and you choose which one you prefer for each macro
       definition.

       See section 6.3.1 for a better way to write the above macro.

4.3.4 Macro Parameters Range

       NASM allows you to expand parameters via special construction
       `%{x:y}' where `x' is the first parameter index and `y' is the last.
       Any index can be either negative or positive but must never be zero.

       For example

       %macro mpar 1-*
            db %{3:5}
       %endmacro
      
       mpar 1,2,3,4,5,6

       expands to `3,4,5' range.

       Even more, the parameters can be reversed so that

       %macro mpar 1-*
            db %{5:3}
       %endmacro
      
       mpar 1,2,3,4,5,6

       expands to `5,4,3' range.

       But even this is not the last. The parameters can be addressed via
       negative indices so NASM will count them reversed. The ones who know
       Python may see the analogue here.

       %macro mpar 1-*
            db %{-1:-3}
       %endmacro
      
       mpar 1,2,3,4,5,6

       expands to `6,5,4' range.

       Note that NASM uses comma to separate parameters being expanded.

       By the way, here is a trick - you might use the index `%{-1:-1'}
       which gives you the last argument passed to a macro.

4.3.5 Default Macro Parameters

       NASM also allows you to define a multi-line macro with a _range_ of
       allowable parameter counts. If you do this, you can specify defaults
       for omitted parameters. So, for example:

       %macro  die 0-1 "Painful program death has occurred."
      
               writefile 2,%1
               mov     ax,0x4c01
               int     0x21
      
       %endmacro

       This macro (which makes use of the `writefile' macro defined in
       section 4.3.3) can be called with an explicit error message, which
       it will display on the error output stream before exiting, or it can
       be called with no parameters, in which case it will use the default
       error message supplied in the macro definition.

       In general, you supply a minimum and maximum number of parameters
       for a macro of this type; the minimum number of parameters are then
       required in the macro call, and then you provide defaults for the
       optional ones. So if a macro definition began with the line

       %macro foobar 1-3 eax,[ebx+2]

       then it could be called with between one and three parameters, and
       `%1' would always be taken from the macro call. `%2', if not
       specified by the macro call, would default to `eax', and `%3' if not
       specified would default to `[ebx+2]'.

       You can provide extra information to a macro by providing too many
       default parameters:

       %macro quux 1 something

       This will trigger a warning by default; see section 2.1.24 for more
       information. When `quux' is invoked, it receives not one but two
       parameters. `something' can be referred to as `%2'. The difference
       between passing `something' this way and writing `something' in the
       macro body is that with this way `something' is evaluated when the
       macro is defined, not when it is expanded.

       You may omit parameter defaults from the macro definition, in which
       case the parameter default is taken to be blank. This can be useful
       for macros which can take a variable number of parameters, since the
       `%0' token (see section 4.3.6) allows you to determine how many
       parameters were really passed to the macro call.

       This defaulting mechanism can be combined with the greedy-parameter
       mechanism; so the `die' macro above could be made more powerful, and
       more useful, by changing the first line of the definition to

       %macro die 0-1+ "Painful program death has occurred.",13,10

       The maximum parameter count can be infinite, denoted by `*'. In this
       case, of course, it is impossible to provide a _full_ set of default
       parameters. Examples of this usage are shown in section 4.3.8.

4.3.6 `%0': Macro Parameter Counter

       The parameter reference `%0' will return a numeric constant giving
       the number of parameters received, that is, if `%0' is n then `%'n
       is the last parameter. `%0' is mostly useful for macros that can
       take a variable number of parameters. It can be used as an argument
       to `%rep' (see section 4.5) in order to iterate through all the
       parameters of a macro. Examples are given in section 4.3.8.

4.3.7 `%00': Label Preceeding Macro

       `%00' will return the label preceeding the macro invocation, if any.
       The label must be on the same line as the macro invocation, may be a
       local label (see section 3.9), and need not end in a colon.

4.3.8 `%rotate': Rotating Macro Parameters

       Unix shell programmers will be familiar with the `shift' shell
       command, which allows the arguments passed to a shell script
       (referenced as `$1', `$2' and so on) to be moved left by one place,
       so that the argument previously referenced as `$2' becomes available
       as `$1', and the argument previously referenced as `$1' is no longer
       available at all.

       NASM provides a similar mechanism, in the form of `%rotate'. As its
       name suggests, it differs from the Unix `shift' in that no
       parameters are lost: parameters rotated off the left end of the
       argument list reappear on the right, and vice versa.

       `%rotate' is invoked with a single numeric argument (which may be an
       expression). The macro parameters are rotated to the left by that
       many places. If the argument to `%rotate' is negative, the macro
       parameters are rotated to the right.

       So a pair of macros to save and restore a set of registers might
       work as follows:

       %macro  multipush 1-*
      
         %rep  %0
               push    %1
         %rotate 1
         %endrep
      
       %endmacro

       This macro invokes the `PUSH' instruction on each of its arguments
       in turn, from left to right. It begins by pushing its first
       argument, `%1', then invokes `%rotate' to move all the arguments one
       place to the left, so that the original second argument is now
       available as `%1'. Repeating this procedure as many times as there
       were arguments (achieved by supplying `%0' as the argument to
       `%rep') causes each argument in turn to be pushed.

       Note also the use of `*' as the maximum parameter count, indicating
       that there is no upper limit on the number of parameters you may
       supply to the `multipush' macro.

       It would be convenient, when using this macro, to have a `POP'
       equivalent, which _didn't_ require the arguments to be given in
       reverse order. Ideally, you would write the `multipush' macro call,
       then cut-and-paste the line to where the pop needed to be done, and
       change the name of the called macro to `multipop', and the macro
       would take care of popping the registers in the opposite order from
       the one in which they were pushed.

       This can be done by the following definition:

       %macro  multipop 1-*
      
         %rep %0
         %rotate -1
               pop     %1
         %endrep
      
       %endmacro

       This macro begins by rotating its arguments one place to the
       _right_, so that the original _last_ argument appears as `%1'. This
       is then popped, and the arguments are rotated right again, so the
       second-to-last argument becomes `%1'. Thus the arguments are
       iterated through in reverse order.

4.3.9 Concatenating Macro Parameters

       NASM can concatenate macro parameters and macro indirection
       constructs on to other text surrounding them. This allows you to
       declare a family of symbols, for example, in a macro definition. If,
       for example, you wanted to generate a table of key codes along with
       offsets into the table, you could code something like

       %macro keytab_entry 2
      
           keypos%1    equ     $-keytab
                       db      %2
      
       %endmacro
      
       keytab:
                 keytab_entry F1,128+1
                 keytab_entry F2,128+2
                 keytab_entry Return,13

       which would expand to

       keytab:
       keyposF1        equ     $-keytab
                       db     128+1
       keyposF2        equ     $-keytab
                       db      128+2
       keyposReturn    equ     $-keytab
                       db      13

       You can just as easily concatenate text on to the other end of a
       macro parameter, by writing `%1foo'.

       If you need to append a _digit_ to a macro parameter, for example
       defining labels `foo1' and `foo2' when passed the parameter `foo',
       you can't code `%11' because that would be taken as the eleventh
       macro parameter. Instead, you must code `%{1}1', which will separate
       the first `1' (giving the number of the macro parameter) from the
       second (literal text to be concatenated to the parameter).

       This concatenation can also be applied to other preprocessor in-line
       objects, such as macro-local labels (section 4.3.2) and context-
       local labels (section 4.7.2). In all cases, ambiguities in syntax
       can be resolved by enclosing everything after the `%' sign and
       before the literal text in braces: so `%{%foo}bar' concatenates the
       text `bar' to the end of the real name of the macro-local label
       `%%foo'. (This is unnecessary, since the form NASM uses for the real
       names of macro-local labels means that the two usages `%{%foo}bar'
       and `%%foobar' would both expand to the same thing anyway;
       nevertheless, the capability is there.)

       The single-line macro indirection construct, `%[...]' (section
       4.1.3), behaves the same way as macro parameters for the purpose of
       concatenation.

       See also the `%+' operator, section 4.1.4.

4.3.10 Condition Codes as Macro Parameters

       NASM can give special treatment to a macro parameter which contains
       a condition code. For a start, you can refer to the macro parameter
       `%1' by means of the alternative syntax `%+1', which informs NASM
       that this macro parameter is supposed to contain a condition code,
       and will cause the preprocessor to report an error message if the
       macro is called with a parameter which is _not_ a valid condition
       code.

       Far more usefully, though, you can refer to the macro parameter by
       means of `%-1', which NASM will expand as the _inverse_ condition
       code. So the `retz' macro defined in section 4.3.2 can be replaced
       by a general conditional-return macro like this:

       %macro  retc 1
      
               j%-1    %%skip
               ret
         %%skip:
      
       %endmacro

       This macro can now be invoked using calls like `retc ne', which will
       cause the conditional-jump instruction in the macro expansion to
       come out as `JE', or `retc po' which will make the jump a `JPE'.

       The `%+1' macro-parameter reference is quite happy to interpret the
       arguments `CXZ' and `ECXZ' as valid condition codes; however, `%-1'
       will report an error if passed either of these, because no inverse
       condition code exists.

4.3.11 Disabling Listing Expansion

       When NASM is generating a listing file from your program, it will
       generally expand multi-line macros by means of writing the macro
       call and then listing each line of the expansion. This allows you to
       see which instructions in the macro expansion are generating what
       code; however, for some macros this clutters the listing up
       unnecessarily.

       NASM therefore provides the `.nolist' qualifier, which you can
       include in a macro definition to inhibit the expansion of the macro
       in the listing file. The `.nolist' qualifier comes directly after
       the number of parameters, like this:

       %macro foo 1.nolist

       Or like this:

       %macro bar 1-5+.nolist a,b,c,d,e,f,g,h

4.3.12 Undefining Multi-Line Macros: `%unmacro'

       Multi-line macros can be removed with the `%unmacro' directive.
       Unlike the `%undef' directive, however, `%unmacro' takes an argument
       specification, and will only remove exact matches with that argument
       specification.

       For example:

       %macro foo 1-3
               ; Do something
       %endmacro
       %unmacro foo 1-3

       removes the previously defined macro `foo', but

       %macro bar 1-3
               ; Do something
       %endmacro
       %unmacro bar 1

       does _not_ remove the macro `bar', since the argument specification
       does not match exactly.

   4.4 Conditional Assembly

       Similarly to the C preprocessor, NASM allows sections of a source
       file to be assembled only if certain conditions are met. The general
       syntax of this feature looks like this:

       %if<condition>
           ; some code which only appears if <condition> is met
       %elif<condition2>
           ; only appears if <condition> is not met but <condition2> is
       %else
           ; this appears if neither <condition> nor <condition2> was met
       %endif

       The inverse forms `%ifn' and `%elifn' are also supported.

       The `%else' clause is optional, as is the `%elif' clause. You can
       have more than one `%elif' clause as well.

       There are a number of variants of the `%if' directive. Each has its
       corresponding `%elif', `%ifn', and `%elifn' directives; for example,
       the equivalents to the `%ifdef' directive are `%elifdef', `%ifndef',
       and `%elifndef'.

4.4.1 `%ifdef': Testing Single-Line Macro Existence

       Beginning a conditional-assembly block with the line `%ifdef MACRO'
       will assemble the subsequent code if, and only if, a single-line
       macro called `MACRO' is defined. If not, then the `%elif' and
       `%else' blocks (if any) will be processed instead.

       For example, when debugging a program, you might want to write code
       such as

                 ; perform some function
       %ifdef DEBUG
                 writefile 2,"Function performed successfully",13,10
       %endif
                 ; go and do something else

       Then you could use the command-line option `-dDEBUG' to create a
       version of the program which produced debugging messages, and remove
       the option to generate the final release version of the program.

       You can test for a macro _not_ being defined by using `%ifndef'
       instead of `%ifdef'. You can also test for macro definitions in
       `%elif' blocks by using `%elifdef' and `%elifndef'.

4.4.2 `%ifmacro': Testing Multi-Line Macro Existence

       The `%ifmacro' directive operates in the same way as the `%ifdef'
       directive, except that it checks for the existence of a multi-line
       macro.

       For example, you may be working with a large project and not have
       control over the macros in a library. You may want to create a macro
       with one name if it doesn't already exist, and another name if one
       with that name does exist.

       The `%ifmacro' is considered true if defining a macro with the given
       name and number of arguments would cause a definitions conflict. For
       example:

       %ifmacro MyMacro 1-3
      
            %error "MyMacro 1-3" causes a conflict with an existing macro.
      
       %else
      
            %macro MyMacro 1-3
      
                    ; insert code to define the macro
      
            %endmacro
      
       %endif

       This will create the macro "MyMacro 1-3" if no macro already exists
       which would conflict with it, and emits a warning if there would be
       a definition conflict.

       You can test for the macro not existing by using the `%ifnmacro'
       instead of `%ifmacro'. Additional tests can be performed in `%elif'
       blocks by using `%elifmacro' and `%elifnmacro'.

4.4.3 `%ifctx': Testing the Context Stack

       The conditional-assembly construct `%ifctx' will cause the
       subsequent code to be assembled if and only if the top context on
       the preprocessor's context stack has the same name as one of the
       arguments. As with `%ifdef', the inverse and `%elif' forms
       `%ifnctx', `%elifctx' and `%elifnctx' are also supported.

       For more details of the context stack, see section 4.7. For a sample
       use of `%ifctx', see section 4.7.6.

4.4.4 `%if': Testing Arbitrary Numeric Expressions

       The conditional-assembly construct `%if expr' will cause the
       subsequent code to be assembled if and only if the value of the
       numeric expression `expr' is non-zero. An example of the use of this
       feature is in deciding when to break out of a `%rep' preprocessor
       loop: see section 4.5 for a detailed example.

       The expression given to `%if', and its counterpart `%elif', is a
       critical expression (see section 3.8).

       `%if' extends the normal NASM expression syntax, by providing a set
       of relational operators which are not normally available in
       expressions. The operators `=', `<', `>', `<=', `>=' and `<>' test
       equality, less-than, greater-than, less-or-equal, greater-or-equal
       and not-equal respectively. The C-like forms `==' and `!=' are
       supported as alternative forms of `=' and `<>'. In addition, low-
       priority logical operators `&&', `^^' and `||' are provided,
       supplying logical AND, logical XOR and logical OR. These work like
       the C logical operators (although C has no logical XOR), in that
       they always return either 0 or 1, and treat any non-zero input as 1
       (so that `^^', for example, returns 1 if exactly one of its inputs
       is zero, and 0 otherwise). The relational operators also return 1
       for true and 0 for false.

       Like other `%if' constructs, `%if' has a counterpart `%elif', and
       negative forms `%ifn' and `%elifn'.

4.4.5 `%ifidn' and `%ifidni': Testing Exact Text Identity

       The construct `%ifidn text1,text2' will cause the subsequent code to
       be assembled if and only if `text1' and `text2', after expanding
       single-line macros, are identical pieces of text. Differences in
       white space are not counted.

       `%ifidni' is similar to `%ifidn', but is case-insensitive.

       For example, the following macro pushes a register or number on the
       stack, and allows you to treat `IP' as a real register:

       %macro  pushparam 1
      
         %ifidni %1,ip
               call    %%label
         %%label:
         %else
               push    %1
         %endif
      
       %endmacro

       Like other `%if' constructs, `%ifidn' has a counterpart `%elifidn',
       and negative forms `%ifnidn' and `%elifnidn'. Similarly, `%ifidni'
       has counterparts `%elifidni', `%ifnidni' and `%elifnidni'.

4.4.6 `%ifid', `%ifnum', `%ifstr': Testing Token Types

       Some macros will want to perform different tasks depending on
       whether they are passed a number, a string, or an identifier. For
       example, a string output macro might want to be able to cope with
       being passed either a string constant or a pointer to an existing
       string.

       The conditional assembly construct `%ifid', taking one parameter
       (which may be blank), assembles the subsequent code if and only if
       the first token in the parameter exists and is an identifier.
       `%ifnum' works similarly, but tests for the token being a numeric
       constant; `%ifstr' tests for it being a string.

       For example, the `writefile' macro defined in section 4.3.3 can be
       extended to take advantage of `%ifstr' in the following fashion:

       %macro writefile 2-3+
      
         %ifstr %2
               jmp     %%endstr
           %if %0 = 3
             %%str:    db      %2,%3
           %else
             %%str:    db      %2
           %endif
             %%endstr: mov     dx,%%str
                       mov     cx,%%endstr-%%str
         %else
                       mov     dx,%2
                       mov     cx,%3
         %endif
                       mov     bx,%1
                       mov     ah,0x40
                       int     0x21
      
       %endmacro

       Then the `writefile' macro can cope with being called in either of
       the following two ways:

               writefile [file], strpointer, length
               writefile [file], "hello", 13, 10

       In the first, `strpointer' is used as the address of an already-
       declared string, and `length' is used as its length; in the second,
       a string is given to the macro, which therefore declares it itself
       and works out the address and length for itself.

       Note the use of `%if' inside the `%ifstr': this is to detect whether
       the macro was passed two arguments (so the string would be a single
       string constant, and `db %2' would be adequate) or more (in which
       case, all but the first two would be lumped together into `%3', and
       `db %2,%3' would be required).

       The usual `%elif'..., `%ifn'..., and `%elifn'... versions exist for
       each of `%ifid', `%ifnum' and `%ifstr'.

4.4.7 `%iftoken': Test for a Single Token

       Some macros will want to do different things depending on if it is
       passed a single token (e.g. paste it to something else using `%+')
       versus a multi-token sequence.

       The conditional assembly construct `%iftoken' assembles the
       subsequent code if and only if the expanded parameters consist of
       exactly one token, possibly surrounded by whitespace.

       For example:

       %iftoken 1

       will assemble the subsequent code, but

       %iftoken -1

       will not, since `-1' contains two tokens: the unary minus operator
       `-', and the number `1'.

       The usual `%eliftoken', `%ifntoken', and `%elifntoken' variants are
       also provided.

4.4.8 `%ifempty': Test for Empty Expansion

       The conditional assembly construct `%ifempty' assembles the
       subsequent code if and only if the expanded parameters do not
       contain any tokens at all, whitespace excepted.

       The usual `%elifempty', `%ifnempty', and `%elifnempty' variants are
       also provided.

4.4.9 `%ifenv': Test If Environment Variable Exists

       The conditional assembly construct `%ifenv' assembles the subsequent
       code if and only if the environment variable referenced by the
       `%!<env>' directive exists.

       The usual `%elifenv', `%ifnenv', and `%elifnenv' variants are also
       provided.

       Just as for `%!<env>' the argument should be written as a string if
       it contains characters that would not be legal in an identifier. See
       section 4.10.2.

   4.5 Preprocessor Loops: `%rep'

       NASM's `TIMES' prefix, though useful, cannot be used to invoke a
       multi-line macro multiple times, because it is processed by NASM
       after macros have already been expanded. Therefore NASM provides
       another form of loop, this time at the preprocessor level: `%rep'.

       The directives `%rep' and `%endrep' (`%rep' takes a numeric
       argument, which can be an expression; `%endrep' takes no arguments)
       can be used to enclose a chunk of code, which is then replicated as
       many times as specified by the preprocessor:

       %assign i 0
       %rep    64
               inc     word [table+2*i]
       %assign i i+1
       %endrep

       This will generate a sequence of 64 `INC' instructions, incrementing
       every word of memory from `[table]' to `[table+126]'.

       For more complex termination conditions, or to break out of a repeat
       loop part way along, you can use the `%exitrep' directive to
       terminate the loop, like this:

       fibonacci:
       %assign i 0
       %assign j 1
       %rep 100
       %if j > 65535
           %exitrep
       %endif
               dw j
       %assign k j+i
       %assign i j
       %assign j k
       %endrep
      
       fib_number equ ($-fibonacci)/2

       This produces a list of all the Fibonacci numbers that will fit in
       16 bits. Note that a maximum repeat count must still be given to
       `%rep'. This is to prevent the possibility of NASM getting into an
       infinite loop in the preprocessor, which (on multitasking or multi-
       user systems) would typically cause all the system memory to be
       gradually used up and other applications to start crashing.

       Note a maximum repeat count is limited by 62 bit number, though it
       is hardly possible that you ever need anything bigger.

   4.6 Source Files and Dependencies

       These commands allow you to split your sources into multiple files.

4.6.1 `%include': Including Other Files

       Using, once again, a very similar syntax to the C preprocessor,
       NASM's preprocessor lets you include other source files into your
       code. This is done by the use of the `%include' directive:

       %include "macros.mac"

       will include the contents of the file `macros.mac' into the source
       file containing the `%include' directive.

       Include files are searched for in the current directory (the
       directory you're in when you run NASM, as opposed to the location of
       the NASM executable or the location of the source file), plus any
       directories specified on the NASM command line using the `-i'
       option.

       The standard C idiom for preventing a file being included more than
       once is just as applicable in NASM: if the file `macros.mac' has the
       form

       %ifndef MACROS_MAC
           %define MACROS_MAC
           ; now define some macros
       %endif

       then including the file more than once will not cause errors,
       because the second time the file is included nothing will happen
       because the macro `MACROS_MAC' will already be defined.

       You can force a file to be included even if there is no `%include'
       directive that explicitly includes it, by using the `-p' option on
       the NASM command line (see section 2.1.17).

4.6.2 `%pathsearch': Search the Include Path

       The `%pathsearch' directive takes a single-line macro name and a
       filename, and declare or redefines the specified single-line macro
       to be the include-path-resolved version of the filename, if the file
       exists (otherwise, it is passed unchanged.)

       For example,

       %pathsearch MyFoo "foo.bin"

       ... with `-Ibins/' in the include path may end up defining the macro
       `MyFoo' to be `"bins/foo.bin"'.

4.6.3 `%depend': Add Dependent Files

       The `%depend' directive takes a filename and adds it to the list of
       files to be emitted as dependency generation when the `-M' options
       and its relatives (see section 2.1.4) are used. It produces no
       output.

       This is generally used in conjunction with `%pathsearch'. For
       example, a simplified version of the standard macro wrapper for the
       `INCBIN' directive looks like:

       %imacro incbin 1-2+ 0
       %pathsearch dep %1
       %depend dep
               incbin dep,%2
       %endmacro

       This first resolves the location of the file into the macro `dep',
       then adds it to the dependency lists, and finally issues the
       assembler-level `INCBIN' directive.

4.6.4 `%use': Include Standard Macro Package

       The `%use' directive is similar to `%include', but rather than
       including the contents of a file, it includes a named standard macro
       package. The standard macro packages are part of NASM, and are
       described in chapter 5.

       Unlike the `%include' directive, package names for the `%use'
       directive do not require quotes, but quotes are permitted. In NASM
       2.04 and 2.05 the unquoted form would be macro-expanded; this is no
       longer true. Thus, the following lines are equivalent:

       %use altreg
       %use 'altreg'

       Standard macro packages are protected from multiple inclusion. When
       a standard macro package is used, a testable single-line macro of
       the form `__USE_'_package_`__' is also defined, see section 4.12.8.

   4.7 The Context Stack

       Having labels that are local to a macro definition is sometimes not
       quite powerful enough: sometimes you want to be able to share labels
       between several macro calls. An example might be a `REPEAT' ...
       `UNTIL' loop, in which the expansion of the `REPEAT' macro would
       need to be able to refer to a label which the `UNTIL' macro had
       defined. However, for such a macro you would also want to be able to
       nest these loops.

       NASM provides this level of power by means of a _context stack_. The
       preprocessor maintains a stack of _contexts_, each of which is
       characterized by a name. You add a new context to the stack using
       the `%push' directive, and remove one using `%pop'. You can define
       labels that are local to a particular context on the stack.

4.7.1 `%push' and `%pop': Creating and Removing Contexts

       The `%push' directive is used to create a new context and place it
       on the top of the context stack. `%push' takes an optional argument,
       which is the name of the context. For example:

       %push    foobar

       This pushes a new context called `foobar' on the stack. You can have
       several contexts on the stack with the same name: they can still be
       distinguished. If no name is given, the context is unnamed (this is
       normally used when both the `%push' and the `%pop' are inside a
       single macro definition.)

       The directive `%pop', taking one optional argument, removes the top
       context from the context stack and destroys it, along with any
       labels associated with it. If an argument is given, it must match
       the name of the current context, otherwise it will issue an error.

4.7.2 Context-Local Labels

       Just as the usage `%%foo' defines a label which is local to the
       particular macro call in which it is used, the usage `%$foo' is used
       to define a label which is local to the context on the top of the
       context stack. So the `REPEAT' and `UNTIL' example given above could
       be implemented by means of:

       %macro repeat 0
      
           %push   repeat
           %$begin:
      
       %endmacro
      
       %macro until 1
      
               j%-1    %$begin
           %pop
      
       %endmacro

       and invoked by means of, for example,

               mov     cx,string
               repeat
               add     cx,3
               scasb
               until   e

       which would scan every fourth byte of a string in search of the byte
       in `AL'.

       If you need to define, or access, labels local to the context
       _below_ the top one on the stack, you can use `%$$foo', or `%$$$foo'
       for the context below that, and so on.

4.7.3 Context-Local Single-Line Macros

       NASM also allows you to define single-line macros which are local to
       a particular context, in just the same way:

       %define %$localmac 3

       will define the single-line macro `%$localmac' to be local to the
       top context on the stack. Of course, after a subsequent `%push', it
       can then still be accessed by the name `%$$localmac'.

4.7.4 Context Fall-Through Lookup

       Context fall-through lookup (automatic searching of outer contexts)
       is a feature that was added in NASM version 0.98.03. Unfortunately,
       this feature is unintuitive and can result in buggy code that would
       have otherwise been prevented by NASM's error reporting. As a
       result, this feature has been _deprecated_. NASM version 2.09 will
       issue a warning when usage of this _deprecated_ feature is detected.
       Starting with NASM version 2.10, usage of this _deprecated_ feature
       will simply result in an _expression syntax error_.

       An example usage of this _deprecated_ feature follows:

       %macro ctxthru 0
       %push ctx1
           %assign %$external 1
               %push ctx2
                   %assign %$internal 1
                   mov eax, %$external
                   mov eax, %$internal
               %pop
       %pop
       %endmacro

       As demonstrated, `%$external' is being defined in the `ctx1' context
       and referenced within the `ctx2' context. With context fall-through
       lookup, referencing an undefined context-local macro like this
       implicitly searches through all outer contexts until a match is made
       or isn't found in any context. As a result, `%$external' referenced
       within the `ctx2' context would implicitly use `%$external' as
       defined in `ctx1'. Most people would expect NASM to issue an error
       in this situation because `%$external' was never defined within
       `ctx2' and also isn't qualified with the proper context depth,
       `%$$external'.

       Here is a revision of the above example with proper context depth:

       %macro ctxthru 0
       %push ctx1
           %assign %$external 1
               %push ctx2
                   %assign %$internal 1
                   mov eax, %$$external
                   mov eax, %$internal
               %pop
       %pop
       %endmacro

       As demonstrated, `%$external' is still being defined in the `ctx1'
       context and referenced within the `ctx2' context. However, the
       reference to `%$external' within `ctx2' has been fully qualified
       with the proper context depth, `%$$external', and thus is no longer
       ambiguous, unintuitive or erroneous.

4.7.5 `%repl': Renaming a Context

       If you need to change the name of the top context on the stack (in
       order, for example, to have it respond differently to `%ifctx'), you
       can execute a `%pop' followed by a `%push'; but this will have the
       side effect of destroying all context-local labels and macros
       associated with the context that was just popped.

       NASM provides the directive `%repl', which _replaces_ a context with
       a different name, without touching the associated macros and labels.
       So you could replace the destructive code

       %pop
       %push   newname

       with the non-destructive version `%repl newname'.

4.7.6 Example Use of the Context Stack: Block IFs

       This example makes use of almost all the context-stack features,
       including the conditional-assembly construct `%ifctx', to implement
       a block IF statement as a set of macros.

       %macro if 1
      
           %push if
           j%-1  %$ifnot
      
       %endmacro
      
       %macro else 0
      
         %ifctx if
               %repl   else
               jmp     %$ifend
               %$ifnot:
         %else
               %error  "expected `if' before `else'"
         %endif
      
       %endmacro
      
       %macro endif 0
      
         %ifctx if
               %$ifnot:
               %pop
         %elifctx      else
               %$ifend:
               %pop
         %else
               %error  "expected `if' or `else' before `endif'"
         %endif
      
       %endmacro

       This code is more robust than the `REPEAT' and `UNTIL' macros given
       in section 4.7.2, because it uses conditional assembly to check that
       the macros are issued in the right order (for example, not calling
       `endif' before `if') and issues a `%error' if they're not.

       In addition, the `endif' macro has to be able to cope with the two
       distinct cases of either directly following an `if', or following an
       `else'. It achieves this, again, by using conditional assembly to do
       different things depending on whether the context on top of the
       stack is `if' or `else'.

       The `else' macro has to preserve the context on the stack, in order
       to have the `%$ifnot' referred to by the `if' macro be the same as
       the one defined by the `endif' macro, but has to change the
       context's name so that `endif' will know there was an intervening
       `else'. It does this by the use of `%repl'.

       A sample usage of these macros might look like:

               cmp     ax,bx
      
               if ae
                      cmp     bx,cx
      
                      if ae
                              mov     ax,cx
                      else
                              mov     ax,bx
                      endif
      
               else
                      cmp     ax,cx
      
                      if ae
                              mov     ax,cx
                      endif
      
               endif

       The block-`IF' macros handle nesting quite happily, by means of
       pushing another context, describing the inner `if', on top of the
       one describing the outer `if'; thus `else' and `endif' always refer
       to the last unmatched `if' or `else'.

   4.8 Stack Relative Preprocessor Directives

       The following preprocessor directives provide a way to use labels to
       refer to local variables allocated on the stack.

       (*) `%arg' (see section 4.8.1)

       (*) `%stacksize' (see section 4.8.2)

       (*) `%local' (see section 4.8.3)

4.8.1 `%arg' Directive

       The `%arg' directive is used to simplify the handling of parameters
       passed on the stack. Stack based parameter passing is used by many
       high level languages, including C, C++ and Pascal.

       While NASM has macros which attempt to duplicate this functionality
       (see section 8.4.5), the syntax is not particularly convenient to
       use and is not TASM compatible. Here is an example which shows the
       use of `%arg' without any external macros:

       some_function:
      
           %push     mycontext        ; save the current context
           %stacksize large           ; tell NASM to use bp
           %arg      i:word, j_ptr:word
      
               mov     ax,[i]
               mov     bx,[j_ptr]
               add     ax,[bx]
               ret
      
           %pop                       ; restore original context

       This is similar to the procedure defined in section 8.4.5 and adds
       the value in i to the value pointed to by j_ptr and returns the sum
       in the ax register. See section 4.7.1 for an explanation of `push'
       and `pop' and the use of context stacks.

4.8.2 `%stacksize' Directive

       The `%stacksize' directive is used in conjunction with the `%arg'
       (see section 4.8.1) and the `%local' (see section 4.8.3) directives.
       It tells NASM the default size to use for subsequent `%arg' and
       `%local' directives. The `%stacksize' directive takes one required
       argument which is one of `flat', `flat64', `large' or `small'.

       %stacksize flat

       This form causes NASM to use stack-based parameter addressing
       relative to `ebp' and it assumes that a near form of call was used
       to get to this label (i.e. that `eip' is on the stack).

       %stacksize flat64

       This form causes NASM to use stack-based parameter addressing
       relative to `rbp' and it assumes that a near form of call was used
       to get to this label (i.e. that `rip' is on the stack).

       %stacksize large

       This form uses `bp' to do stack-based parameter addressing and
       assumes that a far form of call was used to get to this address
       (i.e. that `ip' and `cs' are on the stack).

       %stacksize small

       This form also uses `bp' to address stack parameters, but it is
       different from `large' because it also assumes that the old value of
       bp is pushed onto the stack (i.e. it expects an `ENTER'
       instruction). In other words, it expects that `bp', `ip' and `cs'
       are on the top of the stack, underneath any local space which may
       have been allocated by `ENTER'. This form is probably most useful
       when used in combination with the `%local' directive (see section
       4.8.3).

4.8.3 `%local' Directive

       The `%local' directive is used to simplify the use of local
       temporary stack variables allocated in a stack frame. Automatic
       local variables in C are an example of this kind of variable. The
       `%local' directive is most useful when used with the `%stacksize'
       (see section 4.8.2 and is also compatible with the `%arg' directive
       (see section 4.8.1). It allows simplified reference to variables on
       the stack which have been allocated typically by using the `ENTER'
       instruction. An example of its use is the following:

       silly_swap:
      
           %push mycontext             ; save the current context
           %stacksize small            ; tell NASM to use bp
           %assign %$localsize 0       ; see text for explanation
           %local old_ax:word, old_dx:word
      
               enter   %$localsize,0   ; see text for explanation
               mov     [old_ax],ax     ; swap ax & bx
               mov     [old_dx],dx     ; and swap dx & cx
               mov     ax,bx
               mov     dx,cx
               mov     bx,[old_ax]
               mov     cx,[old_dx]
               leave                   ; restore old bp
               ret                     ;
      
           %pop                        ; restore original context

       The `%$localsize' variable is used internally by the `%local'
       directive and _must_ be defined within the current context before
       the `%local' directive may be used. Failure to do so will result in
       one expression syntax error for each `%local' variable declared. It
       then may be used in the construction of an appropriately sized ENTER
       instruction as shown in the example.

   4.9 Reporting User-Defined Errors: `%error', `%warning', `%fatal'

       The preprocessor directive `%error' will cause NASM to report an
       error if it occurs in assembled code. So if other users are going to
       try to assemble your source files, you can ensure that they define
       the right macros by means of code like this:

       %ifdef F1
           ; do some setup
       %elifdef F2
           ; do some different setup
       %else
           %error "Neither F1 nor F2 was defined."
       %endif

       Then any user who fails to understand the way your code is supposed
       to be assembled will be quickly warned of their mistake, rather than
       having to wait until the program crashes on being run and then not
       knowing what went wrong.

       Similarly, `%warning' issues a warning, but allows assembly to
       continue:

       %ifdef F1
           ; do some setup
       %elifdef F2
           ; do some different setup
       %else
           %warning "Neither F1 nor F2 was defined, assuming F1."
           %define F1
       %endif

       `%error' and `%warning' are issued only on the final assembly pass.
       This makes them safe to use in conjunction with tests that depend on
       symbol values.

       `%fatal' terminates assembly immediately, regardless of pass. This
       is useful when there is no point in continuing the assembly further,
       and doing so is likely just going to cause a spew of confusing error
       messages.

       It is optional for the message string after `%error', `%warning' or
       `%fatal' to be quoted. If it is _not_, then single-line macros are
       expanded in it, which can be used to display more information to the
       user. For example:

       %if foo > 64
           %assign foo_over foo-64
           %error foo is foo_over bytes too large
       %endif

  4.10 Other Preprocessor Directives

       NASM also has preprocessor directives which allow access to
       information from external sources. Currently they include:

       (*) `%line' enables NASM to correctly handle the output of another
           preprocessor (see section 4.10.1).

       (*) `%!' enables NASM to read in the value of an environment
           variable, which can then be used in your program (see section
           4.10.2).

4.10.1 `%line' Directive

       The `%line' directive is used to notify NASM that the input line
       corresponds to a specific line number in another file. Typically
       this other file would be an original source file, with the current
       NASM input being the output of a pre-processor. The `%line'
       directive allows NASM to output messages which indicate the line
       number of the original source file, instead of the file that is
       being read by NASM.

       This preprocessor directive is not generally of use to programmers,
       by may be of interest to preprocessor authors. The usage of the
       `%line' preprocessor directive is as follows:

       %line nnn[+mmm] [filename]

       In this directive, `nnn' identifies the line of the original source
       file which this line corresponds to. `mmm' is an optional parameter
       which specifies a line increment value; each line of the input file
       read in is considered to correspond to `mmm' lines of the original
       source file. Finally, `filename' is an optional parameter which
       specifies the file name of the original source file.

       After reading a `%line' preprocessor directive, NASM will report all
       file name and line numbers relative to the values specified therein.

4.10.2 `%!'`<env>': Read an environment variable.

       The `%!<env>' directive makes it possible to read the value of an
       environment variable at assembly time. This could, for example, be
       used to store the contents of an environment variable into a string,
       which could be used at some other point in your code.

       For example, suppose that you have an environment variable `FOO',
       and you want the contents of `FOO' to be embedded in your program.
       You could do that as follows:

       %defstr FOO          %!FOO

       See section 4.1.8 for notes on the `%defstr' directive.

       If the name of the environment variable contains non-identifier
       characters, you can use string quotes to surround the name of the
       variable, for example:

       %defstr C_colon      %!'C:'

  4.11 Comment Blocks: `%comment'

       The `%comment' and `%endcomment' directives are used to specify a
       block of commented (i.e. unprocessed) code/text. Everything between
       `%comment' and `%endcomment' will be ignored by the preprocessor.

       %comment
           ; some code, text or data to be ignored
       %endcomment

  4.12 Standard Macros

       NASM defines a set of standard macros, which are already defined
       when it starts to process any source file. If you really need a
       program to be assembled with no pre-defined macros, you can use the
       `%clear' directive to empty the preprocessor of everything but
       context-local preprocessor variables and single-line macros.

       Most user-level assembler directives (see chapter 6) are implemented
       as macros which invoke primitive directives; these are described in
       chapter 6. The rest of the standard macro set is described here.

4.12.1 NASM Version Macros

       The single-line macros `__NASM_MAJOR__', `__NASM_MINOR__',
       `__NASM_SUBMINOR__' and `___NASM_PATCHLEVEL__' expand to the major,
       minor, subminor and patch level parts of the version number of NASM
       being used. So, under NASM 0.98.32p1 for example, `__NASM_MAJOR__'
       would be defined to be 0, `__NASM_MINOR__' would be defined as 98,
       `__NASM_SUBMINOR__' would be defined to 32, and
       `___NASM_PATCHLEVEL__' would be defined as 1.

       Additionally, the macro `__NASM_SNAPSHOT__' is defined for
       automatically generated snapshot releases _only_.

4.12.2 `__NASM_VERSION_ID__': NASM Version ID

       The single-line macro `__NASM_VERSION_ID__' expands to a dword
       integer representing the full version number of the version of nasm
       being used. The value is the equivalent to `__NASM_MAJOR__',
       `__NASM_MINOR__', `__NASM_SUBMINOR__' and `___NASM_PATCHLEVEL__'
       concatenated to produce a single doubleword. Hence, for 0.98.32p1,
       the returned number would be equivalent to:

               dd      0x00622001

       or

               db      1,32,98,0

       Note that the above lines are generate exactly the same code, the
       second line is used just to give an indication of the order that the
       separate values will be present in memory.

4.12.3 `__NASM_VER__': NASM Version string

       The single-line macro `__NASM_VER__' expands to a string which
       defines the version number of nasm being used. So, under NASM
       0.98.32 for example,

               db      __NASM_VER__

       would expand to

               db      "0.98.32"

4.12.4 `__FILE__' and `__LINE__': File Name and Line Number

       Like the C preprocessor, NASM allows the user to find out the file
       name and line number containing the current instruction. The macro
       `__FILE__' expands to a string constant giving the name of the
       current input file (which may change through the course of assembly
       if `%include' directives are used), and `__LINE__' expands to a
       numeric constant giving the current line number in the input file.

       These macros could be used, for example, to communicate debugging
       information to a macro, since invoking `__LINE__' inside a macro
       definition (either single-line or multi-line) will return the line
       number of the macro _call_, rather than _definition_. So to
       determine where in a piece of code a crash is occurring, for
       example, one could write a routine `stillhere', which is passed a
       line number in `EAX' and outputs something like `line 155: still
       here'. You could then write a macro

       %macro  notdeadyet 0
      
               push    eax
               mov     eax,__LINE__
               call    stillhere
               pop     eax
      
       %endmacro

       and then pepper your code with calls to `notdeadyet' until you find
       the crash point.

4.12.5 `__BITS__': Current BITS Mode

       The `__BITS__' standard macro is updated every time that the BITS
       mode is set using the `BITS XX' or `[BITS XX]' directive, where XX
       is a valid mode number of 16, 32 or 64. `__BITS__' receives the
       specified mode number and makes it globally available. This can be
       very useful for those who utilize mode-dependent macros.

4.12.6 `__OUTPUT_FORMAT__': Current Output Format

       The `__OUTPUT_FORMAT__' standard macro holds the current Output
       Format, as given by the `-f' option or NASM's default. Type
       `nasm -hf' for a list.

       %ifidn __OUTPUT_FORMAT__, win32
        %define NEWLINE 13, 10
       %elifidn __OUTPUT_FORMAT__, elf32
        %define NEWLINE 10
       %endif

4.12.7 Assembly Date and Time Macros

       NASM provides a variety of macros that represent the timestamp of
       the assembly session.

       (*) The `__DATE__' and `__TIME__' macros give the assembly date and
           time as strings, in ISO 8601 format (`"YYYY-MM-DD"' and
           `"HH:MM:SS"', respectively.)

       (*) The `__DATE_NUM__' and `__TIME_NUM__' macros give the assembly
           date and time in numeric form; in the format `YYYYMMDD' and
           `HHMMSS' respectively.

       (*) The `__UTC_DATE__' and `__UTC_TIME__' macros give the assembly
           date and time in universal time (UTC) as strings, in ISO 8601
           format (`"YYYY-MM-DD"' and `"HH:MM:SS"', respectively.) If the
           host platform doesn't provide UTC time, these macros are
           undefined.

       (*) The `__UTC_DATE_NUM__' and `__UTC_TIME_NUM__' macros give the
           assembly date and time universal time (UTC) in numeric form; in
           the format `YYYYMMDD' and `HHMMSS' respectively. If the host
           platform doesn't provide UTC time, these macros are undefined.

       (*) The `__POSIX_TIME__' macro is defined as a number containing the
           number of seconds since the POSIX epoch, 1 January 1970 00:00:00
           UTC; excluding any leap seconds. This is computed using UTC time
           if available on the host platform, otherwise it is computed
           using the local time as if it was UTC.

       All instances of time and date macros in the same assembly session
       produce consistent output. For example, in an assembly session
       started at 42 seconds after midnight on January 1, 2010 in Moscow
       (timezone UTC+3) these macros would have the following values,
       assuming, of course, a properly configured environment with a
       correct clock:

             __DATE__             "2010-01-01"
             __TIME__             "00:00:42"
             __DATE_NUM__         20100101
             __TIME_NUM__         000042
             __UTC_DATE__         "2009-12-31"
             __UTC_TIME__         "21:00:42"
             __UTC_DATE_NUM__     20091231
             __UTC_TIME_NUM__     210042
             __POSIX_TIME__       1262293242

4.12.8 `__USE_'_package_`__': Package Include Test

       When a standard macro package (see chapter 5) is included with the
       `%use' directive (see section 4.6.4), a single-line macro of the
       form `__USE_'_package_`__' is automatically defined. This allows
       testing if a particular package is invoked or not.

       For example, if the `altreg' package is included (see section 5.1),
       then the macro `__USE_ALTREG__' is defined.

4.12.9 `__PASS__': Assembly Pass

       The macro `__PASS__' is defined to be `1' on preparatory passes, and
       `2' on the final pass. In preprocess-only mode, it is set to `3',
       and when running only to generate dependencies (due to the `-M' or
       `-MG' option, see section 2.1.4) it is set to `0'.

       _Avoid using this macro if at all possible. It is tremendously easy
       to generate very strange errors by misusing it, and the semantics
       may change in future versions of NASM._

4.12.10 `STRUC' and `ENDSTRUC': Declaring Structure Data Types

       The core of NASM contains no intrinsic means of defining data
       structures; instead, the preprocessor is sufficiently powerful that
       data structures can be implemented as a set of macros. The macros
       `STRUC' and `ENDSTRUC' are used to define a structure data type.

       `STRUC' takes one or two parameters. The first parameter is the name
       of the data type. The second, optional parameter is the base offset
       of the structure. The name of the data type is defined as a symbol
       with the value of the base offset, and the name of the data type
       with the suffix `_size' appended to it is defined as an `EQU' giving
       the size of the structure. Once `STRUC' has been issued, you are
       defining the structure, and should define fields using the `RESB'
       family of pseudo-instructions, and then invoke `ENDSTRUC' to finish
       the definition.

       For example, to define a structure called `mytype' containing a
       longword, a word, a byte and a string of bytes, you might code

       struc   mytype
      
         mt_long:      resd    1
         mt_word:      resw    1
         mt_byte:      resb    1
         mt_str:       resb    32
      
       endstruc

       The above code defines six symbols: `mt_long' as 0 (the offset from
       the beginning of a `mytype' structure to the longword field),
       `mt_word' as 4, `mt_byte' as 6, `mt_str' as 7, `mytype_size' as 39,
       and `mytype' itself as zero.

       The reason why the structure type name is defined at zero by default
       is a side effect of allowing structures to work with the local label
       mechanism: if your structure members tend to have the same names in
       more than one structure, you can define the above structure like
       this:

       struc mytype
      
         .long:        resd    1
         .word:        resw    1
         .byte:        resb    1
         .str:         resb    32
      
       endstruc

       This defines the offsets to the structure fields as `mytype.long',
       `mytype.word', `mytype.byte' and `mytype.str'.

       NASM, since it has no _intrinsic_ structure support, does not
       support any form of period notation to refer to the elements of a
       structure once you have one (except the above local-label notation),
       so code such as `mov ax,[mystruc.mt_word]' is not valid. `mt_word'
       is a constant just like any other constant, so the correct syntax is
       `mov ax,[mystruc+mt_word]' or `mov ax,[mystruc+mytype.word]'.

       Sometimes you only have the address of the structure displaced by an
       offset. For example, consider this standard stack frame setup:

       push ebp
       mov ebp, esp
       sub esp, 40

       In this case, you could access an element by subtracting the offset:

       mov [ebp - 40 + mytype.word], ax

       However, if you do not want to repeat this offset, you can use -40
       as a base offset:

       struc mytype, -40

       And access an element this way:

       mov [ebp + mytype.word], ax

4.12.11 `ISTRUC', `AT' and `IEND': Declaring Instances of Structures

       Having defined a structure type, the next thing you typically want
       to do is to declare instances of that structure in your data
       segment. NASM provides an easy way to do this in the `ISTRUC'
       mechanism. To declare a structure of type `mytype' in a program, you
       code something like this:

       mystruc:
           istruc mytype
      
               at mt_long, dd      123456
               at mt_word, dw      1024
               at mt_byte, db      'x'
               at mt_str,  db      'hello, world', 13, 10, 0
      
           iend

       The function of the `AT' macro is to make use of the `TIMES' prefix
       to advance the assembly position to the correct point for the
       specified structure field, and then to declare the specified data.
       Therefore the structure fields must be declared in the same order as
       they were specified in the structure definition.

       If the data to go in a structure field requires more than one source
       line to specify, the remaining source lines can easily come after
       the `AT' line. For example:

               at mt_str,  db      123,134,145,156,167,178,189
                           db      190,100,0

       Depending on personal taste, you can also omit the code part of the
       `AT' line completely, and start the structure field on the next
       line:

               at mt_str
                       db      'hello, world'
                       db      13,10,0

4.12.12 `ALIGN' and `ALIGNB': Data Alignment

       The `ALIGN' and `ALIGNB' macros provides a convenient way to align
       code or data on a word, longword, paragraph or other boundary. (Some
       assemblers call this directive `EVEN'.) The syntax of the `ALIGN'
       and `ALIGNB' macros is

               align   4               ; align on 4-byte boundary
               align   16              ; align on 16-byte boundary
               align   8,db 0          ; pad with 0s rather than NOPs
               align   4,resb 1        ; align to 4 in the BSS
               alignb  4               ; equivalent to previous line

       Both macros require their first argument to be a power of two; they
       both compute the number of additional bytes required to bring the
       length of the current section up to a multiple of that power of two,
       and then apply the `TIMES' prefix to their second argument to
       perform the alignment.

       If the second argument is not specified, the default for `ALIGN' is
       `NOP', and the default for `ALIGNB' is `RESB 1'. So if the second
       argument is specified, the two macros are equivalent. Normally, you
       can just use `ALIGN' in code and data sections and `ALIGNB' in BSS
       sections, and never need the second argument except for special
       purposes.

       `ALIGN' and `ALIGNB', being simple macros, perform no error
       checking: they cannot warn you if their first argument fails to be a
       power of two, or if their second argument generates more than one
       byte of code. In each of these cases they will silently do the wrong
       thing.

       `ALIGNB' (or `ALIGN' with a second argument of `RESB 1') can be used
       within structure definitions:

       struc mytype2
      
         mt_byte:
               resb 1
               alignb 2
         mt_word:
               resw 1
               alignb 4
         mt_long:
               resd 1
         mt_str:
               resb 32
      
       endstruc

       This will ensure that the structure members are sensibly aligned
       relative to the base of the structure.

       A final caveat: `ALIGN' and `ALIGNB' work relative to the beginning
       of the _section_, not the beginning of the address space in the
       final executable. Aligning to a 16-byte boundary when the section
       you're in is only guaranteed to be aligned to a 4-byte boundary, for
       example, is a waste of effort. Again, NASM does not check that the
       section's alignment characteristics are sensible for the use of
       `ALIGN' or `ALIGNB'.

       Both `ALIGN' and `ALIGNB' do call `SECTALIGN' macro implicitly. See
       section 4.12.13 for details.

       See also the `smartalign' standard macro package, section 5.2.

4.12.13 `SECTALIGN': Section Alignment

       The `SECTALIGN' macros provides a way to modify alignment attribute
       of output file section. Unlike the `align=' attribute (which is
       allowed at section definition only) the `SECTALIGN' macro may be
       used at any time.

       For example the directive

       SECTALIGN 16

       sets the section alignment requirements to 16 bytes. Once increased
       it can not be decreased, the magnitude may grow only.

       Note that `ALIGN' (see section 4.12.12) calls the `SECTALIGN' macro
       implicitly so the active section alignment requirements may be
       updated. This is by default behaviour, if for some reason you want
       the `ALIGN' do not call `SECTALIGN' at all use the directive

       SECTALIGN OFF

       It is still possible to turn in on again by

       SECTALIGN ON

Chapter 5: Standard Macro Packages
----------------------------------

       The `%use' directive (see section 4.6.4) includes one of the
       standard macro packages included with the NASM distribution and
       compiled into the NASM binary. It operates like the `%include'
       directive (see section 4.6.1), but the included contents is provided
       by NASM itself.

       The names of standard macro packages are case insensitive, and can
       be quoted or not.

   5.1 `altreg': Alternate Register Names

       The `altreg' standard macro package provides alternate register
       names. It provides numeric register names for all registers (not
       just `R8'-`R15'), the Intel-defined aliases `R8L'-`R15L' for the low
       bytes of register (as opposed to the NASM/AMD standard names `R8B'-
       `R15B'), and the names `R0H'-`R3H' (by analogy with `R0L'-`R3L') for
       `AH', `CH', `DH', and `BH'.

       Example use:

       %use altreg
      
       proc:
             mov r0l,r3h                    ; mov al,bh
             ret

       See also section 11.1.

   5.2 `smartalign': Smart `ALIGN' Macro

       The `smartalign' standard macro package provides for an `ALIGN'
       macro which is more powerful than the default (and backwards-
       compatible) one (see section 4.12.12). When the `smartalign' package
       is enabled, when `ALIGN' is used without a second argument, NASM
       will generate a sequence of instructions more efficient than a
       series of `NOP'. Furthermore, if the padding exceeds a specific
       threshold, then NASM will generate a jump over the entire padding
       sequence.

       The specific instructions generated can be controlled with the new
       `ALIGNMODE' macro. This macro takes two parameters: one mode, and an
       optional jump threshold override. If (for any reason) you need to
       turn off the jump completely just set jump threshold value to -1 (or
       set it to `nojmp'). The following modes are possible:

       (*) `generic': Works on all x86 CPUs and should have reasonable
           performance. The default jump threshold is 8. This is the
           default.

       (*) `nop': Pad out with `NOP' instructions. The only difference
           compared to the standard `ALIGN' macro is that NASM can still
           jump over a large padding area. The default jump threshold is
           16.

       (*) `k7': Optimize for the AMD K7 (Athlon/Althon XP). These
           instructions should still work on all x86 CPUs. The default jump
           threshold is 16.

       (*) `k8': Optimize for the AMD K8 (Opteron/Althon 64). These
           instructions should still work on all x86 CPUs. The default jump
           threshold is 16.

       (*) `p6': Optimize for Intel CPUs. This uses the long `NOP'
           instructions first introduced in Pentium Pro. This is
           incompatible with all CPUs of family 5 or lower, as well as some
           VIA CPUs and several virtualization solutions. The default jump
           threshold is 16.

       The macro `__ALIGNMODE__' is defined to contain the current
       alignment mode. A number of other macros beginning with `__ALIGN_'
       are used internally by this macro package.

   5.3 `fp': Floating-point macros

       This packages contains the following floating-point convenience
       macros:

       %define Inf             __Infinity__
       %define NaN             __QNaN__
       %define QNaN            __QNaN__
       %define SNaN            __SNaN__
        
       %define float8(x)       __float8__(x)
       %define float16(x)      __float16__(x)
       %define float32(x)      __float32__(x)
       %define float64(x)      __float64__(x)
       %define float80m(x)     __float80m__(x)
       %define float80e(x)     __float80e__(x)
       %define float128l(x)    __float128l__(x)
       %define float128h(x)    __float128h__(x)

   5.4 `ifunc': Integer functions

       This package contains a set of macros which implement integer
       functions. These are actually implemented as special operators, but
       are most conveniently accessed via this macro package.

       The macros provided are:

5.4.1 Integer logarithms

       These functions calculate the integer logarithm base 2 of their
       argument, considered as an unsigned integer. The only differences
       between the functions is their behavior if the argument provided is
       not a power of two.

       The function `ilog2e()' (alias `ilog2()') generate an error if the
       argument is not a power of two.

       The function `ilog2w()' generate a warning if the argument is not a
       power of two.

       The function `ilog2f()' rounds the argument down to the nearest
       power of two; if the argument is zero it returns zero.

       The function `ilog2c()' rounds the argument up to the nearest power
       of two.

Chapter 6: Assembler Directives
-------------------------------

       NASM, though it attempts to avoid the bureaucracy of assemblers like
       MASM and TASM, is nevertheless forced to support a _few_ directives.
       These are described in this chapter.

       NASM's directives come in two types: _user-level_ directives and
       _primitive_ directives. Typically, each directive has a user-level
       form and a primitive form. In almost all cases, we recommend that
       users use the user-level forms of the directives, which are
       implemented as macros which call the primitive forms.

       Primitive directives are enclosed in square brackets; user-level
       directives are not.

       In addition to the universal directives described in this chapter,
       each object file format can optionally supply extra directives in
       order to control particular features of that file format. These
       _format-specific_ directives are documented along with the formats
       that implement them, in chapter 7.

   6.1 `BITS': Specifying Target Processor Mode

       The `BITS' directive specifies whether NASM should generate code
       designed to run on a processor operating in 16-bit mode, 32-bit mode
       or 64-bit mode. The syntax is `BITS XX', where XX is 16, 32 or 64.

       In most cases, you should not need to use `BITS' explicitly. The
       `aout', `coff', `elf', `macho', `win32' and `win64' object formats,
       which are designed for use in 32-bit or 64-bit operating systems,
       all cause NASM to select 32-bit or 64-bit mode, respectively, by
       default. The `obj' object format allows you to specify each segment
       you define as either `USE16' or `USE32', and NASM will set its
       operating mode accordingly, so the use of the `BITS' directive is
       once again unnecessary.

       The most likely reason for using the `BITS' directive is to write
       32-bit or 64-bit code in a flat binary file; this is because the
       `bin' output format defaults to 16-bit mode in anticipation of it
       being used most frequently to write DOS `.COM' programs, DOS `.SYS'
       device drivers and boot loader software.

       You do _not_ need to specify `BITS 32' merely in order to use 32-bit
       instructions in a 16-bit DOS program; if you do, the assembler will
       generate incorrect code because it will be writing code targeted at
       a 32-bit platform, to be run on a 16-bit one.

       When NASM is in `BITS 16' mode, instructions which use 32-bit data
       are prefixed with an 0x66 byte, and those referring to 32-bit
       addresses have an 0x67 prefix. In `BITS 32' mode, the reverse is
       true: 32-bit instructions require no prefixes, whereas instructions
       using 16-bit data need an 0x66 and those working on 16-bit addresses
       need an 0x67.

       When NASM is in `BITS 64' mode, most instructions operate the same
       as they do for `BITS 32' mode. However, there are 8 more general and
       SSE registers, and 16-bit addressing is no longer supported.

       The default address size is 64 bits; 32-bit addressing can be
       selected with the 0x67 prefix. The default operand size is still 32
       bits, however, and the 0x66 prefix selects 16-bit operand size. The
       `REX' prefix is used both to select 64-bit operand size, and to
       access the new registers. NASM automatically inserts REX prefixes
       when necessary.

       When the `REX' prefix is used, the processor does not know how to
       address the AH, BH, CH or DH (high 8-bit legacy) registers. Instead,
       it is possible to access the the low 8-bits of the SP, BP SI and DI
       registers as SPL, BPL, SIL and DIL, respectively; but only when the
       REX prefix is used.

       The `BITS' directive has an exactly equivalent primitive form,
       `[BITS 16]', `[BITS 32]' and `[BITS 64]'. The user-level form is a
       macro which has no function other than to call the primitive form.

       Note that the space is neccessary, e.g. `BITS32' will _not_ work!

6.1.1 `USE16' & `USE32': Aliases for BITS

       The ``USE16'' and ``USE32'' directives can be used in place of
       ``BITS 16'' and ``BITS 32'', for compatibility with other
       assemblers.

   6.2 `DEFAULT': Change the assembler defaults

       The `DEFAULT' directive changes the assembler defaults. Normally,
       NASM defaults to a mode where the programmer is expected to
       explicitly specify most features directly. However, this is
       occationally obnoxious, as the explicit form is pretty much the only
       one one wishes to use.

       Currently, `DEFAULT' can set `REL' & `ABS' and `BND' & `NOBND'.

6.2.1 `REL' & `ABS': RIP-relative addressing

       This sets whether registerless instructions in 64-bit mode are
       `RIP'-relative or not. By default, they are absolute unless
       overridden with the `REL' specifier (see section 3.3). However, if
       `DEFAULT REL' is specified, `REL' is default, unless overridden with
       the `ABS' specifier, _except when used with an FS or GS segment
       override_.

       The special handling of `FS' and `GS' overrides are due to the fact
       that these registers are generally used as thread pointers or other
       special functions in 64-bit mode, and generating `RIP'-relative
       addresses would be extremely confusing.

       `DEFAULT REL' is disabled with `DEFAULT ABS'.

6.2.2 `BND' & `NOBND': `BND' prefix

       If `DEFAULT BND' is set, all bnd-prefix available instructions
       following this directive are prefixed with bnd. To override it,
       `NOBND' prefix can be used.

        DEFAULT BND
            call foo            ; BND will be prefixed
            nobnd call foo      ; BND will NOT be prefixed

       `DEFAULT NOBND' can disable `DEFAULT BND' and then `BND' prefix will
       be added only when explicitly specified in code.

       `DEFAULT BND' is expected to be the normal configuration for writing
       MPX-enabled code.

   6.3 `SECTION' or `SEGMENT': Changing and Defining Sections

       The `SECTION' directive (`SEGMENT' is an exactly equivalent synonym)
       changes which section of the output file the code you write will be
       assembled into. In some object file formats, the number and names of
       sections are fixed; in others, the user may make up as many as they
       wish. Hence `SECTION' may sometimes give an error message, or may
       define a new section, if you try to switch to a section that does
       not (yet) exist.

       The Unix object formats, and the `bin' object format (but see
       section 7.1.3, all support the standardized section names `.text',
       `.data' and `.bss' for the code, data and uninitialized-data
       sections. The `obj' format, by contrast, does not recognize these
       section names as being special, and indeed will strip off the
       leading period of any section name that has one.

6.3.1 The `__SECT__' Macro

       The `SECTION' directive is unusual in that its user-level form
       functions differently from its primitive form. The primitive form,
       `[SECTION xyz]', simply switches the current target section to the
       one given. The user-level form, `SECTION xyz', however, first
       defines the single-line macro `__SECT__' to be the primitive
       `[SECTION]' directive which it is about to issue, and then issues
       it. So the user-level directive

               SECTION .text

       expands to the two lines

       %define __SECT__        [SECTION .text]
               [SECTION .text]

       Users may find it useful to make use of this in their own macros.
       For example, the `writefile' macro defined in section 4.3.3 can be
       usefully rewritten in the following more sophisticated form:

       %macro  writefile 2+
      
               [section .data]
      
         %%str:        db      %2
         %%endstr:
      
               __SECT__
      
               mov     dx,%%str
               mov     cx,%%endstr-%%str
               mov     bx,%1
               mov     ah,0x40
               int     0x21
      
       %endmacro

       This form of the macro, once passed a string to output, first
       switches temporarily to the data section of the file, using the
       primitive form of the `SECTION' directive so as not to modify
       `__SECT__'. It then declares its string in the data section, and
       then invokes `__SECT__' to switch back to _whichever_ section the
       user was previously working in. It thus avoids the need, in the
       previous version of the macro, to include a `JMP' instruction to
       jump over the data, and also does not fail if, in a complicated
       `OBJ' format module, the user could potentially be assembling the
       code in any of several separate code sections.

   6.4 `ABSOLUTE': Defining Absolute Labels

       The `ABSOLUTE' directive can be thought of as an alternative form of
       `SECTION': it causes the subsequent code to be directed at no
       physical section, but at the hypothetical section starting at the
       given absolute address. The only instructions you can use in this
       mode are the `RESB' family.

       `ABSOLUTE' is used as follows:

       absolute 0x1A
      
           kbuf_chr    resw    1
           kbuf_free   resw    1
           kbuf        resw    16

       This example describes a section of the PC BIOS data area, at
       segment address 0x40: the above code defines `kbuf_chr' to be 0x1A,
       `kbuf_free' to be 0x1C, and `kbuf' to be 0x1E.

       The user-level form of `ABSOLUTE', like that of `SECTION', redefines
       the `__SECT__' macro when it is invoked.

       `STRUC' and `ENDSTRUC' are defined as macros which use `ABSOLUTE'
       (and also `__SECT__').

       `ABSOLUTE' doesn't have to take an absolute constant as an argument:
       it can take an expression (actually, a critical expression: see
       section 3.8) and it can be a value in a segment. For example, a TSR
       can re-use its setup code as run-time BSS like this:

               org     100h               ; it's a .COM program
      
               jmp     setup              ; setup code comes last
      
               ; the resident part of the TSR goes here
       setup:
               ; now write the code that installs the TSR here
      
       absolute setup
      
       runtimevar1     resw    1
       runtimevar2     resd    20
      
       tsr_end:

       This defines some variables `on top of' the setup code, so that
       after the setup has finished running, the space it took up can be
       re-used as data storage for the running TSR. The symbol `tsr_end'
       can be used to calculate the total size of the part of the TSR that
       needs to be made resident.

   6.5 `EXTERN': Importing Symbols from Other Modules

       `EXTERN' is similar to the MASM directive `EXTRN' and the C keyword
       `extern': it is used to declare a symbol which is not defined
       anywhere in the module being assembled, but is assumed to be defined
       in some other module and needs to be referred to by this one. Not
       every object-file format can support external variables: the `bin'
       format cannot.

       The `EXTERN' directive takes as many arguments as you like. Each
       argument is the name of a symbol:

       extern  _printf
       extern  _sscanf,_fscanf

       Some object-file formats provide extra features to the `EXTERN'
       directive. In all cases, the extra features are used by suffixing a
       colon to the symbol name followed by object-format specific text.
       For example, the `obj' format allows you to declare that the default
       segment base of an external should be the group `dgroup' by means of
       the directive

       extern  _variable:wrt dgroup

       The primitive form of `EXTERN' differs from the user-level form only
       in that it can take only one argument at a time: the support for
       multiple arguments is implemented at the preprocessor level.

       You can declare the same variable as `EXTERN' more than once: NASM
       will quietly ignore the second and later redeclarations. You can't
       declare a variable as `EXTERN' as well as something else, though.

   6.6 `GLOBAL': Exporting Symbols to Other Modules

       `GLOBAL' is the other end of `EXTERN': if one module declares a
       symbol as `EXTERN' and refers to it, then in order to prevent linker
       errors, some other module must actually _define_ the symbol and
       declare it as `GLOBAL'. Some assemblers use the name `PUBLIC' for
       this purpose.

       The `GLOBAL' directive applying to a symbol must appear _before_ the
       definition of the symbol.

       `GLOBAL' uses the same syntax as `EXTERN', except that it must refer
       to symbols which _are_ defined in the same module as the `GLOBAL'
       directive. For example:

       global _main
       _main:
               ; some code

       `GLOBAL', like `EXTERN', allows object formats to define private
       extensions by means of a colon. The `elf' object format, for
       example, lets you specify whether global data items are functions or
       data:

       global  hashlookup:function, hashtable:data

       Like `EXTERN', the primitive form of `GLOBAL' differs from the user-
       level form only in that it can take only one argument at a time.

   6.7 `COMMON': Defining Common Data Areas

       The `COMMON' directive is used to declare _common variables_. A
       common variable is much like a global variable declared in the
       uninitialized data section, so that

       common  intvar  4

       is similar in function to

       global  intvar
       section .bss
      
       intvar  resd    1

       The difference is that if more than one module defines the same
       common variable, then at link time those variables will be _merged_,
       and references to `intvar' in all modules will point at the same
       piece of memory.

       Like `GLOBAL' and `EXTERN', `COMMON' supports object-format specific
       extensions. For example, the `obj' format allows common variables to
       be NEAR or FAR, and the `elf' format allows you to specify the
       alignment requirements of a common variable:

       common  commvar  4:near  ; works in OBJ
       common  intarray 100:4   ; works in ELF: 4 byte aligned

       Once again, like `EXTERN' and `GLOBAL', the primitive form of
       `COMMON' differs from the user-level form only in that it can take
       only one argument at a time.

   6.8 `CPU': Defining CPU Dependencies

       The `CPU' directive restricts assembly to those instructions which
       are available on the specified CPU.

       Options are:

       (*) `CPU 8086' Assemble only 8086 instruction set

       (*) `CPU 186' Assemble instructions up to the 80186 instruction set

       (*) `CPU 286' Assemble instructions up to the 286 instruction set

       (*) `CPU 386' Assemble instructions up to the 386 instruction set

       (*) `CPU 486' 486 instruction set

       (*) `CPU 586' Pentium instruction set

       (*) `CPU PENTIUM' Same as 586

       (*) `CPU 686' P6 instruction set

       (*) `CPU PPRO' Same as 686

       (*) `CPU P2' Same as 686

       (*) `CPU P3' Pentium III (Katmai) instruction sets

       (*) `CPU KATMAI' Same as P3

       (*) `CPU P4' Pentium 4 (Willamette) instruction set

       (*) `CPU WILLAMETTE' Same as P4

       (*) `CPU PRESCOTT' Prescott instruction set

       (*) `CPU X64' x86-64 (x64/AMD64/Intel 64) instruction set

       (*) `CPU IA64' IA64 CPU (in x86 mode) instruction set

       All options are case insensitive. All instructions will be selected
       only if they apply to the selected CPU or lower. By default, all
       instructions are available.

   6.9 `FLOAT': Handling of floating-point constants

       By default, floating-point constants are rounded to nearest, and
       IEEE denormals are supported. The following options can be set to
       alter this behaviour:

       (*) `FLOAT DAZ' Flush denormals to zero

       (*) `FLOAT NODAZ' Do not flush denormals to zero (default)

       (*) `FLOAT NEAR' Round to nearest (default)

       (*) `FLOAT UP' Round up (toward +Infinity)

       (*) `FLOAT DOWN' Round down (toward -Infinity)

       (*) `FLOAT ZERO' Round toward zero

       (*) `FLOAT DEFAULT' Restore default settings

       The standard macros `__FLOAT_DAZ__', `__FLOAT_ROUND__', and
       `__FLOAT__' contain the current state, as long as the programmer has
       avoided the use of the brackeded primitive form, (`[FLOAT]').

       `__FLOAT__' contains the full set of floating-point settings; this
       value can be saved away and invoked later to restore the setting.

Chapter 7: Output Formats
-------------------------

       NASM is a portable assembler, designed to be able to compile on any
       ANSI C-supporting platform and produce output to run on a variety of
       Intel x86 operating systems. For this reason, it has a large number
       of available output formats, selected using the `-f' option on the
       NASM command line. Each of these formats, along with its extensions
       to the base NASM syntax, is detailed in this chapter.

       As stated in section 2.1.1, NASM chooses a default name for your
       output file based on the input file name and the chosen output
       format. This will be generated by removing the extension (`.asm',
       `.s', or whatever you like to use) from the input file name, and
       substituting an extension defined by the output format. The
       extensions are given with each format below.

   7.1 `bin': Flat-Form Binary Output

       The `bin' format does not produce object files: it generates nothing
       in the output file except the code you wrote. Such `pure binary'
       files are used by MS-DOS: `.COM' executables and `.SYS' device
       drivers are pure binary files. Pure binary output is also useful for
       operating system and boot loader development.

       The `bin' format supports multiple section names. For details of how
       NASM handles sections in the `bin' format, see section 7.1.3.

       Using the `bin' format puts NASM by default into 16-bit mode (see
       section 6.1). In order to use `bin' to write 32-bit or 64-bit code,
       such as an OS kernel, you need to explicitly issue the `BITS 32' or
       `BITS 64' directive.

       `bin' has no default output file name extension: instead, it leaves
       your file name as it is once the original extension has been
       removed. Thus, the default is for NASM to assemble `binprog.asm'
       into a binary file called `binprog'.

7.1.1 `ORG': Binary File Program Origin

       The `bin' format provides an additional directive to the list given
       in chapter 6: `ORG'. The function of the `ORG' directive is to
       specify the origin address which NASM will assume the program begins
       at when it is loaded into memory.

       For example, the following code will generate the longword
       `0x00000104':

               org     0x100
               dd      label
       label:

       Unlike the `ORG' directive provided by MASM-compatible assemblers,
       which allows you to jump around in the object file and overwrite
       code you have already generated, NASM's `ORG' does exactly what the
       directive says: _origin_. Its sole function is to specify one offset
       which is added to all internal address references within the
       section; it does not permit any of the trickery that MASM's version
       does. See section 12.1.3 for further comments.

7.1.2 `bin' Extensions to the `SECTION' Directive

       The `bin' output format extends the `SECTION' (or `SEGMENT')
       directive to allow you to specify the alignment requirements of
       segments. This is done by appending the `ALIGN' qualifier to the end
       of the section-definition line. For example,

       section .data   align=16

       switches to the section `.data' and also specifies that it must be
       aligned on a 16-byte boundary.

       The parameter to `ALIGN' specifies how many low bits of the section
       start address must be forced to zero. The alignment value given may
       be any power of two.

7.1.3 Multisection Support for the `bin' Format

       The `bin' format allows the use of multiple sections, of arbitrary
       names, besides the "known" `.text', `.data', and `.bss' names.

       (*) Sections may be designated `progbits' or `nobits'. Default is
           `progbits' (except `.bss', which defaults to `nobits', of
           course).

       (*) Sections can be aligned at a specified boundary following the
           previous section with `align=', or at an arbitrary byte-granular
           position with `start='.

       (*) Sections can be given a virtual start address, which will be
           used for the calculation of all memory references within that
           section with `vstart='.

       (*) Sections can be ordered using `follows='`<section>' or
           `vfollows='`<section>' as an alternative to specifying an
           explicit start address.

       (*) Arguments to `org', `start', `vstart', and `align=' are critical
           expressions. See section 3.8. E.g. `align=(1 << ALIGN_SHIFT)' -
           `ALIGN_SHIFT' must be defined before it is used here.

       (*) Any code which comes before an explicit `SECTION' directive is
           directed by default into the `.text' section.

       (*) If an `ORG' statement is not given, `ORG 0' is used by default.

       (*) The `.bss' section will be placed after the last `progbits'
           section, unless `start=', `vstart=', `follows=', or `vfollows='
           has been specified.

       (*) All sections are aligned on dword boundaries, unless a different
           alignment has been specified.

       (*) Sections may not overlap.

       (*) NASM creates the `section.<secname>.start' for each section,
           which may be used in your code.

7.1.4 Map Files

       Map files can be generated in `-f bin' format by means of the
       `[map]' option. Map types of `all' (default), `brief', `sections',
       `segments', or `symbols' may be specified. Output may be directed to
       `stdout' (default), `stderr', or a specified file. E.g.
       `[map symbols myfile.map]'. No "user form" exists, the square
       brackets must be used.

   7.2 `ith': Intel Hex Output

       The `ith' file format produces Intel hex-format files. Just as the
       `bin' format, this is a flat memory image format with no support for
       relocation or linking. It is usually used with ROM programmers and
       similar utilities.

       All extensions supported by the `bin' file format is also supported
       by the `ith' file format.

       `ith' provides a default output file-name extension of `.ith'.

   7.3 `srec': Motorola S-Records Output

       The `srec' file format produces Motorola S-records files. Just as
       the `bin' format, this is a flat memory image format with no support
       for relocation or linking. It is usually used with ROM programmers
       and similar utilities.

       All extensions supported by the `bin' file format is also supported
       by the `srec' file format.

       `srec' provides a default output file-name extension of `.srec'.

   7.4 `obj': Microsoft OMF Object Files

       The `obj' file format (NASM calls it `obj' rather than `omf' for
       historical reasons) is the one produced by MASM and TASM, which is
       typically fed to 16-bit DOS linkers to produce `.EXE' files. It is
       also the format used by OS/2.

       `obj' provides a default output file-name extension of `.obj'.

       `obj' is not exclusively a 16-bit format, though: NASM has full
       support for the 32-bit extensions to the format. In particular, 32-
       bit `obj' format files are used by Borland's Win32 compilers,
       instead of using Microsoft's newer `win32' object file format.

       The `obj' format does not define any special segment names: you can
       call your segments anything you like. Typical names for segments in
       `obj' format files are `CODE', `DATA' and `BSS'.

       If your source file contains code before specifying an explicit
       `SEGMENT' directive, then NASM will invent its own segment called
       `__NASMDEFSEG' for you.

       When you define a segment in an `obj' file, NASM defines the segment
       name as a symbol as well, so that you can access the segment address
       of the segment. So, for example:

       segment data
      
       dvar:   dw      1234
      
       segment code
      
       function:
               mov     ax,data         ; get segment address of data
               mov     ds,ax           ; and move it into DS
               inc     word [dvar]     ; now this reference will work
               ret

       The `obj' format also enables the use of the `SEG' and `WRT'
       operators, so that you can write code which does things like

       extern  foo
      
             mov   ax,seg foo            ; get preferred segment of foo
             mov   ds,ax
             mov   ax,data               ; a different segment
             mov   es,ax
             mov   ax,[ds:foo]           ; this accesses `foo'
             mov   [es:foo wrt data],bx  ; so does this

7.4.1 `obj' Extensions to the `SEGMENT' Directive

       The `obj' output format extends the `SEGMENT' (or `SECTION')
       directive to allow you to specify various properties of the segment
       you are defining. This is done by appending extra qualifiers to the
       end of the segment-definition line. For example,

       segment code private align=16

       defines the segment `code', but also declares it to be a private
       segment, and requires that the portion of it described in this code
       module must be aligned on a 16-byte boundary.

       The available qualifiers are:

       (*) `PRIVATE', `PUBLIC', `COMMON' and `STACK' specify the
           combination characteristics of the segment. `PRIVATE' segments
           do not get combined with any others by the linker; `PUBLIC' and
           `STACK' segments get concatenated together at link time; and
           `COMMON' segments all get overlaid on top of each other rather
           than stuck end-to-end.

       (*) `ALIGN' is used, as shown above, to specify how many low bits of
           the segment start address must be forced to zero. The alignment
           value given may be any power of two from 1 to 4096; in reality,
           the only values supported are 1, 2, 4, 16, 256 and 4096, so if 8
           is specified it will be rounded up to 16, and 32, 64 and 128
           will all be rounded up to 256, and so on. Note that alignment to
           4096-byte boundaries is a PharLap extension to the format and
           may not be supported by all linkers.

       (*) `CLASS' can be used to specify the segment class; this feature
           indicates to the linker that segments of the same class should
           be placed near each other in the output file. The class name can
           be any word, e.g. `CLASS=CODE'.

       (*) `OVERLAY', like `CLASS', is specified with an arbitrary word as
           an argument, and provides overlay information to an overlay-
           capable linker.

       (*) Segments can be declared as `USE16' or `USE32', which has the
           effect of recording the choice in the object file and also
           ensuring that NASM's default assembly mode when assembling in
           that segment is 16-bit or 32-bit respectively.

       (*) When writing OS/2 object files, you should declare 32-bit
           segments as `FLAT', which causes the default segment base for
           anything in the segment to be the special group `FLAT', and also
           defines the group if it is not already defined.

       (*) The `obj' file format also allows segments to be declared as
           having a pre-defined absolute segment address, although no
           linkers are currently known to make sensible use of this
           feature; nevertheless, NASM allows you to declare a segment such
           as `SEGMENT SCREEN ABSOLUTE=0xB800' if you need to. The
           `ABSOLUTE' and `ALIGN' keywords are mutually exclusive.

       NASM's default segment attributes are `PUBLIC', `ALIGN=1', no class,
       no overlay, and `USE16'.

7.4.2 `GROUP': Defining Groups of Segments

       The `obj' format also allows segments to be grouped, so that a
       single segment register can be used to refer to all the segments in
       a group. NASM therefore supplies the `GROUP' directive, whereby you
       can code

       segment data
      
               ; some data
      
       segment bss
      
               ; some uninitialized data
      
       group dgroup data bss

       which will define a group called `dgroup' to contain the segments
       `data' and `bss'. Like `SEGMENT', `GROUP' causes the group name to
       be defined as a symbol, so that you can refer to a variable `var' in
       the `data' segment as `var wrt data' or as `var wrt dgroup',
       depending on which segment value is currently in your segment
       register.

       If you just refer to `var', however, and `var' is declared in a
       segment which is part of a group, then NASM will default to giving
       you the offset of `var' from the beginning of the _group_, not the
       _segment_. Therefore `SEG var', also, will return the group base
       rather than the segment base.

       NASM will allow a segment to be part of more than one group, but
       will generate a warning if you do this. Variables declared in a
       segment which is part of more than one group will default to being
       relative to the first group that was defined to contain the segment.

       A group does not have to contain any segments; you can still make
       `WRT' references to a group which does not contain the variable you
       are referring to. OS/2, for example, defines the special group
       `FLAT' with no segments in it.

7.4.3 `UPPERCASE': Disabling Case Sensitivity in Output

       Although NASM itself is case sensitive, some OMF linkers are not;
       therefore it can be useful for NASM to output single-case object
       files. The `UPPERCASE' format-specific directive causes all segment,
       group and symbol names that are written to the object file to be
       forced to upper case just before being written. Within a source
       file, NASM is still case-sensitive; but the object file can be
       written entirely in upper case if desired.

       `UPPERCASE' is used alone on a line; it requires no parameters.

7.4.4 `IMPORT': Importing DLL Symbols

       The `IMPORT' format-specific directive defines a symbol to be
       imported from a DLL, for use if you are writing a DLL's import
       library in NASM. You still need to declare the symbol as `EXTERN' as
       well as using the `IMPORT' directive.

       The `IMPORT' directive takes two required parameters, separated by
       white space, which are (respectively) the name of the symbol you
       wish to import and the name of the library you wish to import it
       from. For example:

           import  WSAStartup wsock32.dll

       A third optional parameter gives the name by which the symbol is
       known in the library you are importing it from, in case this is not
       the same as the name you wish the symbol to be known by to your code
       once you have imported it. For example:

           import  asyncsel wsock32.dll WSAAsyncSelect

7.4.5 `EXPORT': Exporting DLL Symbols

       The `EXPORT' format-specific directive defines a global symbol to be
       exported as a DLL symbol, for use if you are writing a DLL in NASM.
       You still need to declare the symbol as `GLOBAL' as well as using
       the `EXPORT' directive.

       `EXPORT' takes one required parameter, which is the name of the
       symbol you wish to export, as it was defined in your source file. An
       optional second parameter (separated by white space from the first)
       gives the _external_ name of the symbol: the name by which you wish
       the symbol to be known to programs using the DLL. If this name is
       the same as the internal name, you may leave the second parameter
       off.

       Further parameters can be given to define attributes of the exported
       symbol. These parameters, like the second, are separated by white
       space. If further parameters are given, the external name must also
       be specified, even if it is the same as the internal name. The
       available attributes are:

       (*) `resident' indicates that the exported name is to be kept
           resident by the system loader. This is an optimisation for
           frequently used symbols imported by name.

       (*) `nodata' indicates that the exported symbol is a function which
           does not make use of any initialized data.

       (*) `parm=NNN', where `NNN' is an integer, sets the number of
           parameter words for the case in which the symbol is a call gate
           between 32-bit and 16-bit segments.

       (*) An attribute which is just a number indicates that the symbol
           should be exported with an identifying number (ordinal), and
           gives the desired number.

       For example:

           export  myfunc
           export  myfunc TheRealMoreFormalLookingFunctionName
           export  myfunc myfunc 1234  ; export by ordinal
           export  myfunc myfunc resident parm=23 nodata

7.4.6 `..start': Defining the Program Entry Point

       `OMF' linkers require exactly one of the object files being linked
       to define the program entry point, where execution will begin when
       the program is run. If the object file that defines the entry point
       is assembled using NASM, you specify the entry point by declaring
       the special symbol `..start' at the point where you wish execution
       to begin.

7.4.7 `obj' Extensions to the `EXTERN' Directive

       If you declare an external symbol with the directive

           extern  foo

       then references such as `mov ax,foo' will give you the offset of
       `foo' from its preferred segment base (as specified in whichever
       module `foo' is actually defined in). So to access the contents of
       `foo' you will usually need to do something like

               mov     ax,seg foo      ; get preferred segment base
               mov     es,ax           ; move it into ES
               mov     ax,[es:foo]     ; and use offset `foo' from it

       This is a little unwieldy, particularly if you know that an external
       is going to be accessible from a given segment or group, say
       `dgroup'. So if `DS' already contained `dgroup', you could simply
       code

               mov     ax,[foo wrt dgroup]

       However, having to type this every time you want to access `foo' can
       be a pain; so NASM allows you to declare `foo' in the alternative
       form

           extern  foo:wrt dgroup

       This form causes NASM to pretend that the preferred segment base of
       `foo' is in fact `dgroup'; so the expression `seg foo' will now
       return `dgroup', and the expression `foo' is equivalent to
       `foo wrt dgroup'.

       This default-`WRT' mechanism can be used to make externals appear to
       be relative to any group or segment in your program. It can also be
       applied to common variables: see section 7.4.8.

7.4.8 `obj' Extensions to the `COMMON' Directive

       The `obj' format allows common variables to be either near or far;
       NASM allows you to specify which your variables should be by the use
       of the syntax

       common  nearvar 2:near   ; `nearvar' is a near common
       common  farvar  10:far   ; and `farvar' is far

       Far common variables may be greater in size than 64Kb, and so the
       OMF specification says that they are declared as a number of
       _elements_ of a given size. So a 10-byte far common variable could
       be declared as ten one-byte elements, five two-byte elements, two
       five-byte elements or one ten-byte element.

       Some `OMF' linkers require the element size, as well as the variable
       size, to match when resolving common variables declared in more than
       one module. Therefore NASM must allow you to specify the element
       size on your far common variables. This is done by the following
       syntax:

       common  c_5by2  10:far 5        ; two five-byte elements
       common  c_2by5  10:far 2        ; five two-byte elements

       If no element size is specified, the default is 1. Also, the `FAR'
       keyword is not required when an element size is specified, since
       only far commons may have element sizes at all. So the above
       declarations could equivalently be

       common  c_5by2  10:5            ; two five-byte elements
       common  c_2by5  10:2            ; five two-byte elements

       In addition to these extensions, the `COMMON' directive in `obj'
       also supports default-`WRT' specification like `EXTERN' does
       (explained in section 7.4.7). So you can also declare things like

       common  foo     10:wrt dgroup
       common  bar     16:far 2:wrt data
       common  baz     24:wrt data:6

   7.5 `win32': Microsoft Win32 Object Files

       The `win32' output format generates Microsoft Win32 object files,
       suitable for passing to Microsoft linkers such as Visual C++. Note
       that Borland Win32 compilers do not use this format, but use `obj'
       instead (see section 7.4).

       `win32' provides a default output file-name extension of `.obj'.

       Note that although Microsoft say that Win32 object files follow the
       `COFF' (Common Object File Format) standard, the object files
       produced by Microsoft Win32 compilers are not compatible with COFF
       linkers such as DJGPP's, and vice versa. This is due to a difference
       of opinion over the precise semantics of PC-relative relocations. To
       produce COFF files suitable for DJGPP, use NASM's `coff' output
       format; conversely, the `coff' format does not produce object files
       that Win32 linkers can generate correct output from.

7.5.1 `win32' Extensions to the `SECTION' Directive

       Like the `obj' format, `win32' allows you to specify additional
       information on the `SECTION' directive line, to control the type and
       properties of sections you declare. Section types and properties are
       generated automatically by NASM for the standard section names
       `.text', `.data' and `.bss', but may still be overridden by these
       qualifiers.

       The available qualifiers are:

       (*) `code', or equivalently `text', defines the section to be a code
           section. This marks the section as readable and executable, but
           not writable, and also indicates to the linker that the type of
           the section is code.

       (*) `data' and `bss' define the section to be a data section,
           analogously to `code'. Data sections are marked as readable and
           writable, but not executable. `data' declares an initialized
           data section, whereas `bss' declares an uninitialized data
           section.

       (*) `rdata' declares an initialized data section that is readable
           but not writable. Microsoft compilers use this section to place
           constants in it.

       (*) `info' defines the section to be an informational section, which
           is not included in the executable file by the linker, but may
           (for example) pass information _to_ the linker. For example,
           declaring an `info'-type section called `.drectve' causes the
           linker to interpret the contents of the section as command-line
           options.

       (*) `align=', used with a trailing number as in `obj', gives the
           alignment requirements of the section. The maximum you may
           specify is 64: the Win32 object file format contains no means to
           request a greater section alignment than this. If alignment is
           not explicitly specified, the defaults are 16-byte alignment for
           code sections, 8-byte alignment for rdata sections and 4-byte
           alignment for data (and BSS) sections. Informational sections
           get a default alignment of 1 byte (no alignment), though the
           value does not matter.

       The defaults assumed by NASM if you do not specify the above
       qualifiers are:

       section .text    code  align=16
       section .data    data  align=4
       section .rdata   rdata align=8
       section .bss     bss   align=4

       Any other section name is treated by default like `.text'.

7.5.2 `win32': Safe Structured Exception Handling

       Among other improvements in Windows XP SP2 and Windows Server 2003
       Microsoft has introduced concept of "safe structured exception
       handling." General idea is to collect handlers' entry points in
       designated read-only table and have alleged entry point verified
       against this table prior exception control is passed to the handler.
       In order for an executable module to be equipped with such "safe
       exception handler table," all object modules on linker command line
       has to comply with certain criteria. If one single module among them
       does not, then the table in question is omitted and above mentioned
       run-time checks will not be performed for application in question.
       Table omission is by default silent and therefore can be easily
       overlooked. One can instruct linker to refuse to produce binary
       without such table by passing `/safeseh' command line option.

       Without regard to this run-time check merits it's natural to expect
       NASM to be capable of generating modules suitable for `/safeseh'
       linking. From developer's viewpoint the problem is two-fold:

       (*) how to adapt modules not deploying exception handlers of their
           own;

       (*) how to adapt/develop modules utilizing custom exception
           handling;

       Former can be easily achieved with any NASM version by adding
       following line to source code:

       $@feat.00 equ 1

       As of version 2.03 NASM adds this absolute symbol automatically. If
       it's not already present to be precise. I.e. if for whatever reason
       developer would choose to assign another value in source file, it
       would still be perfectly possible.

       Registering custom exception handler on the other hand requires
       certain "magic." As of version 2.03 additional directive is
       implemented, `safeseh', which instructs the assembler to produce
       appropriately formatted input data for above mentioned "safe
       exception handler table." Its typical use would be:

       section .text
       extern  _MessageBoxA@16
       %if     __NASM_VERSION_ID__ >= 0x02030000
       safeseh handler         ; register handler as "safe handler"
       %endif
       handler:
               push    DWORD 1 ; MB_OKCANCEL
               push    DWORD caption
               push    DWORD text
               push    DWORD 0
               call    _MessageBoxA@16
               sub     eax,1   ; incidentally suits as return value
                               ; for exception handler
               ret
       global  _main
       _main:
               push    DWORD handler
               push    DWORD [fs:0]
               mov     DWORD [fs:0],esp ; engage exception handler
               xor     eax,eax
               mov     eax,DWORD[eax]   ; cause exception
               pop     DWORD [fs:0]     ; disengage exception handler
               add     esp,4
               ret
       text:   db      'OK to rethrow, CANCEL to generate core dump',0
       caption:db      'SEGV',0
      
       section .drectve info
               db      '/defaultlib:user32.lib /defaultlib:msvcrt.lib '

       As you might imagine, it's perfectly possible to produce .exe binary
       with "safe exception handler table" and yet engage unregistered
       exception handler. Indeed, handler is engaged by simply manipulating
       `[fs:0]' location at run-time, something linker has no power over,
       run-time that is. It should be explicitly mentioned that such
       failure to register handler's entry point with `safeseh' directive
       has undesired side effect at run-time. If exception is raised and
       unregistered handler is to be executed, the application is abruptly
       terminated without any notification whatsoever. One can argue that
       system could at least have logged some kind "non-safe exception
       handler in x.exe at address n" message in event log, but no,
       literally no notification is provided and user is left with no clue
       on what caused application failure.

       Finally, all mentions of linker in this paragraph refer to Microsoft
       linker version 7.x and later. Presence of `@feat.00' symbol and
       input data for "safe exception handler table" causes no backward
       incompatibilities and "safeseh" modules generated by NASM 2.03 and
       later can still be linked by earlier versions or non-Microsoft
       linkers.

   7.6 `win64': Microsoft Win64 Object Files

       The `win64' output format generates Microsoft Win64 object files,
       which is nearly 100% identical to the `win32' object format (section
       7.5) with the exception that it is meant to target 64-bit code and
       the x86-64 platform altogether. This object file is used exactly the
       same as the `win32' object format (section 7.5), in NASM, with
       regard to this exception.

7.6.1 `win64': Writing Position-Independent Code

       While `REL' takes good care of RIP-relative addressing, there is one
       aspect that is easy to overlook for a Win64 programmer: indirect
       references. Consider a switch dispatch table:

               jmp     qword [dsptch+rax*8]
               ...
       dsptch: dq      case0
               dq      case1
               ...

       Even a novice Win64 assembler programmer will soon realize that the
       code is not 64-bit savvy. Most notably linker will refuse to link it
       with

       'ADDR32' relocation to '.text' invalid without /LARGEADDRESSAWARE:NO

       So [s]he will have to split jmp instruction as following:

               lea     rbx,[rel dsptch]
               jmp     qword [rbx+rax*8]

       What happens behind the scene is that effective address in `lea' is
       encoded relative to instruction pointer, or in perfectly position-
       independent manner. But this is only part of the problem! Trouble is
       that in .dll context `caseN' relocations will make their way to the
       final module and might have to be adjusted at .dll load time. To be
       specific when it can't be loaded at preferred address. And when this
       occurs, pages with such relocations will be rendered private to
       current process, which kind of undermines the idea of sharing .dll.
       But no worry, it's trivial to fix:

               lea     rbx,[rel dsptch]
               add     rbx,[rbx+rax*8]
               jmp     rbx
               ...
       dsptch: dq      case0-dsptch
               dq      case1-dsptch
               ...

       NASM version 2.03 and later provides another alternative,
       `wrt ..imagebase' operator, which returns offset from base address
       of the current image, be it .exe or .dll module, therefore the name.
       For those acquainted with PE-COFF format base address denotes start
       of `IMAGE_DOS_HEADER' structure. Here is how to implement switch
       with these image-relative references:

               lea     rbx,[rel dsptch]
               mov     eax,[rbx+rax*4]
               sub     rbx,dsptch wrt ..imagebase
               add     rbx,rax
               jmp     rbx
               ...
       dsptch: dd      case0 wrt ..imagebase
               dd      case1 wrt ..imagebase

       One can argue that the operator is redundant. Indeed, snippet before
       last works just fine with any NASM version and is not even Windows
       specific... The real reason for implementing `wrt ..imagebase' will
       become apparent in next paragraph.

       It should be noted that `wrt ..imagebase' is defined as 32-bit
       operand only:

               dd      label wrt ..imagebase           ; ok
               dq      label wrt ..imagebase           ; bad
               mov     eax,label wrt ..imagebase       ; ok
               mov     rax,label wrt ..imagebase       ; bad

7.6.2 `win64': Structured Exception Handling

       Structured exception handing in Win64 is completely different matter
       from Win32. Upon exception program counter value is noted, and
       linker-generated table comprising start and end addresses of all the
       functions [in given executable module] is traversed and compared to
       the saved program counter. Thus so called `UNWIND_INFO' structure is
       identified. If it's not found, then offending subroutine is assumed
       to be "leaf" and just mentioned lookup procedure is attempted for
       its caller. In Win64 leaf function is such function that does not
       call any other function _nor_ modifies any Win64 non-volatile
       registers, including stack pointer. The latter ensures that it's
       possible to identify leaf function's caller by simply pulling the
       value from the top of the stack.

       While majority of subroutines written in assembler are not calling
       any other function, requirement for non-volatile registers'
       immutability leaves developer with not more than 7 registers and no
       stack frame, which is not necessarily what [s]he counted with.
       Customarily one would meet the requirement by saving non-volatile
       registers on stack and restoring them upon return, so what can go
       wrong? If [and only if] an exception is raised at run-time and no
       `UNWIND_INFO' structure is associated with such "leaf" function, the
       stack unwind procedure will expect to find caller's return address
       on the top of stack immediately followed by its frame. Given that
       developer pushed caller's non-volatile registers on stack, would the
       value on top point at some code segment or even addressable space?
       Well, developer can attempt copying caller's return address to the
       top of stack and this would actually work in some very specific
       circumstances. But unless developer can guarantee that these
       circumstances are always met, it's more appropriate to assume worst
       case scenario, i.e. stack unwind procedure going berserk. Relevant
       question is what happens then? Application is abruptly terminated
       without any notification whatsoever. Just like in Win32 case, one
       can argue that system could at least have logged "unwind procedure
       went berserk in x.exe at address n" in event log, but no, no trace
       of failure is left.

       Now, when we understand significance of the `UNWIND_INFO' structure,
       let's discuss what's in it and/or how it's processed. First of all
       it is checked for presence of reference to custom language-specific
       exception handler. If there is one, then it's invoked. Depending on
       the return value, execution flow is resumed (exception is said to be
       "handled"), _or_ rest of `UNWIND_INFO' structure is processed as
       following. Beside optional reference to custom handler, it carries
       information about current callee's stack frame and where non-
       volatile registers are saved. Information is detailed enough to be
       able to reconstruct contents of caller's non-volatile registers upon
       call to current callee. And so caller's context is reconstructed,
       and then unwind procedure is repeated, i.e. another `UNWIND_INFO'
       structure is associated, this time, with caller's instruction
       pointer, which is then checked for presence of reference to
       language-specific handler, etc. The procedure is recursively
       repeated till exception is handled. As last resort system "handles"
       it by generating memory core dump and terminating the application.

       As for the moment of this writing NASM unfortunately does not
       facilitate generation of above mentioned detailed information about
       stack frame layout. But as of version 2.03 it implements building
       blocks for generating structures involved in stack unwinding. As
       simplest example, here is how to deploy custom exception handler for
       leaf function:

       default rel
       section .text
       extern  MessageBoxA
       handler:
               sub     rsp,40
               mov     rcx,0
               lea     rdx,[text]
               lea     r8,[caption]
               mov     r9,1    ; MB_OKCANCEL
               call    MessageBoxA
               sub     eax,1   ; incidentally suits as return value
                               ; for exception handler
               add     rsp,40
               ret
       global  main
       main:
               xor     rax,rax
               mov     rax,QWORD[rax]  ; cause exception
               ret
       main_end:
       text:   db      'OK to rethrow, CANCEL to generate core dump',0
       caption:db      'SEGV',0
      
       section .pdata  rdata align=4
               dd      main wrt ..imagebase
               dd      main_end wrt ..imagebase
               dd      xmain wrt ..imagebase
       section .xdata  rdata align=8
       xmain:  db      9,0,0,0
               dd      handler wrt ..imagebase
       section .drectve info
               db      '/defaultlib:user32.lib /defaultlib:msvcrt.lib '

       What you see in `.pdata' section is element of the "table comprising
       start and end addresses of function" along with reference to
       associated `UNWIND_INFO' structure. And what you see in `.xdata'
       section is `UNWIND_INFO' structure describing function with no
       frame, but with designated exception handler. References are
       _required_ to be image-relative (which is the real reason for
       implementing `wrt ..imagebase' operator). It should be noted that
       `rdata align=n', as well as `wrt ..imagebase', are optional in these
       two segments' contexts, i.e. can be omitted. Latter means that _all_
       32-bit references, not only above listed required ones, placed into
       these two segments turn out image-relative. Why is it important to
       understand? Developer is allowed to append handler-specific data to
       `UNWIND_INFO' structure, and if [s]he adds a 32-bit reference, then
       [s]he will have to remember to adjust its value to obtain the real
       pointer.

       As already mentioned, in Win64 terms leaf function is one that does
       not call any other function _nor_ modifies any non-volatile
       register, including stack pointer. But it's not uncommon that
       assembler programmer plans to utilize every single register and
       sometimes even have variable stack frame. Is there anything one can
       do with bare building blocks? I.e. besides manually composing fully-
       fledged `UNWIND_INFO' structure, which would surely be considered
       error-prone? Yes, there is. Recall that exception handler is called
       first, before stack layout is analyzed. As it turned out, it's
       perfectly possible to manipulate current callee's context in custom
       handler in manner that permits further stack unwinding. General idea
       is that handler would not actually "handle" the exception, but
       instead restore callee's context, as it was at its entry point and
       thus mimic leaf function. In other words, handler would simply
       undertake part of unwinding procedure. Consider following example:

       function:
               mov     rax,rsp         ; copy rsp to volatile register
               push    r15             ; save non-volatile registers
               push    rbx
               push    rbp
               mov     r11,rsp         ; prepare variable stack frame
               sub     r11,rcx
               and     r11,-64
               mov     QWORD[r11],rax  ; check for exceptions
               mov     rsp,r11         ; allocate stack frame
               mov     QWORD[rsp],rax  ; save original rsp value
       magic_point:
               ...
               mov     r11,QWORD[rsp]  ; pull original rsp value
               mov     rbp,QWORD[r11-24]
               mov     rbx,QWORD[r11-16]
               mov     r15,QWORD[r11-8]
               mov     rsp,r11         ; destroy frame
               ret

       The keyword is that up to `magic_point' original `rsp' value remains
       in chosen volatile register and no non-volatile register, except for
       `rsp', is modified. While past `magic_point' `rsp' remains constant
       till the very end of the `function'. In this case custom language-
       specific exception handler would look like this:

       EXCEPTION_DISPOSITION handler (EXCEPTION_RECORD *rec,ULONG64 frame,
               CONTEXT *context,DISPATCHER_CONTEXT *disp)
       {   ULONG64 *rsp;
           if (context->Rip<(ULONG64)magic_point)
               rsp = (ULONG64 *)context->Rax;
           else
           {   rsp = ((ULONG64 **)context->Rsp)[0];
               context->Rbp = rsp[-3];
               context->Rbx = rsp[-2];
               context->R15 = rsp[-1];
           }
           context->Rsp = (ULONG64)rsp;
      
           memcpy (disp->ContextRecord,context,sizeof(CONTEXT));
           RtlVirtualUnwind(UNW_FLAG_NHANDLER,disp->ImageBase,
               dips->ControlPc,disp->FunctionEntry,disp->ContextRecord,
               &disp->HandlerData,&disp->EstablisherFrame,NULL);
           return ExceptionContinueSearch;
       }

       As custom handler mimics leaf function, corresponding `UNWIND_INFO'
       structure does not have to contain any information about stack frame
       and its layout.

   7.7 `coff': Common Object File Format

       The `coff' output type produces `COFF' object files suitable for
       linking with the DJGPP linker.

       `coff' provides a default output file-name extension of `.o'.

       The `coff' format supports the same extensions to the `SECTION'
       directive as `win32' does, except that the `align' qualifier and the
       `info' section type are not supported.

   7.8 `macho32' and `macho64': Mach Object File Format

       The `macho32' and `macho64' output formts produces `Mach-O' object
       files suitable for linking with the MacOS X linker. `macho' is a
       synonym for `macho32'.

       `macho' provides a default output file-name extension of `.o'.

   7.9 `elf32', `elf64', `elfx32': Executable and Linkable Format Object Files

       The `elf32', `elf64' and `elfx32' output formats generate
       `ELF32 and ELF64' (Executable and Linkable Format) object files, as
       used by Linux as well as Unix System V, including Solaris x86,
       UnixWare and SCO Unix. `elf' provides a default output file-name
       extension of `.o'. `elf' is a synonym for `elf32'.

       The `elfx32' format is used for the x32 ABI, which is a 32-bit ABI
       with the CPU in 64-bit mode.

7.9.1 ELF specific directive `osabi'

       The ELF header specifies the application binary interface for the
       target operating system (OSABI). This field can be set by using the
       `osabi' directive with the numeric value (0-255) of the target
       system. If this directive is not used, the default value will be
       "UNIX System V ABI" (0) which will work on most systems which
       support ELF.

7.9.2 `elf' Extensions to the `SECTION' Directive

       Like the `obj' format, `elf' allows you to specify additional
       information on the `SECTION' directive line, to control the type and
       properties of sections you declare. Section types and properties are
       generated automatically by NASM for the standard section names, but
       may still be overridden by these qualifiers.

       The available qualifiers are:

       (*) `alloc' defines the section to be one which is loaded into
           memory when the program is run. `noalloc' defines it to be one
           which is not, such as an informational or comment section.

       (*) `exec' defines the section to be one which should have execute
           permission when the program is run. `noexec' defines it as one
           which should not.

       (*) `write' defines the section to be one which should be writable
           when the program is run. `nowrite' defines it as one which
           should not.

       (*) `progbits' defines the section to be one with explicit contents
           stored in the object file: an ordinary code or data section, for
           example, `nobits' defines the section to be one with no explicit
           contents given, such as a BSS section.

       (*) `align=', used with a trailing number as in `obj', gives the
           alignment requirements of the section.

       (*) `tls' defines the section to be one which contains thread local
           variables.

       The defaults assumed by NASM if you do not specify the above
       qualifiers are:


       section .text    progbits  alloc   exec    nowrite  align=16
       section .rodata  progbits  alloc   noexec  nowrite  align=4
       section .lrodata progbits  alloc   noexec  nowrite  align=4
       section .data    progbits  alloc   noexec  write    align=4
       section .ldata   progbits  alloc   noexec  write    align=4
       section .bss     nobits    alloc   noexec  write    align=4
       section .lbss    nobits    alloc   noexec  write    align=4
       section .tdata   progbits  alloc   noexec  write    align=4    tls
       section .tbss    nobits    alloc   noexec  write    align=4    tls
       section .comment progbits  noalloc noexec  nowrite  align=1
       section other    progbits  alloc   noexec  nowrite  align=1

       (Any section name other than those in the above table is treated by
       default like `other' in the above table. Please note that section
       names are case sensitive.)

7.9.3 Position-Independent Code: `elf' Special Symbols and `WRT'

       The `ELF' specification contains enough features to allow position-
       independent code (PIC) to be written, which makes ELF shared
       libraries very flexible. However, it also means NASM has to be able
       to generate a variety of ELF specific relocation types in ELF object
       files, if it is to be an assembler which can write PIC.

       Since `ELF' does not support segment-base references, the `WRT'
       operator is not used for its normal purpose; therefore NASM's `elf'
       output format makes use of `WRT' for a different purpose, namely the
       PIC-specific relocation types.

       `elf' defines five special symbols which you can use as the right-
       hand side of the `WRT' operator to obtain PIC relocation types. They
       are `..gotpc', `..gotoff', `..got', `..plt' and `..sym'. Their
       functions are summarized here:

       (*) Referring to the symbol marking the global offset table base
           using `wrt ..gotpc' will end up giving the distance from the
           beginning of the current section to the global offset table.
           (`_GLOBAL_OFFSET_TABLE_' is the standard symbol name used to
           refer to the GOT.) So you would then need to add `$$' to the
           result to get the real address of the GOT.

       (*) Referring to a location in one of your own sections using
           `wrt ..gotoff' will give the distance from the beginning of the
           GOT to the specified location, so that adding on the address of
           the GOT would give the real address of the location you wanted.

       (*) Referring to an external or global symbol using `wrt ..got'
           causes the linker to build an entry _in_ the GOT containing the
           address of the symbol, and the reference gives the distance from
           the beginning of the GOT to the entry; so you can add on the
           address of the GOT, load from the resulting address, and end up
           with the address of the symbol.

       (*) Referring to a procedure name using `wrt ..plt' causes the
           linker to build a procedure linkage table entry for the symbol,
           and the reference gives the address of the PLT entry. You can
           only use this in contexts which would generate a PC-relative
           relocation normally (i.e. as the destination for `CALL' or
           `JMP'), since ELF contains no relocation type to refer to PLT
           entries absolutely.

       (*) Referring to a symbol name using `wrt ..sym' causes NASM to
           write an ordinary relocation, but instead of making the
           relocation relative to the start of the section and then adding
           on the offset to the symbol, it will write a relocation record
           aimed directly at the symbol in question. The distinction is a
           necessary one due to a peculiarity of the dynamic linker.

       A fuller explanation of how to use these relocation types to write
       shared libraries entirely in NASM is given in section 9.2.

7.9.4 Thread Local Storage: `elf' Special Symbols and `WRT'

       (*) In ELF32 mode, referring to an external or global symbol using
           `wrt ..tlsie'  causes the linker to build an entry _in_ the GOT
           containing the offset of the symbol within the TLS block, so you
           can access the value of the symbol with code such as:

              mov  eax,[tid wrt ..tlsie]
              mov  [gs:eax],ebx

       (*) In ELF64 or ELFx32 mode, referring to an external or global
           symbol using `wrt ..gottpoff'  causes the linker to build an
           entry _in_ the GOT containing the offset of the symbol within
           the TLS block, so you can access the value of the symbol with
           code such as:

              mov   rax,[rel tid wrt ..gottpoff]
              mov   rcx,[fs:rax]

7.9.5 `elf' Extensions to the `GLOBAL' Directive

       `ELF' object files can contain more information about a global
       symbol than just its address: they can contain the size of the
       symbol and its type as well. These are not merely debugger
       conveniences, but are actually necessary when the program being
       written is a shared library. NASM therefore supports some extensions
       to the `GLOBAL' directive, allowing you to specify these features.

       You can specify whether a global variable is a function or a data
       object by suffixing the name with a colon and the word `function' or
       `data'. (`object' is a synonym for `data'.) For example:

       global   hashlookup:function, hashtable:data

       exports the global symbol `hashlookup' as a function and `hashtable'
       as a data object.

       Optionally, you can control the ELF visibility of the symbol. Just
       add one of the visibility keywords: `default', `internal', `hidden',
       or `protected'. The default is `default' of course. For example, to
       make `hashlookup' hidden:

       global   hashlookup:function hidden

       You can also specify the size of the data associated with the
       symbol, as a numeric expression (which may involve labels, and even
       forward references) after the type specifier. Like this:

       global  hashtable:data (hashtable.end - hashtable)
      
       hashtable:
               db this,that,theother  ; some data here
       .end:

       This makes NASM automatically calculate the length of the table and
       place that information into the `ELF' symbol table.

       Declaring the type and size of global symbols is necessary when
       writing shared library code. For more information, see section
       9.2.4.

7.9.6 `elf' Extensions to the `COMMON' Directive

       `ELF' also allows you to specify alignment requirements on common
       variables. This is done by putting a number (which must be a power
       of two) after the name and size of the common variable, separated
       (as usual) by a colon. For example, an array of doublewords would
       benefit from 4-byte alignment:

       common  dwordarray 128:4

       This declares the total size of the array to be 128 bytes, and
       requires that it be aligned on a 4-byte boundary.

7.9.7 16-bit code and ELF

       The `ELF32' specification doesn't provide relocations for 8- and 16-
       bit values, but the GNU `ld' linker adds these as an extension. NASM
       can generate GNU-compatible relocations, to allow 16-bit code to be
       linked as ELF using GNU `ld'. If NASM is used with the
       `-w+gnu-elf-extensions' option, a warning is issued when one of
       these relocations is generated.

7.9.8 Debug formats and ELF

       ELF provides debug information in `STABS' and `DWARF' formats. Line
       number information is generated for all executable sections, but
       please note that only the ".text" section is executable by default.

  7.10 `aout': Linux `a.out' Object Files

       The `aout' format generates `a.out' object files, in the form used
       by early Linux systems (current Linux systems use ELF, see section
       7.9.) These differ from other `a.out' object files in that the magic
       number in the first four bytes of the file is different; also, some
       implementations of `a.out', for example NetBSD's, support position-
       independent code, which Linux's implementation does not.

       `a.out' provides a default output file-name extension of `.o'.

       `a.out' is a very simple object format. It supports no special
       directives, no special symbols, no use of `SEG' or `WRT', and no
       extensions to any standard directives. It supports only the three
       standard section names `.text', `.data' and `.bss'.

  7.11 `aoutb': NetBSD/FreeBSD/OpenBSD `a.out' Object Files

       The `aoutb' format generates `a.out' object files, in the form used
       by the various free `BSD Unix' clones, `NetBSD', `FreeBSD' and
       `OpenBSD'. For simple object files, this object format is exactly
       the same as `aout' except for the magic number in the first four
       bytes of the file. However, the `aoutb' format supports
       position-independent code in the same way as the `elf' format, so
       you can use it to write `BSD' shared libraries.

       `aoutb' provides a default output file-name extension of `.o'.

       `aoutb' supports no special directives, no special symbols, and only
       the three standard section names `.text', `.data' and `.bss'.
       However, it also supports the same use of `WRT' as `elf' does, to
       provide position-independent code relocation types. See section
       7.9.3 for full documentation of this feature.

       `aoutb' also supports the same extensions to the `GLOBAL' directive
       as `elf' does: see section 7.9.5 for documentation of this.

  7.12 `as86': Minix/Linux `as86' Object Files

       The Minix/Linux 16-bit assembler `as86' has its own non-standard
       object file format. Although its companion linker `ld86' produces
       something close to ordinary `a.out' binaries as output, the object
       file format used to communicate between `as86' and `ld86' is not
       itself `a.out'.

       NASM supports this format, just in case it is useful, as `as86'.
       `as86' provides a default output file-name extension of `.o'.

       `as86' is a very simple object format (from the NASM user's point of
       view). It supports no special directives, no use of `SEG' or `WRT',
       and no extensions to any standard directives. It supports only the
       three standard section names `.text', `.data' and `.bss'. The only
       special symbol supported is `..start'.

  7.13 `rdf': Relocatable Dynamic Object File Format

       The `rdf' output format produces `RDOFF' object files. `RDOFF'
       (Relocatable Dynamic Object File Format) is a home-grown object-file
       format, designed alongside NASM itself and reflecting in its file
       format the internal structure of the assembler.

       `RDOFF' is not used by any well-known operating systems. Those
       writing their own systems, however, may well wish to use `RDOFF' as
       their object format, on the grounds that it is designed primarily
       for simplicity and contains very little file-header bureaucracy.

       The Unix NASM archive, and the DOS archive which includes sources,
       both contain an `rdoff' subdirectory holding a set of RDOFF
       utilities: an RDF linker, an `RDF' static-library manager, an RDF
       file dump utility, and a program which will load and execute an RDF
       executable under Linux.

       `rdf' supports only the standard section names `.text', `.data' and
       `.bss'.

7.13.1 Requiring a Library: The `LIBRARY' Directive

       `RDOFF' contains a mechanism for an object file to demand a given
       library to be linked to the module, either at load time or run time.
       This is done by the `LIBRARY' directive, which takes one argument
       which is the name of the module:

           library  mylib.rdl

7.13.2 Specifying a Module Name: The `MODULE' Directive

       Special `RDOFF' header record is used to store the name of the
       module. It can be used, for example, by run-time loader to perform
       dynamic linking. `MODULE' directive takes one argument which is the
       name of current module:

           module  mymodname

       Note that when you statically link modules and tell linker to strip
       the symbols from output file, all module names will be stripped too.
       To avoid it, you should start module names with `$', like:

           module  $kernel.core

7.13.3 `rdf' Extensions to the `GLOBAL' Directive

       `RDOFF' global symbols can contain additional information needed by
       the static linker. You can mark a global symbol as exported, thus
       telling the linker do not strip it from target executable or library
       file. Like in `ELF', you can also specify whether an exported symbol
       is a procedure (function) or data object.

       Suffixing the name with a colon and the word `export' you make the
       symbol exported:

           global  sys_open:export

       To specify that exported symbol is a procedure (function), you add
       the word `proc' or `function' after declaration:

           global  sys_open:export proc

       Similarly, to specify exported data object, add the word `data' or
       `object' to the directive:

           global  kernel_ticks:export data

7.13.4 `rdf' Extensions to the `EXTERN' Directive

       By default the `EXTERN' directive in `RDOFF' declares a "pure
       external" symbol (i.e. the static linker will complain if such a
       symbol is not resolved). To declare an "imported" symbol, which must
       be resolved later during a dynamic linking phase, `RDOFF' offers an
       additional `import' modifier. As in `GLOBAL', you can also specify
       whether an imported symbol is a procedure (function) or data object.
       For example:

           library $libc
           extern  _open:import
           extern  _printf:import proc
           extern  _errno:import data

       Here the directive `LIBRARY' is also included, which gives the
       dynamic linker a hint as to where to find requested symbols.

  7.14 `dbg': Debugging Format

       The `dbg' output format is not built into NASM in the default
       configuration. If you are building your own NASM executable from the
       sources, you can define `OF_DBG' in `output/outform.h' or on the
       compiler command line, and obtain the `dbg' output format.

       The `dbg' format does not output an object file as such; instead, it
       outputs a text file which contains a complete list of all the
       transactions between the main body of NASM and the output-format
       back end module. It is primarily intended to aid people who want to
       write their own output drivers, so that they can get a clearer idea
       of the various requests the main program makes of the output driver,
       and in what order they happen.

       For simple files, one can easily use the `dbg' format like this:

       nasm -f dbg filename.asm

       which will generate a diagnostic file called `filename.dbg'.
       However, this will not work well on files which were designed for a
       different object format, because each object format defines its own
       macros (usually user-level forms of directives), and those macros
       will not be defined in the `dbg' format. Therefore it can be useful
       to run NASM twice, in order to do the preprocessing with the native
       object format selected:

       nasm -e -f rdf -o rdfprog.i rdfprog.asm
       nasm -a -f dbg rdfprog.i

       This preprocesses `rdfprog.asm' into `rdfprog.i', keeping the `rdf'
       object format selected in order to make sure RDF special directives
       are converted into primitive form correctly. Then the preprocessed
       source is fed through the `dbg' format to generate the final
       diagnostic output.

       This workaround will still typically not work for programs intended
       for `obj' format, because the `obj' `SEGMENT' and `GROUP' directives
       have side effects of defining the segment and group names as
       symbols; `dbg' will not do this, so the program will not assemble.
       You will have to work around that by defining the symbols yourself
       (using `EXTERN', for example) if you really need to get a `dbg'
       trace of an `obj'-specific source file.

       `dbg' accepts any section name and any directives at all, and logs
       them all to its output file.

Chapter 8: Writing 16-bit Code (DOS, Windows 3/3.1)
---------------------------------------------------

       This chapter attempts to cover some of the common issues encountered
       when writing 16-bit code to run under `MS-DOS' or `Windows 3.x'. It
       covers how to link programs to produce `.EXE' or `.COM' files, how
       to write `.SYS' device drivers, and how to interface assembly
       language code with 16-bit C compilers and with Borland Pascal.

   8.1 Producing `.EXE' Files

       Any large program written under DOS needs to be built as a `.EXE'
       file: only `.EXE' files have the necessary internal structure
       required to span more than one 64K segment. Windows programs, also,
       have to be built as `.EXE' files, since Windows does not support the
       `.COM' format.

       In general, you generate `.EXE' files by using the `obj' output
       format to produce one or more `.OBJ' files, and then linking them
       together using a linker. However, NASM also supports the direct
       generation of simple DOS `.EXE' files using the `bin' output format
       (by using `DB' and `DW' to construct the `.EXE' file header), and a
       macro package is supplied to do this. Thanks to Yann Guidon for
       contributing the code for this.

       NASM may also support `.EXE' natively as another output format in
       future releases.

8.1.1 Using the `obj' Format To Generate `.EXE' Files

       This section describes the usual method of generating `.EXE' files
       by linking `.OBJ' files together.

       Most 16-bit programming language packages come with a suitable
       linker; if you have none of these, there is a free linker called
       VAL, available in `LZH' archive format from `x2ftp.oulu.fi'. An LZH
       archiver can be found at `ftp.simtel.net'. There is another `free'
       linker (though this one doesn't come with sources) called FREELINK,
       available from `www.pcorner.com'. A third, `djlink', written by DJ
       Delorie, is available at `www.delorie.com'. A fourth linker,
       `ALINK', written by Anthony A.J. Williams, is available at
       `alink.sourceforge.net'.

       When linking several `.OBJ' files into a `.EXE' file, you should
       ensure that exactly one of them has a start point defined (using the
       `..start' special symbol defined by the `obj' format: see section
       7.4.6). If no module defines a start point, the linker will not know
       what value to give the entry-point field in the output file header;
       if more than one defines a start point, the linker will not know
       _which_ value to use.

       An example of a NASM source file which can be assembled to a `.OBJ'
       file and linked on its own to a `.EXE' is given here. It
       demonstrates the basic principles of defining a stack, initialising
       the segment registers, and declaring a start point. This file is
       also provided in the `test' subdirectory of the NASM archives, under
       the name `objexe.asm'.

       segment code
      
       ..start:
               mov     ax,data
               mov     ds,ax
               mov     ax,stack
               mov     ss,ax
               mov     sp,stacktop

       This initial piece of code sets up `DS' to point to the data
       segment, and initializes `SS' and `SP' to point to the top of the
       provided stack. Notice that interrupts are implicitly disabled for
       one instruction after a move into `SS', precisely for this
       situation, so that there's no chance of an interrupt occurring
       between the loads of `SS' and `SP' and not having a stack to execute
       on.

       Note also that the special symbol `..start' is defined at the
       beginning of this code, which means that will be the entry point
       into the resulting executable file.

               mov     dx,hello
               mov     ah,9
               int     0x21

       The above is the main program: load `DS:DX' with a pointer to the
       greeting message (`hello' is implicitly relative to the segment
       `data', which was loaded into `DS' in the setup code, so the full
       pointer is valid), and call the DOS print-string function.

               mov     ax,0x4c00
               int     0x21

       This terminates the program using another DOS system call.

       segment data
      
       hello:  db      'hello, world', 13, 10, '$'

       The data segment contains the string we want to display.

       segment stack stack
               resb 64
       stacktop:

       The above code declares a stack segment containing 64 bytes of
       uninitialized stack space, and points `stacktop' at the top of it.
       The directive `segment stack stack' defines a segment _called_
       `stack', and also of _type_ `STACK'. The latter is not necessary to
       the correct running of the program, but linkers are likely to issue
       warnings or errors if your program has no segment of type `STACK'.

       The above file, when assembled into a `.OBJ' file, will link on its
       own to a valid `.EXE' file, which when run will print `hello, world'
       and then exit.

8.1.2 Using the `bin' Format To Generate `.EXE' Files

       The `.EXE' file format is simple enough that it's possible to build
       a `.EXE' file by writing a pure-binary program and sticking a 32-
       byte header on the front. This header is simple enough that it can
       be generated using `DB' and `DW' commands by NASM itself, so that
       you can use the `bin' output format to directly generate `.EXE'
       files.

       Included in the NASM archives, in the `misc' subdirectory, is a file
       `exebin.mac' of macros. It defines three macros: `EXE_begin',
       `EXE_stack' and `EXE_end'.

       To produce a `.EXE' file using this method, you should start by
       using `%include' to load the `exebin.mac' macro package into your
       source file. You should then issue the `EXE_begin' macro call (which
       takes no arguments) to generate the file header data. Then write
       code as normal for the `bin' format - you can use all three standard
       sections `.text', `.data' and `.bss'. At the end of the file you
       should call the `EXE_end' macro (again, no arguments), which defines
       some symbols to mark section sizes, and these symbols are referred
       to in the header code generated by `EXE_begin'.

       In this model, the code you end up writing starts at `0x100', just
       like a `.COM' file - in fact, if you strip off the 32-byte header
       from the resulting `.EXE' file, you will have a valid `.COM'
       program. All the segment bases are the same, so you are limited to a
       64K program, again just like a `.COM' file. Note that an `ORG'
       directive is issued by the `EXE_begin' macro, so you should not
       explicitly issue one of your own.

       You can't directly refer to your segment base value, unfortunately,
       since this would require a relocation in the header, and things
       would get a lot more complicated. So you should get your segment
       base by copying it out of `CS' instead.

       On entry to your `.EXE' file, `SS:SP' are already set up to point to
       the top of a 2Kb stack. You can adjust the default stack size of 2Kb
       by calling the `EXE_stack' macro. For example, to change the stack
       size of your program to 64 bytes, you would call `EXE_stack 64'.

       A sample program which generates a `.EXE' file in this way is given
       in the `test' subdirectory of the NASM archive, as `binexe.asm'.

   8.2 Producing `.COM' Files

       While large DOS programs must be written as `.EXE' files, small ones
       are often better written as `.COM' files. `.COM' files are pure
       binary, and therefore most easily produced using the `bin' output
       format.

8.2.1 Using the `bin' Format To Generate `.COM' Files

       `.COM' files expect to be loaded at offset `100h' into their segment
       (though the segment may change). Execution then begins at `100h',
       i.e. right at the start of the program. So to write a `.COM'
       program, you would create a source file looking like

               org 100h
      
       section .text
      
       start:
               ; put your code here
      
       section .data
      
               ; put data items here
      
       section .bss
      
               ; put uninitialized data here

       The `bin' format puts the `.text' section first in the file, so you
       can declare data or BSS items before beginning to write code if you
       want to and the code will still end up at the front of the file
       where it belongs.

       The BSS (uninitialized data) section does not take up space in the
       `.COM' file itself: instead, addresses of BSS items are resolved to
       point at space beyond the end of the file, on the grounds that this
       will be free memory when the program is run. Therefore you should
       not rely on your BSS being initialized to all zeros when you run.

       To assemble the above program, you should use a command line like

       nasm myprog.asm -fbin -o myprog.com

       The `bin' format would produce a file called `myprog' if no explicit
       output file name were specified, so you have to override it and give
       the desired file name.

8.2.2 Using the `obj' Format To Generate `.COM' Files

       If you are writing a `.COM' program as more than one module, you may
       wish to assemble several `.OBJ' files and link them together into a
       `.COM' program. You can do this, provided you have a linker capable
       of outputting `.COM' files directly (TLINK does this), or
       alternatively a converter program such as `EXE2BIN' to transform the
       `.EXE' file output from the linker into a `.COM' file.

       If you do this, you need to take care of several things:

       (*) The first object file containing code should start its code
           segment with a line like `RESB 100h'. This is to ensure that the
           code begins at offset `100h' relative to the beginning of the
           code segment, so that the linker or converter program does not
           have to adjust address references within the file when
           generating the `.COM' file. Other assemblers use an `ORG'
           directive for this purpose, but `ORG' in NASM is a format-
           specific directive to the `bin' output format, and does not mean
           the same thing as it does in MASM-compatible assemblers.

       (*) You don't need to define a stack segment.

       (*) All your segments should be in the same group, so that every
           time your code or data references a symbol offset, all offsets
           are relative to the same segment base. This is because, when a
           `.COM' file is loaded, all the segment registers contain the
           same value.

   8.3 Producing `.SYS' Files

       MS-DOS device drivers - `.SYS' files - are pure binary files,
       similar to `.COM' files, except that they start at origin zero
       rather than `100h'. Therefore, if you are writing a device driver
       using the `bin' format, you do not need the `ORG' directive, since
       the default origin for `bin' is zero. Similarly, if you are using
       `obj', you do not need the `RESB 100h' at the start of your code
       segment.

       `.SYS' files start with a header structure, containing pointers to
       the various routines inside the driver which do the work. This
       structure should be defined at the start of the code segment, even
       though it is not actually code.

       For more information on the format of `.SYS' files, and the data
       which has to go in the header structure, a list of books is given in
       the Frequently Asked Questions list for the newsgroup
       `comp.os.msdos.programmer'.

   8.4 Interfacing to 16-bit C Programs

       This section covers the basics of writing assembly routines that
       call, or are called from, C programs. To do this, you would
       typically write an assembly module as a `.OBJ' file, and link it
       with your C modules to produce a mixed-language program.

8.4.1 External Symbol Names

       C compilers have the convention that the names of all global symbols
       (functions or data) they define are formed by prefixing an
       underscore to the name as it appears in the C program. So, for
       example, the function a C programmer thinks of as `printf' appears
       to an assembly language programmer as `_printf'. This means that in
       your assembly programs, you can define symbols without a leading
       underscore, and not have to worry about name clashes with C symbols.

       If you find the underscores inconvenient, you can define macros to
       replace the `GLOBAL' and `EXTERN' directives as follows:

       %macro  cglobal 1
      
         global  _%1
         %define %1 _%1
      
       %endmacro
      
       %macro  cextern 1
      
         extern  _%1
         %define %1 _%1
      
       %endmacro

       (These forms of the macros only take one argument at a time; a
       `%rep' construct could solve this.)

       If you then declare an external like this:

       cextern printf

       then the macro will expand it as

       extern  _printf
       %define printf _printf

       Thereafter, you can reference `printf' as if it was a symbol, and
       the preprocessor will put the leading underscore on where necessary.

       The `cglobal' macro works similarly. You must use `cglobal' before
       defining the symbol in question, but you would have had to do that
       anyway if you used `GLOBAL'.

       Also see section 2.1.27.

8.4.2 Memory Models

       NASM contains no mechanism to support the various C memory models
       directly; you have to keep track yourself of which one you are
       writing for. This means you have to keep track of the following
       things:

       (*) In models using a single code segment (tiny, small and compact),
           functions are near. This means that function pointers, when
           stored in data segments or pushed on the stack as function
           arguments, are 16 bits long and contain only an offset field
           (the `CS' register never changes its value, and always gives the
           segment part of the full function address), and that functions
           are called using ordinary near `CALL' instructions and return
           using `RETN' (which, in NASM, is synonymous with `RET' anyway).
           This means both that you should write your own routines to
           return with `RETN', and that you should call external C routines
           with near `CALL' instructions.

       (*) In models using more than one code segment (medium, large and
           huge), functions are far. This means that function pointers are
           32 bits long (consisting of a 16-bit offset followed by a 16-bit
           segment), and that functions are called using `CALL FAR' (or
           `CALL seg:offset') and return using `RETF'. Again, you should
           therefore write your own routines to return with `RETF' and use
           `CALL FAR' to call external routines.

       (*) In models using a single data segment (tiny, small and medium),
           data pointers are 16 bits long, containing only an offset field
           (the `DS' register doesn't change its value, and always gives
           the segment part of the full data item address).

       (*) In models using more than one data segment (compact, large and
           huge), data pointers are 32 bits long, consisting of a 16-bit
           offset followed by a 16-bit segment. You should still be careful
           not to modify `DS' in your routines without restoring it
           afterwards, but `ES' is free for you to use to access the
           contents of 32-bit data pointers you are passed.

       (*) The huge memory model allows single data items to exceed 64K in
           size. In all other memory models, you can access the whole of a
           data item just by doing arithmetic on the offset field of the
           pointer you are given, whether a segment field is present or
           not; in huge model, you have to be more careful of your pointer
           arithmetic.

       (*) In most memory models, there is a _default_ data segment, whose
           segment address is kept in `DS' throughout the program. This
           data segment is typically the same segment as the stack, kept in
           `SS', so that functions' local variables (which are stored on
           the stack) and global data items can both be accessed easily
           without changing `DS'. Particularly large data items are
           typically stored in other segments. However, some memory models
           (though not the standard ones, usually) allow the assumption
           that `SS' and `DS' hold the same value to be removed. Be careful
           about functions' local variables in this latter case.

       In models with a single code segment, the segment is called `_TEXT',
       so your code segment must also go by this name in order to be linked
       into the same place as the main code segment. In models with a
       single data segment, or with a default data segment, it is called
       `_DATA'.

8.4.3 Function Definitions and Function Calls

       The C calling convention in 16-bit programs is as follows. In the
       following description, the words _caller_ and _callee_ are used to
       denote the function doing the calling and the function which gets
       called.

       (*) The caller pushes the function's parameters on the stack, one
           after another, in reverse order (right to left, so that the
           first argument specified to the function is pushed last).

       (*) The caller then executes a `CALL' instruction to pass control to
           the callee. This `CALL' is either near or far depending on the
           memory model.

       (*) The callee receives control, and typically (although this is not
           actually necessary, in functions which do not need to access
           their parameters) starts by saving the value of `SP' in `BP' so
           as to be able to use `BP' as a base pointer to find its
           parameters on the stack. However, the caller was probably doing
           this too, so part of the calling convention states that `BP'
           must be preserved by any C function. Hence the callee, if it is
           going to set up `BP' as a _frame pointer_, must push the
           previous value first.

       (*) The callee may then access its parameters relative to `BP'. The
           word at `[BP]' holds the previous value of `BP' as it was
           pushed; the next word, at `[BP+2]', holds the offset part of the
           return address, pushed implicitly by `CALL'. In a small-model
           (near) function, the parameters start after that, at `[BP+4]';
           in a large-model (far) function, the segment part of the return
           address lives at `[BP+4]', and the parameters begin at `[BP+6]'.
           The leftmost parameter of the function, since it was pushed
           last, is accessible at this offset from `BP'; the others follow,
           at successively greater offsets. Thus, in a function such as
           `printf' which takes a variable number of parameters, the
           pushing of the parameters in reverse order means that the
           function knows where to find its first parameter, which tells it
           the number and type of the remaining ones.

       (*) The callee may also wish to decrease `SP' further, so as to
           allocate space on the stack for local variables, which will then
           be accessible at negative offsets from `BP'.

       (*) The callee, if it wishes to return a value to the caller, should
           leave the value in `AL', `AX' or `DX:AX' depending on the size
           of the value. Floating-point results are sometimes (depending on
           the compiler) returned in `ST0'.

       (*) Once the callee has finished processing, it restores `SP' from
           `BP' if it had allocated local stack space, then pops the
           previous value of `BP', and returns via `RETN' or `RETF'
           depending on memory model.

       (*) When the caller regains control from the callee, the function
           parameters are still on the stack, so it typically adds an
           immediate constant to `SP' to remove them (instead of executing
           a number of slow `POP' instructions). Thus, if a function is
           accidentally called with the wrong number of parameters due to a
           prototype mismatch, the stack will still be returned to a
           sensible state since the caller, which _knows_ how many
           parameters it pushed, does the removing.

       It is instructive to compare this calling convention with that for
       Pascal programs (described in section 8.5.1). Pascal has a simpler
       convention, since no functions have variable numbers of parameters.
       Therefore the callee knows how many parameters it should have been
       passed, and is able to deallocate them from the stack itself by
       passing an immediate argument to the `RET' or `RETF' instruction, so
       the caller does not have to do it. Also, the parameters are pushed
       in left-to-right order, not right-to-left, which means that a
       compiler can give better guarantees about sequence points without
       performance suffering.

       Thus, you would define a function in C style in the following way.
       The following example is for small model:

       global  _myfunc
      
       _myfunc:
               push    bp
               mov     bp,sp
               sub     sp,0x40         ; 64 bytes of local stack space
               mov     bx,[bp+4]       ; first parameter to function
      
               ; some more code
      
               mov     sp,bp           ; undo "sub sp,0x40" above
               pop     bp
               ret

       For a large-model function, you would replace `RET' by `RETF', and
       look for the first parameter at `[BP+6]' instead of `[BP+4]'. Of
       course, if one of the parameters is a pointer, then the offsets of
       _subsequent_ parameters will change depending on the memory model as
       well: far pointers take up four bytes on the stack when passed as a
       parameter, whereas near pointers take up two.

       At the other end of the process, to call a C function from your
       assembly code, you would do something like this:

       extern  _printf
      
             ; and then, further down...
      
             push    word [myint]        ; one of my integer variables
             push    word mystring       ; pointer into my data segment
             call    _printf
             add     sp,byte 4           ; `byte' saves space
      
             ; then those data items...
      
       segment _DATA
      
       myint         dw    1234
       mystring      db    'This number -> %d <- should be 1234',10,0

       This piece of code is the small-model assembly equivalent of the C
       code

           int myint = 1234;
           printf("This number -> %d <- should be 1234\n", myint);

       In large model, the function-call code might look more like this. In
       this example, it is assumed that `DS' already holds the segment base
       of the segment `_DATA'. If not, you would have to initialize it
       first.

             push    word [myint]
             push    word seg mystring   ; Now push the segment, and...
             push    word mystring       ; ... offset of "mystring"
             call    far _printf
             add    sp,byte 6

       The integer value still takes up one word on the stack, since large
       model does not affect the size of the `int' data type. The first
       argument (pushed last) to `printf', however, is a data pointer, and
       therefore has to contain a segment and offset part. The segment
       should be stored second in memory, and therefore must be pushed
       first. (Of course, `PUSH DS' would have been a shorter instruction
       than `PUSH WORD SEG mystring', if `DS' was set up as the above
       example assumed.) Then the actual call becomes a far call, since
       functions expect far calls in large model; and `SP' has to be
       increased by 6 rather than 4 afterwards to make up for the extra
       word of parameters.

8.4.4 Accessing Data Items

       To get at the contents of C variables, or to declare variables which
       C can access, you need only declare the names as `GLOBAL' or
       `EXTERN'. (Again, the names require leading underscores, as stated
       in section 8.4.1.) Thus, a C variable declared as `int i' can be
       accessed from assembler as

       extern _i
      
               mov ax,[_i]

       And to declare your own integer variable which C programs can access
       as `extern int j', you do this (making sure you are assembling in
       the `_DATA' segment, if necessary):

       global  _j
      
       _j      dw      0

       To access a C array, you need to know the size of the components of
       the array. For example, `int' variables are two bytes long, so if a
       C program declares an array as `int a[10]', you can access `a[3]' by
       coding `mov ax,[_a+6]'. (The byte offset 6 is obtained by
       multiplying the desired array index, 3, by the size of the array
       element, 2.) The sizes of the C base types in 16-bit compilers are:
       1 for `char', 2 for `short' and `int', 4 for `long' and `float', and
       8 for `double'.

       To access a C data structure, you need to know the offset from the
       base of the structure to the field you are interested in. You can
       either do this by converting the C structure definition into a NASM
       structure definition (using `STRUC'), or by calculating the one
       offset and using just that.

       To do either of these, you should read your C compiler's manual to
       find out how it organizes data structures. NASM gives no special
       alignment to structure members in its own `STRUC' macro, so you have
       to specify alignment yourself if the C compiler generates it.
       Typically, you might find that a structure like

       struct {
           char c;
           int i;
       } foo;

       might be four bytes long rather than three, since the `int' field
       would be aligned to a two-byte boundary. However, this sort of
       feature tends to be a configurable option in the C compiler, either
       using command-line options or `#pragma' lines, so you have to find
       out how your own compiler does it.

8.4.5 `c16.mac': Helper Macros for the 16-bit C Interface

       Included in the NASM archives, in the `misc' directory, is a file
       `c16.mac' of macros. It defines three macros: `proc', `arg' and
       `endproc'. These are intended to be used for C-style procedure
       definitions, and they automate a lot of the work involved in keeping
       track of the calling convention.

       (An alternative, TASM compatible form of `arg' is also now built
       into NASM's preprocessor. See section 4.8 for details.)

       An example of an assembly function using the macro set is given
       here:

       proc    _nearproc
      
       %$i     arg
       %$j     arg
               mov     ax,[bp + %$i]
               mov     bx,[bp + %$j]
               add     ax,[bx]
      
       endproc

       This defines `_nearproc' to be a procedure taking two arguments, the
       first (`i') an integer and the second (`j') a pointer to an integer.
       It returns `i + *j'.

       Note that the `arg' macro has an `EQU' as the first line of its
       expansion, and since the label before the macro call gets prepended
       to the first line of the expanded macro, the `EQU' works, defining
       `%$i' to be an offset from `BP'. A context-local variable is used,
       local to the context pushed by the `proc' macro and popped by the
       `endproc' macro, so that the same argument name can be used in later
       procedures. Of course, you don't _have_ to do that.

       The macro set produces code for near functions (tiny, small and
       compact-model code) by default. You can have it generate far
       functions (medium, large and huge-model code) by means of coding
       `%define FARCODE'. This changes the kind of return instruction
       generated by `endproc', and also changes the starting point for the
       argument offsets. The macro set contains no intrinsic dependency on
       whether data pointers are far or not.

       `arg' can take an optional parameter, giving the size of the
       argument. If no size is given, 2 is assumed, since it is likely that
       many function parameters will be of type `int'.

       The large-model equivalent of the above function would look like
       this:

       %define FARCODE
      
       proc    _farproc
      
       %$i     arg
       %$j     arg     4
               mov     ax,[bp + %$i]
               mov     bx,[bp + %$j]
               mov     es,[bp + %$j + 2]
               add     ax,[bx]
      
       endproc

       This makes use of the argument to the `arg' macro to define a
       parameter of size 4, because `j' is now a far pointer. When we load
       from `j', we must load a segment and an offset.

   8.5 Interfacing to Borland Pascal Programs

       Interfacing to Borland Pascal programs is similar in concept to
       interfacing to 16-bit C programs. The differences are:

       (*) The leading underscore required for interfacing to C programs is
           not required for Pascal.

       (*) The memory model is always large: functions are far, data
           pointers are far, and no data item can be more than 64K long.
           (Actually, some functions are near, but only those functions
           that are local to a Pascal unit and never called from outside
           it. All assembly functions that Pascal calls, and all Pascal
           functions that assembly routines are able to call, are far.)
           However, all static data declared in a Pascal program goes into
           the default data segment, which is the one whose segment address
           will be in `DS' when control is passed to your assembly code.
           The only things that do not live in the default data segment are
           local variables (they live in the stack segment) and dynamically
           allocated variables. All data _pointers_, however, are far.

       (*) The function calling convention is different - described below.

       (*) Some data types, such as strings, are stored differently.

       (*) There are restrictions on the segment names you are allowed to
           use - Borland Pascal will ignore code or data declared in a
           segment it doesn't like the name of. The restrictions are
           described below.

8.5.1 The Pascal Calling Convention

       The 16-bit Pascal calling convention is as follows. In the following
       description, the words _caller_ and _callee_ are used to denote the
       function doing the calling and the function which gets called.

       (*) The caller pushes the function's parameters on the stack, one
           after another, in normal order (left to right, so that the first
           argument specified to the function is pushed first).

       (*) The caller then executes a far `CALL' instruction to pass
           control to the callee.

       (*) The callee receives control, and typically (although this is not
           actually necessary, in functions which do not need to access
           their parameters) starts by saving the value of `SP' in `BP' so
           as to be able to use `BP' as a base pointer to find its
           parameters on the stack. However, the caller was probably doing
           this too, so part of the calling convention states that `BP'
           must be preserved by any function. Hence the callee, if it is
           going to set up `BP' as a frame pointer, must push the previous
           value first.

       (*) The callee may then access its parameters relative to `BP'. The
           word at `[BP]' holds the previous value of `BP' as it was
           pushed. The next word, at `[BP+2]', holds the offset part of the
           return address, and the next one at `[BP+4]' the segment part.
           The parameters begin at `[BP+6]'. The rightmost parameter of the
           function, since it was pushed last, is accessible at this offset
           from `BP'; the others follow, at successively greater offsets.

       (*) The callee may also wish to decrease `SP' further, so as to
           allocate space on the stack for local variables, which will then
           be accessible at negative offsets from `BP'.

       (*) The callee, if it wishes to return a value to the caller, should
           leave the value in `AL', `AX' or `DX:AX' depending on the size
           of the value. Floating-point results are returned in `ST0'.
           Results of type `Real' (Borland's own custom floating-point data
           type, not handled directly by the FPU) are returned in
           `DX:BX:AX'. To return a result of type `String', the caller
           pushes a pointer to a temporary string before pushing the
           parameters, and the callee places the returned string value at
           that location. The pointer is not a parameter, and should not be
           removed from the stack by the `RETF' instruction.

       (*) Once the callee has finished processing, it restores `SP' from
           `BP' if it had allocated local stack space, then pops the
           previous value of `BP', and returns via `RETF'. It uses the form
           of `RETF' with an immediate parameter, giving the number of
           bytes taken up by the parameters on the stack. This causes the
           parameters to be removed from the stack as a side effect of the
           return instruction.

       (*) When the caller regains control from the callee, the function
           parameters have already been removed from the stack, so it needs
           to do nothing further.

       Thus, you would define a function in Pascal style, taking two
       `Integer'-type parameters, in the following way:

       global  myfunc
      
       myfunc: push    bp
               mov     bp,sp
               sub     sp,0x40         ; 64 bytes of local stack space
               mov     bx,[bp+8]       ; first parameter to function
               mov     bx,[bp+6]       ; second parameter to function
      
               ; some more code
      
               mov     sp,bp           ; undo "sub sp,0x40" above
               pop     bp
               retf    4               ; total size of params is 4

       At the other end of the process, to call a Pascal function from your
       assembly code, you would do something like this:

       extern  SomeFunc
      
              ; and then, further down...
      
              push   word seg mystring   ; Now push the segment, and...
              push   word mystring       ; ... offset of "mystring"
              push   word [myint]        ; one of my variables
              call   far SomeFunc

       This is equivalent to the Pascal code

       procedure SomeFunc(String: PChar; Int: Integer);
           SomeFunc(@mystring, myint);

8.5.2 Borland Pascal Segment Name Restrictions

       Since Borland Pascal's internal unit file format is completely
       different from `OBJ', it only makes a very sketchy job of actually
       reading and understanding the various information contained in a
       real `OBJ' file when it links that in. Therefore an object file
       intended to be linked to a Pascal program must obey a number of
       restrictions:

       (*) Procedures and functions must be in a segment whose name is
           either `CODE', `CSEG', or something ending in `_TEXT'.

       (*) initialized data must be in a segment whose name is either
           `CONST' or something ending in `_DATA'.

       (*) Uninitialized data must be in a segment whose name is either
           `DATA', `DSEG', or something ending in `_BSS'.

       (*) Any other segments in the object file are completely ignored.
           `GROUP' directives and segment attributes are also ignored.

8.5.3 Using `c16.mac' With Pascal Programs

       The `c16.mac' macro package, described in section 8.4.5, can also be
       used to simplify writing functions to be called from Pascal
       programs, if you code `%define PASCAL'. This definition ensures that
       functions are far (it implies `FARCODE'), and also causes procedure
       return instructions to be generated with an operand.

       Defining `PASCAL' does not change the code which calculates the
       argument offsets; you must declare your function's arguments in
       reverse order. For example:

       %define PASCAL
      
       proc    _pascalproc
      
       %$j     arg 4
       %$i     arg
               mov     ax,[bp + %$i]
               mov     bx,[bp + %$j]
               mov     es,[bp + %$j + 2]
               add     ax,[bx]
      
       endproc

       This defines the same routine, conceptually, as the example in
       section 8.4.5: it defines a function taking two arguments, an
       integer and a pointer to an integer, which returns the sum of the
       integer and the contents of the pointer. The only difference between
       this code and the large-model C version is that `PASCAL' is defined
       instead of `FARCODE', and that the arguments are declared in reverse
       order.

Chapter 9: Writing 32-bit Code (Unix, Win32, DJGPP)
---------------------------------------------------

       This chapter attempts to cover some of the common issues involved
       when writing 32-bit code, to run under Win32 or Unix, or to be
       linked with C code generated by a Unix-style C compiler such as
       DJGPP. It covers how to write assembly code to interface with 32-bit
       C routines, and how to write position-independent code for shared
       libraries.

       Almost all 32-bit code, and in particular all code running under
       `Win32', `DJGPP' or any of the PC Unix variants, runs in _flat_
       memory model. This means that the segment registers and paging have
       already been set up to give you the same 32-bit 4Gb address space no
       matter what segment you work relative to, and that you should ignore
       all segment registers completely. When writing flat-model
       application code, you never need to use a segment override or modify
       any segment register, and the code-section addresses you pass to
       `CALL' and `JMP' live in the same address space as the data-section
       addresses you access your variables by and the stack-section
       addresses you access local variables and procedure parameters by.
       Every address is 32 bits long and contains only an offset part.

   9.1 Interfacing to 32-bit C Programs

       A lot of the discussion in section 8.4, about interfacing to 16-bit
       C programs, still applies when working in 32 bits. The absence of
       memory models or segmentation worries simplifies things a lot.

9.1.1 External Symbol Names

       Most 32-bit C compilers share the convention used by 16-bit
       compilers, that the names of all global symbols (functions or data)
       they define are formed by prefixing an underscore to the name as it
       appears in the C program. However, not all of them do: the `ELF'
       specification states that C symbols do _not_ have a leading
       underscore on their assembly-language names.

       The older Linux `a.out' C compiler, all `Win32' compilers, `DJGPP',
       and `NetBSD' and `FreeBSD', all use the leading underscore; for
       these compilers, the macros `cextern' and `cglobal', as given in
       section 8.4.1, will still work. For `ELF', though, the leading
       underscore should not be used.

       See also section 2.1.27.

9.1.2 Function Definitions and Function Calls

       The C calling convention in 32-bit programs is as follows. In the
       following description, the words _caller_ and _callee_ are used to
       denote the function doing the calling and the function which gets
       called.

       (*) The caller pushes the function's parameters on the stack, one
           after another, in reverse order (right to left, so that the
           first argument specified to the function is pushed last).

       (*) The caller then executes a near `CALL' instruction to pass
           control to the callee.

       (*) The callee receives control, and typically (although this is not
           actually necessary, in functions which do not need to access
           their parameters) starts by saving the value of `ESP' in `EBP'
           so as to be able to use `EBP' as a base pointer to find its
           parameters on the stack. However, the caller was probably doing
           this too, so part of the calling convention states that `EBP'
           must be preserved by any C function. Hence the callee, if it is
           going to set up `EBP' as a frame pointer, must push the previous
           value first.

       (*) The callee may then access its parameters relative to `EBP'. The
           doubleword at `[EBP]' holds the previous value of `EBP' as it
           was pushed; the next doubleword, at `[EBP+4]', holds the return
           address, pushed implicitly by `CALL'. The parameters start after
           that, at `[EBP+8]'. The leftmost parameter of the function,
           since it was pushed last, is accessible at this offset from
           `EBP'; the others follow, at successively greater offsets. Thus,
           in a function such as `printf' which takes a variable number of
           parameters, the pushing of the parameters in reverse order means
           that the function knows where to find its first parameter, which
           tells it the number and type of the remaining ones.

       (*) The callee may also wish to decrease `ESP' further, so as to
           allocate space on the stack for local variables, which will then
           be accessible at negative offsets from `EBP'.

       (*) The callee, if it wishes to return a value to the caller, should
           leave the value in `AL', `AX' or `EAX' depending on the size of
           the value. Floating-point results are typically returned in
           `ST0'.

       (*) Once the callee has finished processing, it restores `ESP' from
           `EBP' if it had allocated local stack space, then pops the
           previous value of `EBP', and returns via `RET' (equivalently,
           `RETN').

       (*) When the caller regains control from the callee, the function
           parameters are still on the stack, so it typically adds an
           immediate constant to `ESP' to remove them (instead of executing
           a number of slow `POP' instructions). Thus, if a function is
           accidentally called with the wrong number of parameters due to a
           prototype mismatch, the stack will still be returned to a
           sensible state since the caller, which _knows_ how many
           parameters it pushed, does the removing.

       There is an alternative calling convention used by Win32 programs
       for Windows API calls, and also for functions called _by_ the
       Windows API such as window procedures: they follow what Microsoft
       calls the `__stdcall' convention. This is slightly closer to the
       Pascal convention, in that the callee clears the stack by passing a
       parameter to the `RET' instruction. However, the parameters are
       still pushed in right-to-left order.

       Thus, you would define a function in C style in the following way:

       global  _myfunc
      
       _myfunc:
               push    ebp
               mov     ebp,esp
               sub     esp,0x40        ; 64 bytes of local stack space
               mov     ebx,[ebp+8]     ; first parameter to function
      
               ; some more code
      
               leave                   ; mov esp,ebp / pop ebp
               ret

       At the other end of the process, to call a C function from your
       assembly code, you would do something like this:

       extern  _printf
      
               ; and then, further down...
      
               push    dword [myint]   ; one of my integer variables
               push    dword mystring  ; pointer into my data segment
               call    _printf
               add     esp,byte 8      ; `byte' saves space
      
               ; then those data items...
      
       segment _DATA
      
       myint       dd   1234
       mystring    db   'This number -> %d <- should be 1234',10,0

       This piece of code is the assembly equivalent of the C code

           int myint = 1234;
           printf("This number -> %d <- should be 1234\n", myint);

9.1.3 Accessing Data Items

       To get at the contents of C variables, or to declare variables which
       C can access, you need only declare the names as `GLOBAL' or
       `EXTERN'. (Again, the names require leading underscores, as stated
       in section 9.1.1.) Thus, a C variable declared as `int i' can be
       accessed from assembler as

                 extern _i
                 mov eax,[_i]

       And to declare your own integer variable which C programs can access
       as `extern int j', you do this (making sure you are assembling in
       the `_DATA' segment, if necessary):

                 global _j
       _j        dd 0

       To access a C array, you need to know the size of the components of
       the array. For example, `int' variables are four bytes long, so if a
       C program declares an array as `int a[10]', you can access `a[3]' by
       coding `mov ax,[_a+12]'. (The byte offset 12 is obtained by
       multiplying the desired array index, 3, by the size of the array
       element, 4.) The sizes of the C base types in 32-bit compilers are:
       1 for `char', 2 for `short', 4 for `int', `long' and `float', and 8
       for `double'. Pointers, being 32-bit addresses, are also 4 bytes
       long.

       To access a C data structure, you need to know the offset from the
       base of the structure to the field you are interested in. You can
       either do this by converting the C structure definition into a NASM
       structure definition (using `STRUC'), or by calculating the one
       offset and using just that.

       To do either of these, you should read your C compiler's manual to
       find out how it organizes data structures. NASM gives no special
       alignment to structure members in its own `STRUC' macro, so you have
       to specify alignment yourself if the C compiler generates it.
       Typically, you might find that a structure like

       struct {
           char c;
           int i;
       } foo;

       might be eight bytes long rather than five, since the `int' field
       would be aligned to a four-byte boundary. However, this sort of
       feature is sometimes a configurable option in the C compiler, either
       using command-line options or `#pragma' lines, so you have to find
       out how your own compiler does it.

9.1.4 `c32.mac': Helper Macros for the 32-bit C Interface

       Included in the NASM archives, in the `misc' directory, is a file
       `c32.mac' of macros. It defines three macros: `proc', `arg' and
       `endproc'. These are intended to be used for C-style procedure
       definitions, and they automate a lot of the work involved in keeping
       track of the calling convention.

       An example of an assembly function using the macro set is given
       here:

       proc    _proc32
      
       %$i     arg
       %$j     arg
               mov     eax,[ebp + %$i]
               mov     ebx,[ebp + %$j]
               add     eax,[ebx]
      
       endproc

       This defines `_proc32' to be a procedure taking two arguments, the
       first (`i') an integer and the second (`j') a pointer to an integer.
       It returns `i + *j'.

       Note that the `arg' macro has an `EQU' as the first line of its
       expansion, and since the label before the macro call gets prepended
       to the first line of the expanded macro, the `EQU' works, defining
       `%$i' to be an offset from `BP'. A context-local variable is used,
       local to the context pushed by the `proc' macro and popped by the
       `endproc' macro, so that the same argument name can be used in later
       procedures. Of course, you don't _have_ to do that.

       `arg' can take an optional parameter, giving the size of the
       argument. If no size is given, 4 is assumed, since it is likely that
       many function parameters will be of type `int' or pointers.

   9.2 Writing NetBSD/FreeBSD/OpenBSD and Linux/ELF Shared Libraries

       `ELF' replaced the older `a.out' object file format under Linux
       because it contains support for position-independent code (PIC),
       which makes writing shared libraries much easier. NASM supports the
       `ELF' position-independent code features, so you can write Linux
       `ELF' shared libraries in NASM.

       NetBSD, and its close cousins FreeBSD and OpenBSD, take a different
       approach by hacking PIC support into the `a.out' format. NASM
       supports this as the `aoutb' output format, so you can write BSD
       shared libraries in NASM too.

       The operating system loads a PIC shared library by memory-mapping
       the library file at an arbitrarily chosen point in the address space
       of the running process. The contents of the library's code section
       must therefore not depend on where it is loaded in memory.

       Therefore, you cannot get at your variables by writing code like
       this:

               mov     eax,[myvar]             ; WRONG

       Instead, the linker provides an area of memory called the _global
       offset table_, or GOT; the GOT is situated at a constant distance
       from your library's code, so if you can find out where your library
       is loaded (which is typically done using a `CALL' and `POP'
       combination), you can obtain the address of the GOT, and you can
       then load the addresses of your variables out of linker-generated
       entries in the GOT.

       The _data_ section of a PIC shared library does not have these
       restrictions: since the data section is writable, it has to be
       copied into memory anyway rather than just paged in from the library
       file, so as long as it's being copied it can be relocated too. So
       you can put ordinary types of relocation in the data section without
       too much worry (but see section 9.2.4 for a caveat).

9.2.1 Obtaining the Address of the GOT

       Each code module in your shared library should define the GOT as an
       external symbol:

       extern  _GLOBAL_OFFSET_TABLE_   ; in ELF
       extern  __GLOBAL_OFFSET_TABLE_  ; in BSD a.out

       At the beginning of any function in your shared library which plans
       to access your data or BSS sections, you must first calculate the
       address of the GOT. This is typically done by writing the function
       in this form:

       func:   push    ebp
               mov     ebp,esp
               push    ebx
               call    .get_GOT
       .get_GOT:
               pop     ebx
               add     ebx,_GLOBAL_OFFSET_TABLE_+$$-.get_GOT wrt ..gotpc
      
               ; the function body comes here
      
               mov     ebx,[ebp-4]
               mov     esp,ebp
               pop     ebp
               ret

       (For BSD, again, the symbol `_GLOBAL_OFFSET_TABLE' requires a second
       leading underscore.)

       The first two lines of this function are simply the standard C
       prologue to set up a stack frame, and the last three lines are
       standard C function epilogue. The third line, and the fourth to last
       line, save and restore the `EBX' register, because PIC shared
       libraries use this register to store the address of the GOT.

       The interesting bit is the `CALL' instruction and the following two
       lines. The `CALL' and `POP' combination obtains the address of the
       label `.get_GOT', without having to know in advance where the
       program was loaded (since the `CALL' instruction is encoded relative
       to the current position). The `ADD' instruction makes use of one of
       the special PIC relocation types: GOTPC relocation. With the
       `WRT ..gotpc' qualifier specified, the symbol referenced (here
       `_GLOBAL_OFFSET_TABLE_', the special symbol assigned to the GOT) is
       given as an offset from the beginning of the section. (Actually,
       `ELF' encodes it as the offset from the operand field of the `ADD'
       instruction, but NASM simplifies this deliberately, so you do things
       the same way for both `ELF' and `BSD'.) So the instruction then
       _adds_ the beginning of the section, to get the real address of the
       GOT, and subtracts the value of `.get_GOT' which it knows is in
       `EBX'. Therefore, by the time that instruction has finished, `EBX'
       contains the address of the GOT.

       If you didn't follow that, don't worry: it's never necessary to
       obtain the address of the GOT by any other means, so you can put
       those three instructions into a macro and safely ignore them:

       %macro  get_GOT 0
      
               call    %%getgot
         %%getgot:
               pop     ebx
               add     ebx,_GLOBAL_OFFSET_TABLE_+$$-%%getgot wrt ..gotpc
      
       %endmacro

9.2.2 Finding Your Local Data Items

       Having got the GOT, you can then use it to obtain the addresses of
       your data items. Most variables will reside in the sections you have
       declared; they can be accessed using the `..gotoff' special `WRT'
       type. The way this works is like this:

               lea     eax,[ebx+myvar wrt ..gotoff]

       The expression `myvar wrt ..gotoff' is calculated, when the shared
       library is linked, to be the offset to the local variable `myvar'
       from the beginning of the GOT. Therefore, adding it to `EBX' as
       above will place the real address of `myvar' in `EAX'.

       If you declare variables as `GLOBAL' without specifying a size for
       them, they are shared between code modules in the library, but do
       not get exported from the library to the program that loaded it.
       They will still be in your ordinary data and BSS sections, so you
       can access them in the same way as local variables, using the above
       `..gotoff' mechanism.

       Note that due to a peculiarity of the way BSD `a.out' format handles
       this relocation type, there must be at least one non-local symbol in
       the same section as the address you're trying to access.

9.2.3 Finding External and Common Data Items

       If your library needs to get at an external variable (external to
       the _library_, not just to one of the modules within it), you must
       use the `..got' type to get at it. The `..got' type, instead of
       giving you the offset from the GOT base to the variable, gives you
       the offset from the GOT base to a GOT _entry_ containing the address
       of the variable. The linker will set up this GOT entry when it
       builds the library, and the dynamic linker will place the correct
       address in it at load time. So to obtain the address of an external
       variable `extvar' in `EAX', you would code

               mov     eax,[ebx+extvar wrt ..got]

       This loads the address of `extvar' out of an entry in the GOT. The
       linker, when it builds the shared library, collects together every
       relocation of type `..got', and builds the GOT so as to ensure it
       has every necessary entry present.

       Common variables must also be accessed in this way.

9.2.4 Exporting Symbols to the Library User

       If you want to export symbols to the user of the library, you have
       to declare whether they are functions or data, and if they are data,
       you have to give the size of the data item. This is because the
       dynamic linker has to build procedure linkage table entries for any
       exported functions, and also moves exported data items away from the
       library's data section in which they were declared.

       So to export a function to users of the library, you must use

       global  func:function           ; declare it as a function
      
       func:   push    ebp
      
               ; etc.

       And to export a data item such as an array, you would have to code

       global  array:data array.end-array      ; give the size too
      
       array:  resd    128
       .end:

       Be careful: If you export a variable to the library user, by
       declaring it as `GLOBAL' and supplying a size, the variable will end
       up living in the data section of the main program, rather than in
       your library's data section, where you declared it. So you will have
       to access your own global variable with the `..got' mechanism rather
       than `..gotoff', as if it were external (which, effectively, it has
       become).

       Equally, if you need to store the address of an exported global in
       one of your data sections, you can't do it by means of the standard
       sort of code:

       dataptr:        dd      global_data_item        ; WRONG

       NASM will interpret this code as an ordinary relocation, in which
       `global_data_item' is merely an offset from the beginning of the
       `.data' section (or whatever); so this reference will end up
       pointing at your data section instead of at the exported global
       which resides elsewhere.

       Instead of the above code, then, you must write

       dataptr:        dd      global_data_item wrt ..sym

       which makes use of the special `WRT' type `..sym' to instruct NASM
       to search the symbol table for a particular symbol at that address,
       rather than just relocating by section base.

       Either method will work for functions: referring to one of your
       functions by means of

       funcptr:        dd      my_function

       will give the user the address of the code you wrote, whereas

       funcptr:        dd      my_function wrt ..sym

       will give the address of the procedure linkage table for the
       function, which is where the calling program will _believe_ the
       function lives. Either address is a valid way to call the function.

9.2.5 Calling Procedures Outside the Library

       Calling procedures outside your shared library has to be done by
       means of a _procedure linkage table_, or PLT. The PLT is placed at a
       known offset from where the library is loaded, so the library code
       can make calls to the PLT in a position-independent way. Within the
       PLT there is code to jump to offsets contained in the GOT, so
       function calls to other shared libraries or to routines in the main
       program can be transparently passed off to their real destinations.

       To call an external routine, you must use another special PIC
       relocation type, `WRT ..plt'. This is much easier than the GOT-based
       ones: you simply replace calls such as `CALL printf' with the PLT-
       relative version `CALL printf WRT ..plt'.

9.2.6 Generating the Library File

       Having written some code modules and assembled them to `.o' files,
       you then generate your shared library with a command such as

       ld -shared -o library.so module1.o module2.o       # for ELF
       ld -Bshareable -o library.so module1.o module2.o   # for BSD

       For ELF, if your shared library is going to reside in system
       directories such as `/usr/lib' or `/lib', it is usually worth using
       the `-soname' flag to the linker, to store the final library file
       name, with a version number, into the library:

       ld -shared -soname library.so.1 -o library.so.1.2 *.o

       You would then copy `library.so.1.2' into the library directory, and
       create `library.so.1' as a symbolic link to it.

Chapter 10: Mixing 16 and 32 Bit Code
-------------------------------------

       This chapter tries to cover some of the issues, largely related to
       unusual forms of addressing and jump instructions, encountered when
       writing operating system code such as protected-mode initialisation
       routines, which require code that operates in mixed segment sizes,
       such as code in a 16-bit segment trying to modify data in a 32-bit
       one, or jumps between different-size segments.

  10.1 Mixed-Size Jumps

       The most common form of mixed-size instruction is the one used when
       writing a 32-bit OS: having done your setup in 16-bit mode, such as
       loading the kernel, you then have to boot it by switching into
       protected mode and jumping to the 32-bit kernel start address. In a
       fully 32-bit OS, this tends to be the _only_ mixed-size instruction
       you need, since everything before it can be done in pure 16-bit
       code, and everything after it can be pure 32-bit.

       This jump must specify a 48-bit far address, since the target
       segment is a 32-bit one. However, it must be assembled in a 16-bit
       segment, so just coding, for example,

               jmp     0x1234:0x56789ABC       ; wrong!

       will not work, since the offset part of the address will be
       truncated to `0x9ABC' and the jump will be an ordinary 16-bit far
       one.

       The Linux kernel setup code gets round the inability of `as86' to
       generate the required instruction by coding it manually, using `DB'
       instructions. NASM can go one better than that, by actually
       generating the right instruction itself. Here's how to do it right:

               jmp     dword 0x1234:0x56789ABC         ; right

       The `DWORD' prefix (strictly speaking, it should come _after_ the
       colon, since it is declaring the _offset_ field to be a doubleword;
       but NASM will accept either form, since both are unambiguous) forces
       the offset part to be treated as far, in the assumption that you are
       deliberately writing a jump from a 16-bit segment to a 32-bit one.

       You can do the reverse operation, jumping from a 32-bit segment to a
       16-bit one, by means of the `WORD' prefix:

               jmp     word 0x8765:0x4321      ; 32 to 16 bit

       If the `WORD' prefix is specified in 16-bit mode, or the `DWORD'
       prefix in 32-bit mode, they will be ignored, since each is
       explicitly forcing NASM into a mode it was in anyway.

  10.2 Addressing Between Different-Size Segments

       If your OS is mixed 16 and 32-bit, or if you are writing a DOS
       extender, you are likely to have to deal with some 16-bit segments
       and some 32-bit ones. At some point, you will probably end up
       writing code in a 16-bit segment which has to access data in a 32-
       bit segment, or vice versa.

       If the data you are trying to access in a 32-bit segment lies within
       the first 64K of the segment, you may be able to get away with using
       an ordinary 16-bit addressing operation for the purpose; but sooner
       or later, you will want to do 32-bit addressing from 16-bit mode.

       The easiest way to do this is to make sure you use a register for
       the address, since any effective address containing a 32-bit
       register is forced to be a 32-bit address. So you can do

               mov     eax,offset_into_32_bit_segment_specified_by_fs
               mov     dword [fs:eax],0x11223344

       This is fine, but slightly cumbersome (since it wastes an
       instruction and a register) if you already know the precise offset
       you are aiming at. The x86 architecture does allow 32-bit effective
       addresses to specify nothing but a 4-byte offset, so why shouldn't
       NASM be able to generate the best instruction for the purpose?

       It can. As in section 10.1, you need only prefix the address with
       the `DWORD' keyword, and it will be forced to be a 32-bit address:

               mov     dword [fs:dword my_offset],0x11223344

       Also as in section 10.1, NASM is not fussy about whether the `DWORD'
       prefix comes before or after the segment override, so arguably a
       nicer-looking way to code the above instruction is

               mov     dword [dword fs:my_offset],0x11223344

       Don't confuse the `DWORD' prefix _outside_ the square brackets,
       which controls the size of the data stored at the address, with the
       one `inside' the square brackets which controls the length of the
       address itself. The two can quite easily be different:

               mov     word [dword 0x12345678],0x9ABC

       This moves 16 bits of data to an address specified by a 32-bit
       offset.

       You can also specify `WORD' or `DWORD' prefixes along with the `FAR'
       prefix to indirect far jumps or calls. For example:

               call    dword far [fs:word 0x4321]

       This instruction contains an address specified by a 16-bit offset;
       it loads a 48-bit far pointer from that (16-bit segment and 32-bit
       offset), and calls that address.

  10.3 Other Mixed-Size Instructions

       The other way you might want to access data might be using the
       string instructions (`LODSx', `STOSx' and so on) or the `XLATB'
       instruction. These instructions, since they take no parameters,
       might seem to have no easy way to make them perform 32-bit
       addressing when assembled in a 16-bit segment.

       This is the purpose of NASM's `a16', `a32' and `a64' prefixes. If
       you are coding `LODSB' in a 16-bit segment but it is supposed to be
       accessing a string in a 32-bit segment, you should load the desired
       address into `ESI' and then code

               a32     lodsb

       The prefix forces the addressing size to 32 bits, meaning that
       `LODSB' loads from `[DS:ESI]' instead of `[DS:SI]'. To access a
       string in a 16-bit segment when coding in a 32-bit one, the
       corresponding `a16' prefix can be used.

       The `a16', `a32' and `a64' prefixes can be applied to any
       instruction in NASM's instruction table, but most of them can
       generate all the useful forms without them. The prefixes are
       necessary only for instructions with implicit addressing: `CMPSx',
       `SCASx', `LODSx', `STOSx', `MOVSx', `INSx', `OUTSx', and `XLATB'.
       Also, the various push and pop instructions (`PUSHA' and `POPF' as
       well as the more usual `PUSH' and `POP') can accept `a16', `a32' or
       `a64' prefixes to force a particular one of `SP', `ESP' or `RSP' to
       be used as a stack pointer, in case the stack segment in use is a
       different size from the code segment.

       `PUSH' and `POP', when applied to segment registers in 32-bit mode,
       also have the slightly odd behaviour that they push and pop 4 bytes
       at a time, of which the top two are ignored and the bottom two give
       the value of the segment register being manipulated. To force the
       16-bit behaviour of segment-register push and pop instructions, you
       can use the operand-size prefix `o16':

               o16 push    ss
               o16 push    ds

       This code saves a doubleword of stack space by fitting two segment
       registers into the space which would normally be consumed by pushing
       one.

       (You can also use the `o32' prefix to force the 32-bit behaviour
       when in 16-bit mode, but this seems less useful.)

Chapter 11: Writing 64-bit Code (Unix, Win64)
---------------------------------------------

       This chapter attempts to cover some of the common issues involved
       when writing 64-bit code, to run under Win64 or Unix. It covers how
       to write assembly code to interface with 64-bit C routines, and how
       to write position-independent code for shared libraries.

       All 64-bit code uses a flat memory model, since segmentation is not
       available in 64-bit mode. The one exception is the `FS' and `GS'
       registers, which still add their bases.

       Position independence in 64-bit mode is significantly simpler, since
       the processor supports `RIP'-relative addressing directly; see the
       `REL' keyword (section 3.3). On most 64-bit platforms, it is
       probably desirable to make that the default, using the directive
       `DEFAULT REL' (section 6.2).

       64-bit programming is relatively similar to 32-bit programming, but
       of course pointers are 64 bits long; additionally, all existing
       platforms pass arguments in registers rather than on the stack.
       Furthermore, 64-bit platforms use SSE2 by default for floating
       point. Please see the ABI documentation for your platform.

       64-bit platforms differ in the sizes of the fundamental datatypes,
       not just from 32-bit platforms but from each other. If a specific
       size data type is desired, it is probably best to use the types
       defined in the Standard C header `<inttypes.h>'.

       In 64-bit mode, the default instruction size is still 32 bits. When
       loading a value into a 32-bit register (but not an 8- or 16-bit
       register), the upper 32 bits of the corresponding 64-bit register
       are set to zero.

  11.1 Register Names in 64-bit Mode

       NASM uses the following names for general-purpose registers in 64-
       bit mode, for 8-, 16-, 32- and 64-bit references, respectively:

            AL/AH, CL/CH, DL/DH, BL/BH, SPL, BPL, SIL, DIL, R8B-R15B
            AX, CX, DX, BX, SP, BP, SI, DI, R8W-R15W
            EAX, ECX, EDX, EBX, ESP, EBP, ESI, EDI, R8D-R15D
            RAX, RCX, RDX, RBX, RSP, RBP, RSI, RDI, R8-R15

       This is consistent with the AMD documentation and most other
       assemblers. The Intel documentation, however, uses the names
       `R8L-R15L' for 8-bit references to the higher registers. It is
       possible to use those names by definiting them as macros; similarly,
       if one wants to use numeric names for the low 8 registers, define
       them as macros. The standard macro package `altreg' (see section
       5.1) can be used for this purpose.

  11.2 Immediates and Displacements in 64-bit Mode

       In 64-bit mode, immediates and displacements are generally only 32
       bits wide. NASM will therefore truncate most displacements and
       immediates to 32 bits.

       The only instruction which takes a full 64-bit immediate is:

            MOV reg64,imm64

       NASM will produce this instruction whenever the programmer uses
       `MOV' with an immediate into a 64-bit register. If this is not
       desirable, simply specify the equivalent 32-bit register, which will
       be automatically zero-extended by the processor, or specify the
       immediate as `DWORD':

            mov rax,foo             ; 64-bit immediate
            mov rax,qword foo       ; (identical)
            mov eax,foo             ; 32-bit immediate, zero-extended
            mov rax,dword foo       ; 32-bit immediate, sign-extended

       The length of these instructions are 10, 5 and 7 bytes,
       respectively.

       The only instructions which take a full 64-bit _displacement_ is
       loading or storing, using `MOV', `AL', `AX', `EAX' or `RAX' (but no
       other registers) to an absolute 64-bit address. Since this is a
       relatively rarely used instruction (64-bit code generally uses
       relative addressing), the programmer has to explicitly declare the
       displacement size as `QWORD':

            default abs
      
            mov eax,[foo]           ; 32-bit absolute disp, sign-extended
            mov eax,[a32 foo]       ; 32-bit absolute disp, zero-extended
            mov eax,[qword foo]     ; 64-bit absolute disp
      
            default rel
      
            mov eax,[foo]           ; 32-bit relative disp
            mov eax,[a32 foo]       ; d:o, address truncated to 32 bits(!)
            mov eax,[qword foo]     ; error
            mov eax,[abs qword foo] ; 64-bit absolute disp

       A sign-extended absolute displacement can access from -2 GB to +2
       GB; a zero-extended absolute displacement can access from 0 to 4 GB.

  11.3 Interfacing to 64-bit C Programs (Unix)

       On Unix, the 64-bit ABI is defined by the document:

       `http://www.nasm.us/links/unix64abi'

       Although written for AT&T-syntax assembly, the concepts apply
       equally well for NASM-style assembly. What follows is a simplified
       summary.

       The first six integer arguments (from the left) are passed in `RDI',
       `RSI', `RDX', `RCX', `R8', and `R9', in that order. Additional
       integer arguments are passed on the stack. These registers, plus
       `RAX', `R10' and `R11' are destroyed by function calls, and thus are
       available for use by the function without saving.

       Integer return values are passed in `RAX' and `RDX', in that order.

       Floating point is done using SSE registers, except for
       `long double'. Floating-point arguments are passed in `XMM0' to
       `XMM7'; return is `XMM0' and `XMM1'. `long double' are passed on the
       stack, and returned in `ST0' and `ST1'.

       All SSE and x87 registers are destroyed by function calls.

       On 64-bit Unix, `long' is 64 bits.

       Integer and SSE register arguments are counted separately, so for
       the case of

            void foo(long a, double b, int c)

       `a' is passed in `RDI', `b' in `XMM0', and `c' in `ESI'.

  11.4 Interfacing to 64-bit C Programs (Win64)

       The Win64 ABI is described at:

       `http://www.nasm.us/links/win64abi'

       What follows is a simplified summary.

       The first four integer arguments are passed in `RCX', `RDX', `R8'
       and `R9', in that order. Additional integer arguments are passed on
       the stack. These registers, plus `RAX', `R10' and `R11' are
       destroyed by function calls, and thus are available for use by the
       function without saving.

       Integer return values are passed in `RAX' only.

       Floating point is done using SSE registers, except for
       `long double'. Floating-point arguments are passed in `XMM0' to
       `XMM3'; return is `XMM0' only.

       On Win64, `long' is 32 bits; `long long' or `_int64' is 64 bits.

       Integer and SSE register arguments are counted together, so for the
       case of

            void foo(long long a, double b, int c)

       `a' is passed in `RCX', `b' in `XMM1', and `c' in `R8D'.

Chapter 12: Troubleshooting
---------------------------

       This chapter describes some of the common problems that users have
       been known to encounter with NASM, and answers them. It also gives
       instructions for reporting bugs in NASM if you find a difficulty
       that isn't listed here.

  12.1 Common Problems

12.1.1 NASM Generates Inefficient Code

       We sometimes get `bug' reports about NASM generating inefficient, or
       even `wrong', code on instructions such as `ADD ESP,8'. This is a
       deliberate design feature, connected to predictability of output:
       NASM, on seeing `ADD ESP,8', will generate the form of the
       instruction which leaves room for a 32-bit offset. You need to code
       `ADD ESP,BYTE 8' if you want the space-efficient form of the
       instruction. This isn't a bug, it's user error: if you prefer to
       have NASM produce the more efficient code automatically enable
       optimization with the `-O' option (see section 2.1.22).

12.1.2 My Jumps are Out of Range

       Similarly, people complain that when they issue conditional jumps
       (which are `SHORT' by default) that try to jump too far, NASM
       reports `short jump out of range' instead of making the jumps
       longer.

       This, again, is partly a predictability issue, but in fact has a
       more practical reason as well. NASM has no means of being told what
       type of processor the code it is generating will be run on; so it
       cannot decide for itself that it should generate `Jcc NEAR' type
       instructions, because it doesn't know that it's working for a 386 or
       above. Alternatively, it could replace the out-of-range short `JNE'
       instruction with a very short `JE' instruction that jumps over a
       `JMP NEAR'; this is a sensible solution for processors below a 386,
       but hardly efficient on processors which have good branch prediction
       _and_ could have used `JNE NEAR' instead. So, once again, it's up to
       the user, not the assembler, to decide what instructions should be
       generated. See section 2.1.22.

12.1.3 `ORG' Doesn't Work

       People writing boot sector programs in the `bin' format often
       complain that `ORG' doesn't work the way they'd like: in order to
       place the `0xAA55' signature word at the end of a 512-byte boot
       sector, people who are used to MASM tend to code

               ORG 0
      
               ; some boot sector code
      
               ORG 510
               DW 0xAA55

       This is not the intended use of the `ORG' directive in NASM, and
       will not work. The correct way to solve this problem in NASM is to
       use the `TIMES' directive, like this:

               ORG 0
      
               ; some boot sector code
      
               TIMES 510-($-$$) DB 0
               DW 0xAA55

       The `TIMES' directive will insert exactly enough zero bytes into the
       output to move the assembly point up to 510. This method also has
       the advantage that if you accidentally fill your boot sector too
       full, NASM will catch the problem at assembly time and report it, so
       you won't end up with a boot sector that you have to disassemble to
       find out what's wrong with it.

12.1.4 `TIMES' Doesn't Work

       The other common problem with the above code is people who write the
       `TIMES' line as

               TIMES 510-$ DB 0

       by reasoning that `$' should be a pure number, just like 510, so the
       difference between them is also a pure number and can happily be fed
       to `TIMES'.

       NASM is a _modular_ assembler: the various component parts are
       designed to be easily separable for re-use, so they don't exchange
       information unnecessarily. In consequence, the `bin' output format,
       even though it has been told by the `ORG' directive that the `.text'
       section should start at 0, does not pass that information back to
       the expression evaluator. So from the evaluator's point of view, `$'
       isn't a pure number: it's an offset from a section base. Therefore
       the difference between `$' and 510 is also not a pure number, but
       involves a section base. Values involving section bases cannot be
       passed as arguments to `TIMES'.

       The solution, as in the previous section, is to code the `TIMES'
       line in the form

               TIMES 510-($-$$) DB 0

       in which `$' and `$$' are offsets from the same section base, and so
       their difference is a pure number. This will solve the problem and
       generate sensible code.

  12.2 Bugs

       We have never yet released a version of NASM with any _known_ bugs.
       That doesn't usually stop there being plenty we didn't know about,
       though. Any that you find should be reported firstly via the
       `bugtracker' at `http://www.nasm.us/' (click on "Bug Tracker"), or
       if that fails then through one of the contacts in section 1.2.

       Please read section 2.2 first, and don't report the bug if it's
       listed in there as a deliberate feature. (If you think the feature
       is badly thought out, feel free to send us reasons why you think it
       should be changed, but don't just send us mail saying `This is a
       bug' if the documentation says we did it on purpose.) Then read
       section 12.1, and don't bother reporting the bug if it's listed
       there.

       If you do report a bug, _please_ give us all of the following
       information:

       (*) What operating system you're running NASM under. DOS, Linux,
           NetBSD, Win16, Win32, VMS (I'd be impressed), whatever.

       (*) If you're running NASM under DOS or Win32, tell us whether
           you've compiled your own executable from the DOS source archive,
           or whether you were using the standard distribution binaries out
           of the archive. If you were using a locally built executable,
           try to reproduce the problem using one of the standard binaries,
           as this will make it easier for us to reproduce your problem
           prior to fixing it.

       (*) Which version of NASM you're using, and exactly how you invoked
           it. Give us the precise command line, and the contents of the
           `NASMENV' environment variable if any.

       (*) Which versions of any supplementary programs you're using, and
           how you invoked them. If the problem only becomes visible at
           link time, tell us what linker you're using, what version of it
           you've got, and the exact linker command line. If the problem
           involves linking against object files generated by a compiler,
           tell us what compiler, what version, and what command line or
           options you used. (If you're compiling in an IDE, please try to
           reproduce the problem with the command-line version of the
           compiler.)

       (*) If at all possible, send us a NASM source file which exhibits
           the problem. If this causes copyright problems (e.g. you can
           only reproduce the bug in restricted-distribution code) then
           bear in mind the following two points: firstly, we guarantee
           that any source code sent to us for the purposes of debugging
           NASM will be used _only_ for the purposes of debugging NASM, and
           that we will delete all our copies of it as soon as we have
           found and fixed the bug or bugs in question; and secondly, we
           would prefer _not_ to be mailed large chunks of code anyway. The
           smaller the file, the better. A three-line sample file that does
           nothing useful _except_ demonstrate the problem is much easier
           to work with than a fully fledged ten-thousand-line program. (Of
           course, some errors _do_ only crop up in large files, so this
           may not be possible.)

       (*) A description of what the problem actually _is_. `It doesn't
           work' is _not_ a helpful description! Please describe exactly
           what is happening that shouldn't be, or what isn't happening
           that should. Examples might be: `NASM generates an error message
           saying Line 3 for an error that's actually on Line 5'; `NASM
           generates an error message that I believe it shouldn't be
           generating at all'; `NASM fails to generate an error message
           that I believe it _should_ be generating'; `the object file
           produced from this source code crashes my linker'; `the ninth
           byte of the output file is 66 and I think it should be 77
           instead'.

       (*) If you believe the output file from NASM to be faulty, send it
           to us. That allows us to determine whether our own copy of NASM
           generates the same file, or whether the problem is related to
           portability issues between our development platforms and yours.
           We can handle binary files mailed to us as MIME attachments,
           uuencoded, and even BinHex. Alternatively, we may be able to
           provide an FTP site you can upload the suspect files to; but
           mailing them is easier for us.

       (*) Any other information or data files that might be helpful. If,
           for example, the problem involves NASM failing to generate an
           object file while TASM can generate an equivalent file without
           trouble, then send us _both_ object files, so we can see what
           TASM is doing differently from us.

Appendix A: Ndisasm
-------------------

       The Netwide Disassembler, NDISASM

   A.1 Introduction

       The Netwide Disassembler is a small companion program to the Netwide
       Assembler, NASM. It seemed a shame to have an x86 assembler,
       complete with a full instruction table, and not make as much use of
       it as possible, so here's a disassembler which shares the
       instruction table (and some other bits of code) with NASM.

       The Netwide Disassembler does nothing except to produce
       disassemblies of _binary_ source files. NDISASM does not have any
       understanding of object file formats, like `objdump', and it will
       not understand `DOS .EXE' files like `debug' will. It just
       disassembles.

   A.2 Getting Started: Installation

       See section 1.3 for installation instructions. NDISASM, like NASM,
       has a `man page' which you may want to put somewhere useful, if you
       are on a Unix system.

   A.3 Running NDISASM

       To disassemble a file, you will typically use a command of the form

              ndisasm -b {16|32|64} filename

       NDISASM can disassemble 16-, 32- or 64-bit code equally easily,
       provided of course that you remember to specify which it is to work
       with. If no `-b' switch is present, NDISASM works in 16-bit mode by
       default. The `-u' switch (for USE32) also invokes 32-bit mode.

       Two more command line options are `-r' which reports the version
       number of NDISASM you are running, and `-h' which gives a short
       summary of command line options.

A.3.1 COM Files: Specifying an Origin

       To disassemble a `DOS .COM' file correctly, a disassembler must
       assume that the first instruction in the file is loaded at address
       `0x100', rather than at zero. NDISASM, which assumes by default that
       any file you give it is loaded at zero, will therefore need to be
       informed of this.

       The `-o' option allows you to declare a different origin for the
       file you are disassembling. Its argument may be expressed in any of
       the NASM numeric formats: decimal by default, if it begins with
       ``$'' or ``0x'' or ends in ``H'' it's `hex', if it ends in ``Q''
       it's `octal', and if it ends in ``B'' it's `binary'.

       Hence, to disassemble a `.COM' file:

              ndisasm -o100h filename.com

       will do the trick.

A.3.2 Code Following Data: Synchronisation

       Suppose you are disassembling a file which contains some data which
       isn't machine code, and _then_ contains some machine code. NDISASM
       will faithfully plough through the data section, producing machine
       instructions wherever it can (although most of them will look
       bizarre, and some may have unusual prefixes, e.g.
       ``FS OR AX,0x240A''), and generating `DB' instructions ever so often
       if it's totally stumped. Then it will reach the code section.

       Supposing NDISASM has just finished generating a strange machine
       instruction from part of the data section, and its file position is
       now one byte _before_ the beginning of the code section. It's
       entirely possible that another spurious instruction will get
       generated, starting with the final byte of the data section, and
       then the correct first instruction in the code section will not be
       seen because the starting point skipped over it. This isn't really
       ideal.

       To avoid this, you can specify a ``synchronisation'' point, or
       indeed as many synchronisation points as you like (although NDISASM
       can only handle 2147483647 sync points internally). The definition
       of a sync point is this: NDISASM guarantees to hit sync points
       exactly during disassembly. If it is thinking about generating an
       instruction which would cause it to jump over a sync point, it will
       discard that instruction and output a ``db'' instead. So it _will_
       start disassembly exactly from the sync point, and so you _will_ see
       all the instructions in your code section.

       Sync points are specified using the `-s' option: they are measured
       in terms of the program origin, not the file position. So if you
       want to synchronize after 32 bytes of a `.COM' file, you would have
       to do

              ndisasm -o100h -s120h file.com

       rather than

              ndisasm -o100h -s20h file.com

       As stated above, you can specify multiple sync markers if you need
       to, just by repeating the `-s' option.

A.3.3 Mixed Code and Data: Automatic (Intelligent) Synchronisation

       Suppose you are disassembling the boot sector of a `DOS' floppy
       (maybe it has a virus, and you need to understand the virus so that
       you know what kinds of damage it might have done you). Typically,
       this will contain a `JMP' instruction, then some data, then the rest
       of the code. So there is a very good chance of NDISASM being
       _misaligned_ when the data ends and the code begins. Hence a sync
       point is needed.

       On the other hand, why should you have to specify the sync point
       manually? What you'd do in order to find where the sync point would
       be, surely, would be to read the `JMP' instruction, and then to use
       its target address as a sync point. So can NDISASM do that for you?

       The answer, of course, is yes: using either of the synonymous
       switches `-a' (for automatic sync) or `-i' (for intelligent sync)
       will enable `auto-sync' mode. Auto-sync mode automatically generates
       a sync point for any forward-referring PC-relative jump or call
       instruction that NDISASM encounters. (Since NDISASM is one-pass, if
       it encounters a PC-relative jump whose target has already been
       processed, there isn't much it can do about it...)

       Only PC-relative jumps are processed, since an absolute jump is
       either through a register (in which case NDISASM doesn't know what
       the register contains) or involves a segment address (in which case
       the target code isn't in the same segment that NDISASM is working
       in, and so the sync point can't be placed anywhere useful).

       For some kinds of file, this mechanism will automatically put sync
       points in all the right places, and save you from having to place
       any sync points manually. However, it should be stressed that auto-
       sync mode is _not_ guaranteed to catch all the sync points, and you
       may still have to place some manually.

       Auto-sync mode doesn't prevent you from declaring manual sync
       points: it just adds automatically generated ones to the ones you
       provide. It's perfectly feasible to specify `-i' _and_ some `-s'
       options.

       Another caveat with auto-sync mode is that if, by some unpleasant
       fluke, something in your data section should disassemble to a PC-
       relative call or jump instruction, NDISASM may obediently place a
       sync point in a totally random place, for example in the middle of
       one of the instructions in your code section. So you may end up with
       a wrong disassembly even if you use auto-sync. Again, there isn't
       much I can do about this. If you have problems, you'll have to use
       manual sync points, or use the `-k' option (documented below) to
       suppress disassembly of the data area.

A.3.4 Other Options

       The `-e' option skips a header on the file, by ignoring the first N
       bytes. This means that the header is _not_ counted towards the
       disassembly offset: if you give `-e10 -o10', disassembly will start
       at byte 10 in the file, and this will be given offset 10, not 20.

       The `-k' option is provided with two comma-separated numeric
       arguments, the first of which is an assembly offset and the second
       is a number of bytes to skip. This _will_ count the skipped bytes
       towards the assembly offset: its use is to suppress disassembly of a
       data section which wouldn't contain anything you wanted to see
       anyway.

   A.4 Bugs and Improvements

       There are no known bugs. However, any you find, with patches if
       possible, should be sent to `nasm-bugs@lists.sourceforge.net', or to
       the developer's site at `http://www.nasm.us/' and we'll try to fix
       them. Feel free to send contributions and new features as well.

Appendix B: Instruction List
----------------------------

   B.1 Introduction

       The following sections show the instructions which NASM currently
       supports. For each instruction, there is a separate entry for each
       supported addressing mode. The third column shows the processor type
       in which the instruction was introduced and, when appropriate, one
       or more usage flags.

B.1.1 Special instructions...

       DB                                         
       DW                                         
       DD                                         
       DQ                                         
       DT                                         
       DO                                         
       DY                                         
       DZ                                         
       RESB             imm                      8086
       RESW                                       
       RESD                                       
       RESQ                                       
       REST                                       
       RESO                                       
       RESY                                       
       RESZ                                      

B.1.2 Conventional instructions

       AAA                                       8086,NOLONG
       AAD                                       8086,NOLONG
       AAD              imm                      8086,NOLONG
       AAM                                       8086,NOLONG
       AAM              imm                      8086,NOLONG
       AAS                                       8086,NOLONG
       ADC              mem,reg8                 8086,LOCK
       ADC              reg8,reg8                8086
       ADC              mem,reg16                8086,LOCK
       ADC              reg16,reg16              8086
       ADC              mem,reg32                386,LOCK
       ADC              reg32,reg32              386
       ADC              mem,reg64                X64,LOCK
       ADC              reg64,reg64              X64
       ADC              reg8,mem                 8086
       ADC              reg8,reg8                8086
       ADC              reg16,mem                8086
       ADC              reg16,reg16              8086
       ADC              reg32,mem                386
       ADC              reg32,reg32              386
       ADC              reg64,mem                X64
       ADC              reg64,reg64              X64
       ADC              rm16,imm8                8086,LOCK
       ADC              rm32,imm8                386,LOCK
       ADC              rm64,imm8                X64,LOCK
       ADC              reg_al,imm               8086
       ADC              reg_ax,sbyteword         8086,ND
       ADC              reg_ax,imm               8086
       ADC              reg_eax,sbytedword       386,ND
       ADC              reg_eax,imm              386
       ADC              reg_rax,sbytedword       X64,ND
       ADC              reg_rax,imm              X64
       ADC              rm8,imm                  8086,LOCK
       ADC              rm16,sbyteword           8086,LOCK,ND
       ADC              rm16,imm                 8086,LOCK
       ADC              rm32,sbytedword          386,LOCK,ND
       ADC              rm32,imm                 386,LOCK
       ADC              rm64,sbytedword          X64,LOCK,ND
       ADC              rm64,imm                 X64,LOCK
       ADC              mem,imm8                 8086,LOCK,ND
       ADC              mem,sbyteword16          8086,LOCK,ND
       ADC              mem,imm16                8086,LOCK
       ADC              mem,sbytedword32         386,LOCK,ND
       ADC              mem,imm32                386,LOCK
       ADC              rm8,imm                  8086,LOCK,ND,NOLONG
       ADD              mem,reg8                 8086,LOCK
       ADD              reg8,reg8                8086
       ADD              mem,reg16                8086,LOCK
       ADD              reg16,reg16              8086
       ADD              mem,reg32                386,LOCK
       ADD              reg32,reg32              386
       ADD              mem,reg64                X64,LOCK
       ADD              reg64,reg64              X64
       ADD              reg8,mem                 8086
       ADD              reg8,reg8                8086
       ADD              reg16,mem                8086
       ADD              reg16,reg16              8086
       ADD              reg32,mem                386
       ADD              reg32,reg32              386
       ADD              reg64,mem                X64
       ADD              reg64,reg64              X64
       ADD              rm16,imm8                8086,LOCK
       ADD              rm32,imm8                386,LOCK
       ADD              rm64,imm8                X64,LOCK
       ADD              reg_al,imm               8086
       ADD              reg_ax,sbyteword         8086,ND
       ADD              reg_ax,imm               8086
       ADD              reg_eax,sbytedword       386,ND
       ADD              reg_eax,imm              386
       ADD              reg_rax,sbytedword       X64,ND
       ADD              reg_rax,imm              X64
       ADD              rm8,imm                  8086,LOCK
       ADD              rm16,sbyteword           8086,LOCK,ND
       ADD              rm16,imm                 8086,LOCK
       ADD              rm32,sbytedword          386,LOCK,ND
       ADD              rm32,imm                 386,LOCK
       ADD              rm64,sbytedword          X64,LOCK,ND
       ADD              rm64,imm                 X64,LOCK
       ADD              mem,imm8                 8086,LOCK
       ADD              mem,sbyteword16          8086,LOCK,ND
       ADD              mem,imm16                8086,LOCK
       ADD              mem,sbytedword32         386,LOCK,ND
       ADD              mem,imm32                386,LOCK
       ADD              rm8,imm                  8086,LOCK,ND,NOLONG
       AND              mem,reg8                 8086,LOCK
       AND              reg8,reg8                8086
       AND              mem,reg16                8086,LOCK
       AND              reg16,reg16              8086
       AND              mem,reg32                386,LOCK
       AND              reg32,reg32              386
       AND              mem,reg64                X64,LOCK
       AND              reg64,reg64              X64
       AND              reg8,mem                 8086
       AND              reg8,reg8                8086
       AND              reg16,mem                8086
       AND              reg16,reg16              8086
       AND              reg32,mem                386
       AND              reg32,reg32              386
       AND              reg64,mem                X64
       AND              reg64,reg64              X64
       AND              rm16,imm8                8086,LOCK
       AND              rm32,imm8                386,LOCK
       AND              rm64,imm8                X64,LOCK
       AND              reg_al,imm               8086
       AND              reg_ax,sbyteword         8086,ND
       AND              reg_ax,imm               8086
       AND              reg_eax,sbytedword       386,ND
       AND              reg_eax,imm              386
       AND              reg_rax,sbytedword       X64,ND
       AND              reg_rax,imm              X64
       AND              rm8,imm                  8086,LOCK
       AND              rm16,sbyteword           8086,LOCK,ND
       AND              rm16,imm                 8086,LOCK
       AND              rm32,sbytedword          386,LOCK,ND
       AND              rm32,imm                 386,LOCK
       AND              rm64,sbytedword          X64,LOCK,ND
       AND              rm64,imm                 X64,LOCK
       AND              mem,imm8                 8086,LOCK
       AND              mem,sbyteword16          8086,LOCK,ND
       AND              mem,imm16                8086,LOCK
       AND              mem,sbytedword32         386,LOCK,ND
       AND              mem,imm32                386,LOCK
       AND              rm8,imm                  8086,LOCK,ND,NOLONG
       ARPL             mem,reg16                286,PROT,NOLONG
       ARPL             reg16,reg16              286,PROT,NOLONG
       BB0_RESET                                 PENT,CYRIX,ND
       BB1_RESET                                 PENT,CYRIX,ND
       BOUND            reg16,mem                186,NOLONG
       BOUND            reg32,mem                386,NOLONG
       BSF              reg16,mem                386
       BSF              reg16,reg16              386
       BSF              reg32,mem                386
       BSF              reg32,reg32              386
       BSF              reg64,mem                X64
       BSF              reg64,reg64              X64
       BSR              reg16,mem                386
       BSR              reg16,reg16              386
       BSR              reg32,mem                386
       BSR              reg32,reg32              386
       BSR              reg64,mem                X64
       BSR              reg64,reg64              X64
       BSWAP            reg32                    486
       BSWAP            reg64                    X64
       BT               mem,reg16                386
       BT               reg16,reg16              386
       BT               mem,reg32                386
       BT               reg32,reg32              386
       BT               mem,reg64                X64
       BT               reg64,reg64              X64
       BT               rm16,imm                 386
       BT               rm32,imm                 386
       BT               rm64,imm                 X64
       BTC              mem,reg16                386,LOCK
       BTC              reg16,reg16              386
       BTC              mem,reg32                386,LOCK
       BTC              reg32,reg32              386
       BTC              mem,reg64                X64,LOCK
       BTC              reg64,reg64              X64
       BTC              rm16,imm                 386,LOCK
       BTC              rm32,imm                 386,LOCK
       BTC              rm64,imm                 X64,LOCK
       BTR              mem,reg16                386,LOCK
       BTR              reg16,reg16              386
       BTR              mem,reg32                386,LOCK
       BTR              reg32,reg32              386
       BTR              mem,reg64                X64,LOCK
       BTR              reg64,reg64              X64
       BTR              rm16,imm                 386,LOCK
       BTR              rm32,imm                 386,LOCK
       BTR              rm64,imm                 X64,LOCK
       BTS              mem,reg16                386,LOCK
       BTS              reg16,reg16              386
       BTS              mem,reg32                386,LOCK
       BTS              reg32,reg32              386
       BTS              mem,reg64                X64,LOCK
       BTS              reg64,reg64              X64
       BTS              rm16,imm                 386,LOCK
       BTS              rm32,imm                 386,LOCK
       BTS              rm64,imm                 X64,LOCK
       CALL             imm                      8086,BND
       CALL             imm|near                 8086,ND,BND
       CALL             imm|far                  8086,ND,NOLONG
       CALL             imm16                    8086,NOLONG,BND
       CALL             imm16|near               8086,ND,NOLONG,BND
       CALL             imm16|far                8086,ND,NOLONG
       CALL             imm32                    386,NOLONG,BND
       CALL             imm32|near               386,ND,NOLONG,BND
       CALL             imm32|far                386,ND,NOLONG
       CALL             imm64                    X64,BND
       CALL             imm64|near               X64,ND,BND
       CALL             imm:imm                  8086,NOLONG
       CALL             imm16:imm                8086,NOLONG
       CALL             imm:imm16                8086,NOLONG
       CALL             imm32:imm                386,NOLONG
       CALL             imm:imm32                386,NOLONG
       CALL             mem|far                  8086,NOLONG
       CALL             mem|far                  X64
       CALL             mem16|far                8086
       CALL             mem32|far                386
       CALL             mem64|far                X64
       CALL             mem|near                 8086,ND,BND
       CALL             rm16|near                8086,NOLONG,ND,BND
       CALL             rm32|near                386,NOLONG,ND,BND
       CALL             rm64|near                X64,ND,BND
       CALL             mem                      8086,BND
       CALL             rm16                     8086,NOLONG,BND
       CALL             rm32                     386,NOLONG,BND
       CALL             rm64                     X64,BND
       CBW                                       8086
       CDQ                                       386
       CDQE                                      X64
       CLC                                       8086
       CLD                                       8086
       CLI                                       8086
       CLTS                                      286,PRIV
       CMC                                       8086
       CMP              mem,reg8                 8086
       CMP              reg8,reg8                8086
       CMP              mem,reg16                8086
       CMP              reg16,reg16              8086
       CMP              mem,reg32                386
       CMP              reg32,reg32              386
       CMP              mem,reg64                X64
       CMP              reg64,reg64              X64
       CMP              reg8,mem                 8086
       CMP              reg8,reg8                8086
       CMP              reg16,mem                8086
       CMP              reg16,reg16              8086
       CMP              reg32,mem                386
       CMP              reg32,reg32              386
       CMP              reg64,mem                X64
       CMP              reg64,reg64              X64
       CMP              rm16,imm8                8086
       CMP              rm32,imm8                386
       CMP              rm64,imm8                X64
       CMP              reg_al,imm               8086
       CMP              reg_ax,sbyteword         8086,ND
       CMP              reg_ax,imm               8086
       CMP              reg_eax,sbytedword       386,ND
       CMP              reg_eax,imm              386
       CMP              reg_rax,sbytedword       X64,ND
       CMP              reg_rax,imm              X64
       CMP              rm8,imm                  8086
       CMP              rm16,sbyteword           8086,ND
       CMP              rm16,imm                 8086
       CMP              rm32,sbytedword          386,ND
       CMP              rm32,imm                 386
       CMP              rm64,sbytedword          X64,ND
       CMP              rm64,imm                 X64
       CMP              mem,imm8                 8086
       CMP              mem,sbyteword16          8086,ND
       CMP              mem,imm16                8086
       CMP              mem,sbytedword32         386,ND
       CMP              mem,imm32                386
       CMP              rm8,imm                  8086,ND,NOLONG
       CMPSB                                     8086
       CMPSD                                     386
       CMPSQ                                     X64
       CMPSW                                     8086
       CMPXCHG          mem,reg8                 PENT,LOCK
       CMPXCHG          reg8,reg8                PENT
       CMPXCHG          mem,reg16                PENT,LOCK
       CMPXCHG          reg16,reg16              PENT
       CMPXCHG          mem,reg32                PENT,LOCK
       CMPXCHG          reg32,reg32              PENT
       CMPXCHG          mem,reg64                X64,LOCK
       CMPXCHG          reg64,reg64              X64
       CMPXCHG486       mem,reg8                 486,UNDOC,ND,LOCK
       CMPXCHG486       reg8,reg8                486,UNDOC,ND
       CMPXCHG486       mem,reg16                486,UNDOC,ND,LOCK
       CMPXCHG486       reg16,reg16              486,UNDOC,ND
       CMPXCHG486       mem,reg32                486,UNDOC,ND,LOCK
       CMPXCHG486       reg32,reg32              486,UNDOC,ND
       CMPXCHG8B        mem                      PENT,LOCK
       CMPXCHG16B       mem                      X64,LOCK
       CPUID                                     PENT
       CPU_READ                                  PENT,CYRIX
       CPU_WRITE                                 PENT,CYRIX
       CQO                                       X64
       CWD                                       8086
       CWDE                                      386
       DAA                                       8086,NOLONG
       DAS                                       8086,NOLONG
       DEC              reg16                    8086,NOLONG
       DEC              reg32                    386,NOLONG
       DEC              rm8                      8086,LOCK
       DEC              rm16                     8086,LOCK
       DEC              rm32                     386,LOCK
       DEC              rm64                     X64,LOCK
       DIV              rm8                      8086
       DIV              rm16                     8086
       DIV              rm32                     386
       DIV              rm64                     X64
       DMINT                                     P6,CYRIX
       EMMS                                      PENT,MMX
       ENTER            imm,imm                  186
       EQU              imm                      8086
       EQU              imm:imm                  8086
       F2XM1                                     8086,FPU
       FABS                                      8086,FPU
       FADD             mem32                    8086,FPU
       FADD             mem64                    8086,FPU
       FADD             fpureg|to                8086,FPU
       FADD             fpureg                   8086,FPU
       FADD             fpureg,fpu0              8086,FPU
       FADD             fpu0,fpureg              8086,FPU
       FADD                                      8086,FPU,ND
       FADDP            fpureg                   8086,FPU
       FADDP            fpureg,fpu0              8086,FPU
       FADDP                                     8086,FPU,ND
       FBLD             mem80                    8086,FPU
       FBLD             mem                      8086,FPU
       FBSTP            mem80                    8086,FPU
       FBSTP            mem                      8086,FPU
       FCHS                                      8086,FPU
       FCLEX                                     8086,FPU
       FCMOVB           fpureg                   P6,FPU
       FCMOVB           fpu0,fpureg              P6,FPU
       FCMOVB                                    P6,FPU,ND
       FCMOVBE          fpureg                   P6,FPU
       FCMOVBE          fpu0,fpureg              P6,FPU
       FCMOVBE                                   P6,FPU,ND
       FCMOVE           fpureg                   P6,FPU
       FCMOVE           fpu0,fpureg              P6,FPU
       FCMOVE                                    P6,FPU,ND
       FCMOVNB          fpureg                   P6,FPU
       FCMOVNB          fpu0,fpureg              P6,FPU
       FCMOVNB                                   P6,FPU,ND
       FCMOVNBE         fpureg                   P6,FPU
       FCMOVNBE         fpu0,fpureg              P6,FPU
       FCMOVNBE                                  P6,FPU,ND
       FCMOVNE          fpureg                   P6,FPU
       FCMOVNE          fpu0,fpureg              P6,FPU
       FCMOVNE                                   P6,FPU,ND
       FCMOVNU          fpureg                   P6,FPU
       FCMOVNU          fpu0,fpureg              P6,FPU
       FCMOVNU                                   P6,FPU,ND
       FCMOVU           fpureg                   P6,FPU
       FCMOVU           fpu0,fpureg              P6,FPU
       FCMOVU                                    P6,FPU,ND
       FCOM             mem32                    8086,FPU
       FCOM             mem64                    8086,FPU
       FCOM             fpureg                   8086,FPU
       FCOM             fpu0,fpureg              8086,FPU
       FCOM                                      8086,FPU,ND
       FCOMI            fpureg                   P6,FPU
       FCOMI            fpu0,fpureg              P6,FPU
       FCOMI                                     P6,FPU,ND
       FCOMIP           fpureg                   P6,FPU
       FCOMIP           fpu0,fpureg              P6,FPU
       FCOMIP                                    P6,FPU,ND
       FCOMP            mem32                    8086,FPU
       FCOMP            mem64                    8086,FPU
       FCOMP            fpureg                   8086,FPU
       FCOMP            fpu0,fpureg              8086,FPU
       FCOMP                                     8086,FPU,ND
       FCOMPP                                    8086,FPU
       FCOS                                      386,FPU
       FDECSTP                                   8086,FPU
       FDISI                                     8086,FPU
       FDIV             mem32                    8086,FPU
       FDIV             mem64                    8086,FPU
       FDIV             fpureg|to                8086,FPU
       FDIV             fpureg                   8086,FPU
       FDIV             fpureg,fpu0              8086,FPU
       FDIV             fpu0,fpureg              8086,FPU
       FDIV                                      8086,FPU,ND
       FDIVP            fpureg                   8086,FPU
       FDIVP            fpureg,fpu0              8086,FPU
       FDIVP                                     8086,FPU,ND
       FDIVR            mem32                    8086,FPU
       FDIVR            mem64                    8086,FPU
       FDIVR            fpureg|to                8086,FPU
       FDIVR            fpureg,fpu0              8086,FPU
       FDIVR            fpureg                   8086,FPU
       FDIVR            fpu0,fpureg              8086,FPU
       FDIVR                                     8086,FPU,ND
       FDIVRP           fpureg                   8086,FPU
       FDIVRP           fpureg,fpu0              8086,FPU
       FDIVRP                                    8086,FPU,ND
       FEMMS                                     PENT,3DNOW
       FENI                                      8086,FPU
       FFREE            fpureg                   8086,FPU
       FFREE                                     8086,FPU
       FFREEP           fpureg                   286,FPU,UNDOC
       FFREEP                                    286,FPU,UNDOC
       FIADD            mem32                    8086,FPU
       FIADD            mem16                    8086,FPU
       FICOM            mem32                    8086,FPU
       FICOM            mem16                    8086,FPU
       FICOMP           mem32                    8086,FPU
       FICOMP           mem16                    8086,FPU
       FIDIV            mem32                    8086,FPU
       FIDIV            mem16                    8086,FPU
       FIDIVR           mem32                    8086,FPU
       FIDIVR           mem16                    8086,FPU
       FILD             mem32                    8086,FPU
       FILD             mem16                    8086,FPU
       FILD             mem64                    8086,FPU
       FIMUL            mem32                    8086,FPU
       FIMUL            mem16                    8086,FPU
       FINCSTP                                   8086,FPU
       FINIT                                     8086,FPU
       FIST             mem32                    8086,FPU
       FIST             mem16                    8086,FPU
       FISTP            mem32                    8086,FPU
       FISTP            mem16                    8086,FPU
       FISTP            mem64                    8086,FPU
       FISTTP           mem16                    PRESCOTT,FPU
       FISTTP           mem32                    PRESCOTT,FPU
       FISTTP           mem64                    PRESCOTT,FPU
       FISUB            mem32                    8086,FPU
       FISUB            mem16                    8086,FPU
       FISUBR           mem32                    8086,FPU
       FISUBR           mem16                    8086,FPU
       FLD              mem32                    8086,FPU
       FLD              mem64                    8086,FPU
       FLD              mem80                    8086,FPU
       FLD              fpureg                   8086,FPU
       FLD                                       8086,FPU,ND
       FLD1                                      8086,FPU
       FLDCW            mem                      8086,FPU,SW
       FLDENV           mem                      8086,FPU
       FLDL2E                                    8086,FPU
       FLDL2T                                    8086,FPU
       FLDLG2                                    8086,FPU
       FLDLN2                                    8086,FPU
       FLDPI                                     8086,FPU
       FLDZ                                      8086,FPU
       FMUL             mem32                    8086,FPU
       FMUL             mem64                    8086,FPU
       FMUL             fpureg|to                8086,FPU
       FMUL             fpureg,fpu0              8086,FPU
       FMUL             fpureg                   8086,FPU
       FMUL             fpu0,fpureg              8086,FPU
       FMUL                                      8086,FPU,ND
       FMULP            fpureg                   8086,FPU
       FMULP            fpureg,fpu0              8086,FPU
       FMULP                                     8086,FPU,ND
       FNCLEX                                    8086,FPU
       FNDISI                                    8086,FPU
       FNENI                                     8086,FPU
       FNINIT                                    8086,FPU
       FNOP                                      8086,FPU
       FNSAVE           mem                      8086,FPU
       FNSTCW           mem                      8086,FPU,SW
       FNSTENV          mem                      8086,FPU
       FNSTSW           mem                      8086,FPU,SW
       FNSTSW           reg_ax                   286,FPU
       FPATAN                                    8086,FPU
       FPREM                                     8086,FPU
       FPREM1                                    386,FPU
       FPTAN                                     8086,FPU
       FRNDINT                                   8086,FPU
       FRSTOR           mem                      8086,FPU
       FSAVE            mem                      8086,FPU
       FSCALE                                    8086,FPU
       FSETPM                                    286,FPU
       FSIN                                      386,FPU
       FSINCOS                                   386,FPU
       FSQRT                                     8086,FPU
       FST              mem32                    8086,FPU
       FST              mem64                    8086,FPU
       FST              fpureg                   8086,FPU
       FST                                       8086,FPU,ND
       FSTCW            mem                      8086,FPU,SW
       FSTENV           mem                      8086,FPU
       FSTP             mem32                    8086,FPU
       FSTP             mem64                    8086,FPU
       FSTP             mem80                    8086,FPU
       FSTP             fpureg                   8086,FPU
       FSTP                                      8086,FPU,ND
       FSTSW            mem                      8086,FPU,SW
       FSTSW            reg_ax                   286,FPU
       FSUB             mem32                    8086,FPU
       FSUB             mem64                    8086,FPU
       FSUB             fpureg|to                8086,FPU
       FSUB             fpureg,fpu0              8086,FPU
       FSUB             fpureg                   8086,FPU
       FSUB             fpu0,fpureg              8086,FPU
       FSUB                                      8086,FPU,ND
       FSUBP            fpureg                   8086,FPU
       FSUBP            fpureg,fpu0              8086,FPU
       FSUBP                                     8086,FPU,ND
       FSUBR            mem32                    8086,FPU
       FSUBR            mem64                    8086,FPU
       FSUBR            fpureg|to                8086,FPU
       FSUBR            fpureg,fpu0              8086,FPU
       FSUBR            fpureg                   8086,FPU
       FSUBR            fpu0,fpureg              8086,FPU
       FSUBR                                     8086,FPU,ND
       FSUBRP           fpureg                   8086,FPU
       FSUBRP           fpureg,fpu0              8086,FPU
       FSUBRP                                    8086,FPU,ND
       FTST                                      8086,FPU
       FUCOM            fpureg                   386,FPU
       FUCOM            fpu0,fpureg              386,FPU
       FUCOM                                     386,FPU,ND
       FUCOMI           fpureg                   P6,FPU
       FUCOMI           fpu0,fpureg              P6,FPU
       FUCOMI                                    P6,FPU,ND
       FUCOMIP          fpureg                   P6,FPU
       FUCOMIP          fpu0,fpureg              P6,FPU
       FUCOMIP                                   P6,FPU,ND
       FUCOMP           fpureg                   386,FPU
       FUCOMP           fpu0,fpureg              386,FPU
       FUCOMP                                    386,FPU,ND
       FUCOMPP                                   386,FPU
       FXAM                                      8086,FPU
       FXCH             fpureg                   8086,FPU
       FXCH             fpureg,fpu0              8086,FPU
       FXCH             fpu0,fpureg              8086,FPU
       FXCH                                      8086,FPU,ND
       FXTRACT                                   8086,FPU
       FYL2X                                     8086,FPU
       FYL2XP1                                   8086,FPU
       HLT                                       8086,PRIV
       IBTS             mem,reg16                386,SW,UNDOC,ND
       IBTS             reg16,reg16              386,UNDOC,ND
       IBTS             mem,reg32                386,SD,UNDOC,ND
       IBTS             reg32,reg32              386,UNDOC,ND
       ICEBP                                     386,ND
       IDIV             rm8                      8086
       IDIV             rm16                     8086
       IDIV             rm32                     386
       IDIV             rm64                     X64
       IMUL             rm8                      8086
       IMUL             rm16                     8086
       IMUL             rm32                     386
       IMUL             rm64                     X64
       IMUL             reg16,mem                386
       IMUL             reg16,reg16              386
       IMUL             reg32,mem                386
       IMUL             reg32,reg32              386
       IMUL             reg64,mem                X64
       IMUL             reg64,reg64              X64
       IMUL             reg16,mem,imm8           186
       IMUL             reg16,mem,sbyteword      186,ND
       IMUL             reg16,mem,imm16          186
       IMUL             reg16,mem,imm            186,ND
       IMUL             reg16,reg16,imm8         186
       IMUL             reg16,reg16,sbyteword    186,ND
       IMUL             reg16,reg16,imm16        186
       IMUL             reg16,reg16,imm          186,ND
       IMUL             reg32,mem,imm8           386
       IMUL             reg32,mem,sbytedword     386,ND
       IMUL             reg32,mem,imm32          386
       IMUL             reg32,mem,imm            386,ND
       IMUL             reg32,reg32,imm8         386
       IMUL             reg32,reg32,sbytedword   386,ND
       IMUL             reg32,reg32,imm32        386
       IMUL             reg32,reg32,imm          386,ND
       IMUL             reg64,mem,imm8           X64
       IMUL             reg64,mem,sbytedword     X64,ND
       IMUL             reg64,mem,imm32          X64
       IMUL             reg64,mem,imm            X64,ND
       IMUL             reg64,reg64,imm8         X64
       IMUL             reg64,reg64,sbytedword   X64,ND
       IMUL             reg64,reg64,imm32        X64
       IMUL             reg64,reg64,imm          X64,ND
       IMUL             reg16,imm8               186
       IMUL             reg16,sbyteword          186,ND
       IMUL             reg16,imm16              186
       IMUL             reg16,imm                186,ND
       IMUL             reg32,imm8               386
       IMUL             reg32,sbytedword         386,ND
       IMUL             reg32,imm32              386
       IMUL             reg32,imm                386,ND
       IMUL             reg64,imm8               X64
       IMUL             reg64,sbytedword         X64,ND
       IMUL             reg64,imm32              X64
       IMUL             reg64,imm                X64,ND
       IN               reg_al,imm               8086
       IN               reg_ax,imm               8086
       IN               reg_eax,imm              386
       IN               reg_al,reg_dx            8086
       IN               reg_ax,reg_dx            8086
       IN               reg_eax,reg_dx           386
       INC              reg16                    8086,NOLONG
       INC              reg32                    386,NOLONG
       INC              rm8                      8086,LOCK
       INC              rm16                     8086,LOCK
       INC              rm32                     386,LOCK
       INC              rm64                     X64,LOCK
       INCBIN                                    
       INSB                                      186
       INSD                                      386
       INSW                                      186
       INT              imm                      8086
       INT01                                     386,ND
       INT1                                      386
       INT03                                     8086,ND
       INT3                                      8086
       INTO                                      8086,NOLONG
       INVD                                      486,PRIV
       INVPCID          reg32,mem128             FUTURE,INVPCID,PRIV,NOLONG
       INVPCID          reg64,mem128             FUTURE,INVPCID,PRIV,LONG
       INVLPG           mem                      486,PRIV
       INVLPGA          reg_ax,reg_ecx           X86_64,AMD,NOLONG
       INVLPGA          reg_eax,reg_ecx          X86_64,AMD
       INVLPGA          reg_rax,reg_ecx          X64,AMD
       INVLPGA                                   X86_64,AMD
       IRET                                      8086
       IRETD                                     386
       IRETQ                                     X64
       IRETW                                     8086
       JCXZ             imm                      8086,NOLONG
       JECXZ            imm                      386
       JRCXZ            imm                      X64
       JMP              imm|short                8086
       JMP              imm                      8086,ND
       JMP              imm                      8086,BND
       JMP              imm|near                 8086,ND,BND
       JMP              imm|far                  8086,ND,NOLONG
       JMP              imm16                    8086,NOLONG,BND
       JMP              imm16|near               8086,ND,NOLONG,BND
       JMP              imm16|far                8086,ND,NOLONG
       JMP              imm32                    386,NOLONG,BND
       JMP              imm32|near               386,ND,NOLONG,BND
       JMP              imm32|far                386,ND,NOLONG
       JMP              imm64                    X64,BND
       JMP              imm64|near               X64,ND,BND
       JMP              imm:imm                  8086,NOLONG
       JMP              imm16:imm                8086,NOLONG
       JMP              imm:imm16                8086,NOLONG
       JMP              imm32:imm                386,NOLONG
       JMP              imm:imm32                386,NOLONG
       JMP              mem|far                  8086,NOLONG
       JMP              mem|far                  X64
       JMP              mem16|far                8086
       JMP              mem32|far                386
       JMP              mem64|far                X64
       JMP              mem|near                 8086,ND,BND
       JMP              rm16|near                8086,NOLONG,ND,BND
       JMP              rm32|near                386,NOLONG,ND,BND
       JMP              rm64|near                X64,ND,BND
       JMP              mem                      8086,BND
       JMP              rm16                     8086,NOLONG,BND
       JMP              rm32                     386,NOLONG,BND
       JMP              rm64                     X64,BND
       JMPE             imm                      IA64
       JMPE             imm16                    IA64
       JMPE             imm32                    IA64
       JMPE             rm16                     IA64
       JMPE             rm32                     IA64
       LAHF                                      8086
       LAR              reg16,mem                286,PROT,SW
       LAR              reg16,reg16              286,PROT
       LAR              reg16,reg32              386,PROT
       LAR              reg16,reg64              X64,PROT,ND
       LAR              reg32,mem                386,PROT,SW
       LAR              reg32,reg16              386,PROT
       LAR              reg32,reg32              386,PROT
       LAR              reg32,reg64              X64,PROT,ND
       LAR              reg64,mem                X64,PROT,SW
       LAR              reg64,reg16              X64,PROT
       LAR              reg64,reg32              X64,PROT
       LAR              reg64,reg64              X64,PROT
       LDS              reg16,mem                8086,NOLONG
       LDS              reg32,mem                386,NOLONG
       LEA              reg16,mem                8086
       LEA              reg32,mem                386
       LEA              reg64,mem                X64
       LEAVE                                     186
       LES              reg16,mem                8086,NOLONG
       LES              reg32,mem                386,NOLONG
       LFENCE                                    X64,AMD
       LFS              reg16,mem                386
       LFS              reg32,mem                386
       LFS              reg64,mem                X64
       LGDT             mem                      286,PRIV
       LGS              reg16,mem                386
       LGS              reg32,mem                386
       LGS              reg64,mem                X64
       LIDT             mem                      286,PRIV
       LLDT             mem                      286,PROT,PRIV
       LLDT             mem16                    286,PROT,PRIV
       LLDT             reg16                    286,PROT,PRIV
       LMSW             mem                      286,PRIV
       LMSW             mem16                    286,PRIV
       LMSW             reg16                    286,PRIV
       LOADALL                                   386,UNDOC,ND
       LOADALL286                                286,UNDOC,ND
       LODSB                                     8086
       LODSD                                     386
       LODSQ                                     X64
       LODSW                                     8086
       LOOP             imm                      8086
       LOOP             imm,reg_cx               8086,NOLONG
       LOOP             imm,reg_ecx              386
       LOOP             imm,reg_rcx              X64
       LOOPE            imm                      8086
       LOOPE            imm,reg_cx               8086,NOLONG
       LOOPE            imm,reg_ecx              386
       LOOPE            imm,reg_rcx              X64
       LOOPNE           imm                      8086
       LOOPNE           imm,reg_cx               8086,NOLONG
       LOOPNE           imm,reg_ecx              386
       LOOPNE           imm,reg_rcx              X64
       LOOPNZ           imm                      8086
       LOOPNZ           imm,reg_cx               8086,NOLONG
       LOOPNZ           imm,reg_ecx              386
       LOOPNZ           imm,reg_rcx              X64
       LOOPZ            imm                      8086
       LOOPZ            imm,reg_cx               8086,NOLONG
       LOOPZ            imm,reg_ecx              386
       LOOPZ            imm,reg_rcx              X64
       LSL              reg16,mem                286,PROT,SW
       LSL              reg16,reg16              286,PROT
       LSL              reg16,reg32              386,PROT
       LSL              reg16,reg64              X64,PROT,ND
       LSL              reg32,mem                386,PROT,SW
       LSL              reg32,reg16              386,PROT
       LSL              reg32,reg32              386,PROT
       LSL              reg32,reg64              X64,PROT,ND
       LSL              reg64,mem                X64,PROT,SW
       LSL              reg64,reg16              X64,PROT
       LSL              reg64,reg32              X64,PROT
       LSL              reg64,reg64              X64,PROT
       LSS              reg16,mem                386
       LSS              reg32,mem                386
       LSS              reg64,mem                X64
       LTR              mem                      286,PROT,PRIV
       LTR              mem16                    286,PROT,PRIV
       LTR              reg16                    286,PROT,PRIV
       MFENCE                                    X64,AMD
       MONITOR                                   PRESCOTT
       MONITOR          reg_eax,reg_ecx,reg_edx  PRESCOTT,NOLONG,ND
       MONITOR          reg_rax,reg_ecx,reg_edx  X64,ND
       MOV              mem,reg_sreg             8086,SW
       MOV              reg16,reg_sreg           8086
       MOV              reg32,reg_sreg           386
       MOV              reg64,reg_sreg           X64,OPT,ND
       MOV              rm64,reg_sreg            X64
       MOV              reg_sreg,mem             8086,SW
       MOV              reg_sreg,reg16           8086,OPT,ND
       MOV              reg_sreg,reg32           386,OPT,ND
       MOV              reg_sreg,reg64           X64,OPT,ND
       MOV              reg_sreg,reg16           8086
       MOV              reg_sreg,reg32           386
       MOV              reg_sreg,rm64            X64
       MOV              reg_al,mem_offs          8086
       MOV              reg_ax,mem_offs          8086
       MOV              reg_eax,mem_offs         386
       MOV              reg_rax,mem_offs         X64
       MOV              mem_offs,reg_al          8086,NOHLE
       MOV              mem_offs,reg_ax          8086,NOHLE
       MOV              mem_offs,reg_eax         386,NOHLE
       MOV              mem_offs,reg_rax         X64,NOHLE
       MOV              reg32,reg_creg           386,PRIV,NOLONG
       MOV              reg64,reg_creg           X64,PRIV
       MOV              reg_creg,reg32           386,PRIV,NOLONG
       MOV              reg_creg,reg64           X64,PRIV
       MOV              reg32,reg_dreg           386,PRIV,NOLONG
       MOV              reg64,reg_dreg           X64,PRIV
       MOV              reg_dreg,reg32           386,PRIV,NOLONG
       MOV              reg_dreg,reg64           X64,PRIV
       MOV              reg32,reg_treg           386,NOLONG,ND
       MOV              reg_treg,reg32           386,NOLONG,ND
       MOV              mem,reg8                 8086
       MOV              reg8,reg8                8086
       MOV              mem,reg16                8086
       MOV              reg16,reg16              8086
       MOV              mem,reg32                386
       MOV              reg32,reg32              386
       MOV              mem,reg64                X64
       MOV              reg64,reg64              X64
       MOV              reg8,mem                 8086
       MOV              reg8,reg8                8086
       MOV              reg16,mem                8086
       MOV              reg16,reg16              8086
       MOV              reg32,mem                386
       MOV              reg32,reg32              386
       MOV              reg64,mem                X64
       MOV              reg64,reg64              X64
       MOV              reg8,imm                 8086
       MOV              reg16,imm                8086
       MOV              reg32,imm                386
       MOV              reg64,udword             X64,OPT,ND
       MOV              reg64,sdword             X64,OPT,ND
       MOV              reg64,imm                X64
       MOV              rm8,imm                  8086
       MOV              rm16,imm                 8086
       MOV              rm32,imm                 386
       MOV              rm64,imm                 X64
       MOV              rm64,imm32               X64
       MOV              mem,imm8                 8086
       MOV              mem,imm16                8086
       MOV              mem,imm32                386
       MOVD             mmxreg,rm32              PENT,MMX,SD
       MOVD             rm32,mmxreg              PENT,MMX,SD
       MOVD             mmxreg,rm64              X64,MMX,SX,ND
       MOVD             rm64,mmxreg              X64,MMX,SX,ND
       MOVQ             mmxreg,mmxrm             PENT,MMX
       MOVQ             mmxrm,mmxreg             PENT,MMX
       MOVQ             mmxreg,rm64              X64,MMX
       MOVQ             rm64,mmxreg              X64,MMX
       MOVSB                                     8086
       MOVSD                                     386
       MOVSQ                                     X64
       MOVSW                                     8086
       MOVSX            reg16,mem                386
       MOVSX            reg16,reg8               386
       MOVSX            reg32,rm8                386
       MOVSX            reg32,rm16               386
       MOVSX            reg64,rm8                X64
       MOVSX            reg64,rm16               X64
       MOVSXD           reg64,rm32               X64
       MOVSX            reg64,rm32               X64,ND
       MOVZX            reg16,mem                386
       MOVZX            reg16,reg8               386
       MOVZX            reg32,rm8                386
       MOVZX            reg32,rm16               386
       MOVZX            reg64,rm8                X64
       MOVZX            reg64,rm16               X64
       MUL              rm8                      8086
       MUL              rm16                     8086
       MUL              rm32                     386
       MUL              rm64                     X64
       MWAIT                                     PRESCOTT
       MWAIT            reg_eax,reg_ecx          PRESCOTT,ND
       NEG              rm8                      8086,LOCK
       NEG              rm16                     8086,LOCK
       NEG              rm32                     386,LOCK
       NEG              rm64                     X64,LOCK
       NOP                                       8086
       NOP              rm16                     P6
       NOP              rm32                     P6
       NOP              rm64                     X64
       NOT              rm8                      8086,LOCK
       NOT              rm16                     8086,LOCK
       NOT              rm32                     386,LOCK
       NOT              rm64                     X64,LOCK
       OR               mem,reg8                 8086,LOCK
       OR               reg8,reg8                8086
       OR               mem,reg16                8086,LOCK
       OR               reg16,reg16              8086
       OR               mem,reg32                386,LOCK
       OR               reg32,reg32              386
       OR               mem,reg64                X64,LOCK
       OR               reg64,reg64              X64
       OR               reg8,mem                 8086
       OR               reg8,reg8                8086
       OR               reg16,mem                8086
       OR               reg16,reg16              8086
       OR               reg32,mem                386
       OR               reg32,reg32              386
       OR               reg64,mem                X64
       OR               reg64,reg64              X64
       OR               rm16,imm8                8086,LOCK
       OR               rm32,imm8                386,LOCK
       OR               rm64,imm8                X64,LOCK
       OR               reg_al,imm               8086
       OR               reg_ax,sbyteword         8086,ND
       OR               reg_ax,imm               8086
       OR               reg_eax,sbytedword       386,ND
       OR               reg_eax,imm              386
       OR               reg_rax,sbytedword       X64,ND
       OR               reg_rax,imm              X64
       OR               rm8,imm                  8086,LOCK
       OR               rm16,sbyteword           8086,LOCK,ND
       OR               rm16,imm                 8086,LOCK
       OR               rm32,sbytedword          386,LOCK,ND
       OR               rm32,imm                 386,LOCK
       OR               rm64,sbytedword          X64,LOCK,ND
       OR               rm64,imm                 X64,LOCK
       OR               mem,imm8                 8086,LOCK
       OR               mem,sbyteword16          8086,LOCK,ND
       OR               mem,imm16                8086,LOCK
       OR               mem,sbytedword32         386,LOCK,ND
       OR               mem,imm32                386,LOCK
       OR               rm8,imm                  8086,LOCK,ND,NOLONG
       OUT              imm,reg_al               8086
       OUT              imm,reg_ax               8086
       OUT              imm,reg_eax              386
       OUT              reg_dx,reg_al            8086
       OUT              reg_dx,reg_ax            8086
       OUT              reg_dx,reg_eax           386
       OUTSB                                     186
       OUTSD                                     386
       OUTSW                                     186
       PACKSSDW         mmxreg,mmxrm             PENT,MMX
       PACKSSWB         mmxreg,mmxrm             PENT,MMX
       PACKUSWB         mmxreg,mmxrm             PENT,MMX
       PADDB            mmxreg,mmxrm             PENT,MMX
       PADDD            mmxreg,mmxrm             PENT,MMX
       PADDSB           mmxreg,mmxrm             PENT,MMX
       PADDSIW          mmxreg,mmxrm             PENT,MMX,CYRIX
       PADDSW           mmxreg,mmxrm             PENT,MMX
       PADDUSB          mmxreg,mmxrm             PENT,MMX
       PADDUSW          mmxreg,mmxrm             PENT,MMX
       PADDW            mmxreg,mmxrm             PENT,MMX
       PAND             mmxreg,mmxrm             PENT,MMX
       PANDN            mmxreg,mmxrm             PENT,MMX
       PAUSE                                     8086
       PAVEB            mmxreg,mmxrm             PENT,MMX,CYRIX
       PAVGUSB          mmxreg,mmxrm             PENT,3DNOW
       PCMPEQB          mmxreg,mmxrm             PENT,MMX
       PCMPEQD          mmxreg,mmxrm             PENT,MMX
       PCMPEQW          mmxreg,mmxrm             PENT,MMX
       PCMPGTB          mmxreg,mmxrm             PENT,MMX
       PCMPGTD          mmxreg,mmxrm             PENT,MMX
       PCMPGTW          mmxreg,mmxrm             PENT,MMX
       PDISTIB          mmxreg,mem               PENT,MMX,CYRIX
       PF2ID            mmxreg,mmxrm             PENT,3DNOW
       PFACC            mmxreg,mmxrm             PENT,3DNOW
       PFADD            mmxreg,mmxrm             PENT,3DNOW
       PFCMPEQ          mmxreg,mmxrm             PENT,3DNOW
       PFCMPGE          mmxreg,mmxrm             PENT,3DNOW
       PFCMPGT          mmxreg,mmxrm             PENT,3DNOW
       PFMAX            mmxreg,mmxrm             PENT,3DNOW
       PFMIN            mmxreg,mmxrm             PENT,3DNOW
       PFMUL            mmxreg,mmxrm             PENT,3DNOW
       PFRCP            mmxreg,mmxrm             PENT,3DNOW
       PFRCPIT1         mmxreg,mmxrm             PENT,3DNOW
       PFRCPIT2         mmxreg,mmxrm             PENT,3DNOW
       PFRSQIT1         mmxreg,mmxrm             PENT,3DNOW
       PFRSQRT          mmxreg,mmxrm             PENT,3DNOW
       PFSUB            mmxreg,mmxrm             PENT,3DNOW
       PFSUBR           mmxreg,mmxrm             PENT,3DNOW
       PI2FD            mmxreg,mmxrm             PENT,3DNOW
       PMACHRIW         mmxreg,mem               PENT,MMX,CYRIX
       PMADDWD          mmxreg,mmxrm             PENT,MMX
       PMAGW            mmxreg,mmxrm             PENT,MMX,CYRIX
       PMULHRIW         mmxreg,mmxrm             PENT,MMX,CYRIX
       PMULHRWA         mmxreg,mmxrm             PENT,3DNOW
       PMULHRWC         mmxreg,mmxrm             PENT,MMX,CYRIX
       PMULHW           mmxreg,mmxrm             PENT,MMX
       PMULLW           mmxreg,mmxrm             PENT,MMX
       PMVGEZB          mmxreg,mem               PENT,MMX,CYRIX
       PMVLZB           mmxreg,mem               PENT,MMX,CYRIX
       PMVNZB           mmxreg,mem               PENT,MMX,CYRIX
       PMVZB            mmxreg,mem               PENT,MMX,CYRIX
       POP              reg16                    8086
       POP              reg32                    386,NOLONG
       POP              reg64                    X64
       POP              rm16                     8086
       POP              rm32                     386,NOLONG
       POP              rm64                     X64
       POP              reg_es                   8086,NOLONG
       POP              reg_cs                   8086,UNDOC,ND
       POP              reg_ss                   8086,NOLONG
       POP              reg_ds                   8086,NOLONG
       POP              reg_fs                   386
       POP              reg_gs                   386
       POPA                                      186,NOLONG
       POPAD                                     386,NOLONG
       POPAW                                     186,NOLONG
       POPF                                      8086
       POPFD                                     386,NOLONG
       POPFQ                                     X64
       POPFW                                     8086
       POR              mmxreg,mmxrm             PENT,MMX
       PREFETCH         mem                      PENT,3DNOW
       PREFETCHW        mem                      PENT,3DNOW
       PSLLD            mmxreg,mmxrm             PENT,MMX
       PSLLD            mmxreg,imm               PENT,MMX
       PSLLQ            mmxreg,mmxrm             PENT,MMX
       PSLLQ            mmxreg,imm               PENT,MMX
       PSLLW            mmxreg,mmxrm             PENT,MMX
       PSLLW            mmxreg,imm               PENT,MMX
       PSRAD            mmxreg,mmxrm             PENT,MMX
       PSRAD            mmxreg,imm               PENT,MMX
       PSRAW            mmxreg,mmxrm             PENT,MMX
       PSRAW            mmxreg,imm               PENT,MMX
       PSRLD            mmxreg,mmxrm             PENT,MMX
       PSRLD            mmxreg,imm               PENT,MMX
       PSRLQ            mmxreg,mmxrm             PENT,MMX
       PSRLQ            mmxreg,imm               PENT,MMX
       PSRLW            mmxreg,mmxrm             PENT,MMX
       PSRLW            mmxreg,imm               PENT,MMX
       PSUBB            mmxreg,mmxrm             PENT,MMX
       PSUBD            mmxreg,mmxrm             PENT,MMX
       PSUBSB           mmxreg,mmxrm             PENT,MMX
       PSUBSIW          mmxreg,mmxrm             PENT,MMX,CYRIX
       PSUBSW           mmxreg,mmxrm             PENT,MMX
       PSUBUSB          mmxreg,mmxrm             PENT,MMX
       PSUBUSW          mmxreg,mmxrm             PENT,MMX
       PSUBW            mmxreg,mmxrm             PENT,MMX
       PUNPCKHBW        mmxreg,mmxrm             PENT,MMX
       PUNPCKHDQ        mmxreg,mmxrm             PENT,MMX
       PUNPCKHWD        mmxreg,mmxrm             PENT,MMX
       PUNPCKLBW        mmxreg,mmxrm             PENT,MMX
       PUNPCKLDQ        mmxreg,mmxrm             PENT,MMX
       PUNPCKLWD        mmxreg,mmxrm             PENT,MMX
       PUSH             reg16                    8086
       PUSH             reg32                    386,NOLONG
       PUSH             reg64                    X64
       PUSH             rm16                     8086
       PUSH             rm32                     386,NOLONG
       PUSH             rm64                     X64
       PUSH             reg_es                   8086,NOLONG
       PUSH             reg_cs                   8086,NOLONG
       PUSH             reg_ss                   8086,NOLONG
       PUSH             reg_ds                   8086,NOLONG
       PUSH             reg_fs                   386
       PUSH             reg_gs                   386
       PUSH             imm8                     186
       PUSH             sbyteword16              186,AR0,SIZE,ND
       PUSH             imm16                    186,AR0,SIZE
       PUSH             sbytedword32             386,NOLONG,AR0,SIZE,ND
       PUSH             imm32                    386,NOLONG,AR0,SIZE
       PUSH             sbytedword32             386,NOLONG,SD,ND
       PUSH             imm32                    386,NOLONG,SD
       PUSH             sbytedword64             X64,AR0,SIZE,ND
       PUSH             imm64                    X64,AR0,SIZE
       PUSH             sbytedword32             X64,AR0,SIZE,ND
       PUSH             imm32                    X64,AR0,SIZE
       PUSHA                                     186,NOLONG
       PUSHAD                                    386,NOLONG
       PUSHAW                                    186,NOLONG
       PUSHF                                     8086
       PUSHFD                                    386,NOLONG
       PUSHFQ                                    X64
       PUSHFW                                    8086
       PXOR             mmxreg,mmxrm             PENT,MMX
       RCL              rm8,unity                8086
       RCL              rm8,reg_cl               8086
       RCL              rm8,imm8                 186
       RCL              rm16,unity               8086
       RCL              rm16,reg_cl              8086
       RCL              rm16,imm8                186
       RCL              rm32,unity               386
       RCL              rm32,reg_cl              386
       RCL              rm32,imm8                386
       RCL              rm64,unity               X64
       RCL              rm64,reg_cl              X64
       RCL              rm64,imm8                X64
       RCR              rm8,unity                8086
       RCR              rm8,reg_cl               8086
       RCR              rm8,imm8                 186
       RCR              rm16,unity               8086
       RCR              rm16,reg_cl              8086
       RCR              rm16,imm8                186
       RCR              rm32,unity               386
       RCR              rm32,reg_cl              386
       RCR              rm32,imm8                386
       RCR              rm64,unity               X64
       RCR              rm64,reg_cl              X64
       RCR              rm64,imm8                X64
       RDSHR            rm32                     P6,CYRIXM
       RDMSR                                     PENT,PRIV
       RDPMC                                     P6
       RDTSC                                     PENT
       RDTSCP                                    X86_64
       RET                                       8086,BND
       RET              imm                      8086,SW,BND
       RETF                                      8086
       RETF             imm                      8086,SW
       RETN                                      8086,BND
       RETN             imm                      8086,SW,BND
       ROL              rm8,unity                8086
       ROL              rm8,reg_cl               8086
       ROL              rm8,imm8                 186
       ROL              rm16,unity               8086
       ROL              rm16,reg_cl              8086
       ROL              rm16,imm8                186
       ROL              rm32,unity               386
       ROL              rm32,reg_cl              386
       ROL              rm32,imm8                386
       ROL              rm64,unity               X64
       ROL              rm64,reg_cl              X64
       ROL              rm64,imm8                X64
       ROR              rm8,unity                8086
       ROR              rm8,reg_cl               8086
       ROR              rm8,imm8                 186
       ROR              rm16,unity               8086
       ROR              rm16,reg_cl              8086
       ROR              rm16,imm8                186
       ROR              rm32,unity               386
       ROR              rm32,reg_cl              386
       ROR              rm32,imm8                386
       ROR              rm64,unity               X64
       ROR              rm64,reg_cl              X64
       ROR              rm64,imm8                X64
       RDM                                       P6,CYRIX,ND
       RSDC             reg_sreg,mem80           486,CYRIXM
       RSLDT            mem80                    486,CYRIXM
       RSM                                       PENTM
       RSTS             mem80                    486,CYRIXM
       SAHF                                      8086
       SAL              rm8,unity                8086,ND
       SAL              rm8,reg_cl               8086,ND
       SAL              rm8,imm8                 186,ND
       SAL              rm16,unity               8086,ND
       SAL              rm16,reg_cl              8086,ND
       SAL              rm16,imm8                186,ND
       SAL              rm32,unity               386,ND
       SAL              rm32,reg_cl              386,ND
       SAL              rm32,imm8                386,ND
       SAL              rm64,unity               X64,ND
       SAL              rm64,reg_cl              X64,ND
       SAL              rm64,imm8                X64,ND
       SALC                                      8086,UNDOC
       SAR              rm8,unity                8086
       SAR              rm8,reg_cl               8086
       SAR              rm8,imm8                 186
       SAR              rm16,unity               8086
       SAR              rm16,reg_cl              8086
       SAR              rm16,imm8                186
       SAR              rm32,unity               386
       SAR              rm32,reg_cl              386
       SAR              rm32,imm8                386
       SAR              rm64,unity               X64
       SAR              rm64,reg_cl              X64
       SAR              rm64,imm8                X64
       SBB              mem,reg8                 8086,LOCK
       SBB              reg8,reg8                8086
       SBB              mem,reg16                8086,LOCK
       SBB              reg16,reg16              8086
       SBB              mem,reg32                386,LOCK
       SBB              reg32,reg32              386
       SBB              mem,reg64                X64,LOCK
       SBB              reg64,reg64              X64
       SBB              reg8,mem                 8086
       SBB              reg8,reg8                8086
       SBB              reg16,mem                8086
       SBB              reg16,reg16              8086
       SBB              reg32,mem                386
       SBB              reg32,reg32              386
       SBB              reg64,mem                X64
       SBB              reg64,reg64              X64
       SBB              rm16,imm8                8086,LOCK
       SBB              rm32,imm8                386,LOCK
       SBB              rm64,imm8                X64,LOCK
       SBB              reg_al,imm               8086
       SBB              reg_ax,sbyteword         8086,ND
       SBB              reg_ax,imm               8086
       SBB              reg_eax,sbytedword       386,ND
       SBB              reg_eax,imm              386
       SBB              reg_rax,sbytedword       X64,ND
       SBB              reg_rax,imm              X64
       SBB              rm8,imm                  8086,LOCK
       SBB              rm16,sbyteword           8086,LOCK,ND
       SBB              rm16,imm                 8086,LOCK
       SBB              rm32,sbytedword          386,LOCK,ND
       SBB              rm32,imm                 386,LOCK
       SBB              rm64,sbytedword          X64,LOCK,ND
       SBB              rm64,imm                 X64,LOCK
       SBB              mem,imm8                 8086,LOCK
       SBB              mem,sbyteword16          8086,LOCK,ND
       SBB              mem,imm16                8086,LOCK
       SBB              mem,sbytedword32         386,LOCK,ND
       SBB              mem,imm32                386,LOCK
       SBB              rm8,imm                  8086,LOCK,ND,NOLONG
       SCASB                                     8086
       SCASD                                     386
       SCASQ                                     X64
       SCASW                                     8086
       SFENCE                                    X64,AMD
       SGDT             mem                      286
       SHL              rm8,unity                8086
       SHL              rm8,reg_cl               8086
       SHL              rm8,imm8                 186
       SHL              rm16,unity               8086
       SHL              rm16,reg_cl              8086
       SHL              rm16,imm8                186
       SHL              rm32,unity               386
       SHL              rm32,reg_cl              386
       SHL              rm32,imm8                386
       SHL              rm64,unity               X64
       SHL              rm64,reg_cl              X64
       SHL              rm64,imm8                X64
       SHLD             mem,reg16,imm            3862
       SHLD             reg16,reg16,imm          3862
       SHLD             mem,reg32,imm            3862
       SHLD             reg32,reg32,imm          3862
       SHLD             mem,reg64,imm            X642
       SHLD             reg64,reg64,imm          X642
       SHLD             mem,reg16,reg_cl         386
       SHLD             reg16,reg16,reg_cl       386
       SHLD             mem,reg32,reg_cl         386
       SHLD             reg32,reg32,reg_cl       386
       SHLD             mem,reg64,reg_cl         X64
       SHLD             reg64,reg64,reg_cl       X64
       SHR              rm8,unity                8086
       SHR              rm8,reg_cl               8086
       SHR              rm8,imm8                 186
       SHR              rm16,unity               8086
       SHR              rm16,reg_cl              8086
       SHR              rm16,imm8                186
       SHR              rm32,unity               386
       SHR              rm32,reg_cl              386
       SHR              rm32,imm8                386
       SHR              rm64,unity               X64
       SHR              rm64,reg_cl              X64
       SHR              rm64,imm8                X64
       SHRD             mem,reg16,imm            3862
       SHRD             reg16,reg16,imm          3862
       SHRD             mem,reg32,imm            3862
       SHRD             reg32,reg32,imm          3862
       SHRD             mem,reg64,imm            X642
       SHRD             reg64,reg64,imm          X642
       SHRD             mem,reg16,reg_cl         386
       SHRD             reg16,reg16,reg_cl       386
       SHRD             mem,reg32,reg_cl         386
       SHRD             reg32,reg32,reg_cl       386
       SHRD             mem,reg64,reg_cl         X64
       SHRD             reg64,reg64,reg_cl       X64
       SIDT             mem                      286
       SLDT             mem                      286
       SLDT             mem16                    286
       SLDT             reg16                    286
       SLDT             reg32                    386
       SLDT             reg64                    X64,ND
       SLDT             reg64                    X64
       SKINIT                                    X64
       SMI                                       386,UNDOC
       SMINT                                     P6,CYRIX,ND
       SMINTOLD                                  486,CYRIX,ND
       SMSW             mem                      286
       SMSW             mem16                    286
       SMSW             reg16                    286
       SMSW             reg32                    386
       STC                                       8086
       STD                                       8086
       STI                                       8086
       STOSB                                     8086
       STOSD                                     386
       STOSQ                                     X64
       STOSW                                     8086
       STR              mem                      286,PROT
       STR              mem16                    286,PROT
       STR              reg16                    286,PROT
       STR              reg32                    386,PROT
       STR              reg64                    X64
       SUB              mem,reg8                 8086,LOCK
       SUB              reg8,reg8                8086
       SUB              mem,reg16                8086,LOCK
       SUB              reg16,reg16              8086
       SUB              mem,reg32                386,LOCK
       SUB              reg32,reg32              386
       SUB              mem,reg64                X64,LOCK
       SUB              reg64,reg64              X64
       SUB              reg8,mem                 8086
       SUB              reg8,reg8                8086
       SUB              reg16,mem                8086
       SUB              reg16,reg16              8086
       SUB              reg32,mem                386
       SUB              reg32,reg32              386
       SUB              reg64,mem                X64
       SUB              reg64,reg64              X64
       SUB              rm16,imm8                8086,LOCK
       SUB              rm32,imm8                386,LOCK
       SUB              rm64,imm8                X64,LOCK
       SUB              reg_al,imm               8086
       SUB              reg_ax,sbyteword         8086,ND
       SUB              reg_ax,imm               8086
       SUB              reg_eax,sbytedword       386,ND
       SUB              reg_eax,imm              386
       SUB              reg_rax,sbytedword       X64,ND
       SUB              reg_rax,imm              X64
       SUB              rm8,imm                  8086,LOCK
       SUB              rm16,sbyteword           8086,LOCK,ND
       SUB              rm16,imm                 8086,LOCK
       SUB              rm32,sbytedword          386,LOCK,ND
       SUB              rm32,imm                 386,LOCK
       SUB              rm64,sbytedword          X64,LOCK,ND
       SUB              rm64,imm                 X64,LOCK
       SUB              mem,imm8                 8086,LOCK
       SUB              mem,sbyteword16          8086,LOCK,ND
       SUB              mem,imm16                8086,LOCK
       SUB              mem,sbytedword32         386,LOCK,ND
       SUB              mem,imm32                386,LOCK
       SUB              rm8,imm                  8086,LOCK,ND,NOLONG
       SVDC             mem80,reg_sreg           486,CYRIXM
       SVLDT            mem80                    486,CYRIXM,ND
       SVTS             mem80                    486,CYRIXM
       SWAPGS                                    X64
       SYSCALL                                   P6,AMD
       SYSENTER                                  P6
       SYSEXIT                                   P6,PRIV
       SYSRET                                    P6,PRIV,AMD
       TEST             mem,reg8                 8086
       TEST             reg8,reg8                8086
       TEST             mem,reg16                8086
       TEST             reg16,reg16              8086
       TEST             mem,reg32                386
       TEST             reg32,reg32              386
       TEST             mem,reg64                X64
       TEST             reg64,reg64              X64
       TEST             reg8,mem                 8086
       TEST             reg16,mem                8086
       TEST             reg32,mem                386
       TEST             reg64,mem                X64
       TEST             reg_al,imm               8086
       TEST             reg_ax,imm               8086
       TEST             reg_eax,imm              386
       TEST             reg_rax,imm              X64
       TEST             rm8,imm                  8086
       TEST             rm16,imm                 8086
       TEST             rm32,imm                 386
       TEST             rm64,imm                 X64
       TEST             mem,imm8                 8086
       TEST             mem,imm16                8086
       TEST             mem,imm32                386
       UD0                                       186,UNDOC
       UD1                                       186,UNDOC
       UD2B                                      186,UNDOC,ND
       UD2                                       186
       UD2A                                      186,ND
       UMOV             mem,reg8                 386,UNDOC,ND
       UMOV             reg8,reg8                386,UNDOC,ND
       UMOV             mem,reg16                386,UNDOC,ND
       UMOV             reg16,reg16              386,UNDOC,ND
       UMOV             mem,reg32                386,UNDOC,ND
       UMOV             reg32,reg32              386,UNDOC,ND
       UMOV             reg8,mem                 386,UNDOC,ND
       UMOV             reg8,reg8                386,UNDOC,ND
       UMOV             reg16,mem                386,UNDOC,ND
       UMOV             reg16,reg16              386,UNDOC,ND
       UMOV             reg32,mem                386,UNDOC,ND
       UMOV             reg32,reg32              386,UNDOC,ND
       VERR             mem                      286,PROT
       VERR             mem16                    286,PROT
       VERR             reg16                    286,PROT
       VERW             mem                      286,PROT
       VERW             mem16                    286,PROT
       VERW             reg16                    286,PROT
       FWAIT                                     8086
       WBINVD                                    486,PRIV
       WRSHR            rm32                     P6,CYRIXM
       WRMSR                                     PENT,PRIV
       XADD             mem,reg8                 486,LOCK
       XADD             reg8,reg8                486
       XADD             mem,reg16                486,LOCK
       XADD             reg16,reg16              486
       XADD             mem,reg32                486,LOCK
       XADD             reg32,reg32              486
       XADD             mem,reg64                X64,LOCK
       XADD             reg64,reg64              X64
       XBTS             reg16,mem                386,SW,UNDOC,ND
       XBTS             reg16,reg16              386,UNDOC,ND
       XBTS             reg32,mem                386,SD,UNDOC,ND
       XBTS             reg32,reg32              386,UNDOC,ND
       XCHG             reg_ax,reg16             8086
       XCHG             reg_eax,reg32na          386
       XCHG             reg_rax,reg64            X64
       XCHG             reg16,reg_ax             8086
       XCHG             reg32na,reg_eax          386
       XCHG             reg64,reg_rax            X64
       XCHG             reg_eax,reg_eax          386,NOLONG
       XCHG             reg8,mem                 8086,LOCK
       XCHG             reg8,reg8                8086
       XCHG             reg16,mem                8086,LOCK
       XCHG             reg16,reg16              8086
       XCHG             reg32,mem                386,LOCK
       XCHG             reg32,reg32              386
       XCHG             reg64,mem                X64,LOCK
       XCHG             reg64,reg64              X64
       XCHG             mem,reg8                 8086,LOCK
       XCHG             reg8,reg8                8086
       XCHG             mem,reg16                8086,LOCK
       XCHG             reg16,reg16              8086
       XCHG             mem,reg32                386,LOCK
       XCHG             reg32,reg32              386
       XCHG             mem,reg64                X64,LOCK
       XCHG             reg64,reg64              X64
       XLATB                                     8086
       XLAT                                      8086
       XOR              mem,reg8                 8086,LOCK
       XOR              reg8,reg8                8086
       XOR              mem,reg16                8086,LOCK
       XOR              reg16,reg16              8086
       XOR              mem,reg32                386,LOCK
       XOR              reg32,reg32              386
       XOR              mem,reg64                X64,LOCK
       XOR              reg64,reg64              X64
       XOR              reg8,mem                 8086
       XOR              reg8,reg8                8086
       XOR              reg16,mem                8086
       XOR              reg16,reg16              8086
       XOR              reg32,mem                386
       XOR              reg32,reg32              386
       XOR              reg64,mem                X64
       XOR              reg64,reg64              X64
       XOR              rm16,imm8                8086,LOCK
       XOR              rm32,imm8                386,LOCK
       XOR              rm64,imm8                X64,LOCK
       XOR              reg_al,imm               8086
       XOR              reg_ax,sbyteword         8086,ND
       XOR              reg_ax,imm               8086
       XOR              reg_eax,sbytedword       386,ND
       XOR              reg_eax,imm              386
       XOR              reg_rax,sbytedword       X64,ND
       XOR              reg_rax,imm              X64
       XOR              rm8,imm                  8086,LOCK
       XOR              rm16,sbyteword           8086,LOCK,ND
       XOR              rm16,imm                 8086,LOCK
       XOR              rm32,sbytedword          386,LOCK,ND
       XOR              rm32,imm                 386,LOCK
       XOR              rm64,sbytedword          X64,LOCK,ND
       XOR              rm64,imm                 X64,LOCK
       XOR              mem,imm8                 8086,LOCK
       XOR              mem,sbyteword16          8086,LOCK,ND
       XOR              mem,imm16                8086,LOCK
       XOR              mem,sbytedword32         386,LOCK,ND
       XOR              mem,imm32                386,LOCK
       XOR              rm8,imm                  8086,LOCK,ND,NOLONG
       CMOVcc           reg16,mem                P6
       CMOVcc           reg16,reg16              P6
       CMOVcc           reg32,mem                P6
       CMOVcc           reg32,reg32              P6
       CMOVcc           reg64,mem                X64
       CMOVcc           reg64,reg64              X64
       Jcc              imm|near                 386,BND
       Jcc              imm16|near               386,NOLONG,BND
       Jcc              imm32|near               386,NOLONG,BND
       Jcc              imm64|near               X64,BND
       Jcc              imm|short                8086,ND,BND
       Jcc              imm                      8086,ND,BND
       Jcc              imm                      386,ND,BND
       Jcc              imm                      8086,ND,BND
       Jcc              imm                      8086,BND
       SETcc            mem                      386
       SETcc            reg8                     386

B.1.3 Katmai Streaming SIMD instructions (SSE -- a.k.a. KNI, XMM, MMX2)

       ADDPS            xmmreg,xmmrm128          KATMAI,SSE
       ADDSS            xmmreg,xmmrm32           KATMAI,SSE
       ANDNPS           xmmreg,xmmrm128          KATMAI,SSE
       ANDPS            xmmreg,xmmrm128          KATMAI,SSE
       CMPEQPS          xmmreg,xmmrm128          KATMAI,SSE
       CMPEQSS          xmmreg,xmmrm32           KATMAI,SSE
       CMPLEPS          xmmreg,xmmrm128          KATMAI,SSE
       CMPLESS          xmmreg,xmmrm32           KATMAI,SSE
       CMPLTPS          xmmreg,xmmrm128          KATMAI,SSE
       CMPLTSS          xmmreg,xmmrm32           KATMAI,SSE
       CMPNEQPS         xmmreg,xmmrm128          KATMAI,SSE
       CMPNEQSS         xmmreg,xmmrm32           KATMAI,SSE
       CMPNLEPS         xmmreg,xmmrm128          KATMAI,SSE
       CMPNLESS         xmmreg,xmmrm32           KATMAI,SSE
       CMPNLTPS         xmmreg,xmmrm128          KATMAI,SSE
       CMPNLTSS         xmmreg,xmmrm32           KATMAI,SSE
       CMPORDPS         xmmreg,xmmrm128          KATMAI,SSE
       CMPORDSS         xmmreg,xmmrm32           KATMAI,SSE
       CMPUNORDPS       xmmreg,xmmrm128          KATMAI,SSE
       CMPUNORDSS       xmmreg,xmmrm32           KATMAI,SSE
       CMPPS            xmmreg,mem,imm           KATMAI,SSE
       CMPPS            xmmreg,xmmreg,imm        KATMAI,SSE
       CMPSS            xmmreg,mem,imm           KATMAI,SSE
       CMPSS            xmmreg,xmmreg,imm        KATMAI,SSE
       COMISS           xmmreg,xmmrm32           KATMAI,SSE
       CVTPI2PS         xmmreg,mmxrm64           KATMAI,SSE,MMX
       CVTPS2PI         mmxreg,xmmrm64           KATMAI,SSE,MMX
       CVTSI2SS         xmmreg,mem               KATMAI,SSE,SD,AR1,ND
       CVTSI2SS         xmmreg,rm32              KATMAI,SSE,SD,AR1
       CVTSI2SS         xmmreg,rm64              X64,SSE,AR1
       CVTSS2SI         reg32,xmmreg             KATMAI,SSE,SD,AR1
       CVTSS2SI         reg32,mem                KATMAI,SSE,SD,AR1
       CVTSS2SI         reg64,xmmreg             X64,SSE,SD,AR1
       CVTSS2SI         reg64,mem                X64,SSE,SD,AR1
       CVTTPS2PI        mmxreg,xmmrm             KATMAI,SSE,MMX
       CVTTSS2SI        reg32,xmmrm              KATMAI,SSE,SD,AR1
       CVTTSS2SI        reg64,xmmrm              X64,SSE,SD,AR1
       DIVPS            xmmreg,xmmrm128          KATMAI,SSE
       DIVSS            xmmreg,xmmrm32           KATMAI,SSE
       LDMXCSR          mem32                    KATMAI,SSE
       MAXPS            xmmreg,xmmrm128          KATMAI,SSE
       MAXSS            xmmreg,xmmrm32           KATMAI,SSE
       MINPS            xmmreg,xmmrm128          KATMAI,SSE
       MINSS            xmmreg,xmmrm32           KATMAI,SSE
       MOVAPS           xmmreg,xmmrm128          KATMAI,SSE
       MOVAPS           xmmrm128,xmmreg          KATMAI,SSE
       MOVHPS           xmmreg,mem64             KATMAI,SSE
       MOVHPS           mem64,xmmreg             KATMAI,SSE
       MOVLHPS          xmmreg,xmmreg            KATMAI,SSE
       MOVLPS           xmmreg,mem64             KATMAI,SSE
       MOVLPS           mem64,xmmreg             KATMAI,SSE
       MOVHLPS          xmmreg,xmmreg            KATMAI,SSE
       MOVMSKPS         reg32,xmmreg             KATMAI,SSE
       MOVMSKPS         reg64,xmmreg             X64,SSE
       MOVNTPS          mem128,xmmreg            KATMAI,SSE
       MOVSS            xmmreg,xmmrm32           KATMAI,SSE
       MOVSS            mem32,xmmreg             KATMAI,SSE
       MOVSS            xmmreg,xmmreg            KATMAI,SSE
       MOVUPS           xmmreg,xmmrm128          KATMAI,SSE
       MOVUPS           xmmrm128,xmmreg          KATMAI,SSE
       MULPS            xmmreg,xmmrm128          KATMAI,SSE
       MULSS            xmmreg,xmmrm32           KATMAI,SSE
       ORPS             xmmreg,xmmrm128          KATMAI,SSE
       RCPPS            xmmreg,xmmrm128          KATMAI,SSE
       RCPSS            xmmreg,xmmrm32           KATMAI,SSE
       RSQRTPS          xmmreg,xmmrm128          KATMAI,SSE
       RSQRTSS          xmmreg,xmmrm32           KATMAI,SSE
       SHUFPS           xmmreg,xmmrm128,imm8     KATMAI,SSE
       SQRTPS           xmmreg,xmmrm128          KATMAI,SSE
       SQRTSS           xmmreg,xmmrm32           KATMAI,SSE
       STMXCSR          mem32                    KATMAI,SSE
       SUBPS            xmmreg,xmmrm128          KATMAI,SSE
       SUBSS            xmmreg,xmmrm32           KATMAI,SSE
       UCOMISS          xmmreg,xmmrm32           KATMAI,SSE
       UNPCKHPS         xmmreg,xmmrm128          KATMAI,SSE
       UNPCKLPS         xmmreg,xmmrm128          KATMAI,SSE
       XORPS            xmmreg,xmmrm128          KATMAI,SSE

B.1.4 Introduced in Deschutes but necessary for SSE support

       FXRSTOR          mem                      P6,SSE,FPU
       FXRSTOR64        mem                      X64,SSE,FPU
       FXSAVE           mem                      P6,SSE,FPU
       FXSAVE64         mem                      X64,SSE,FPU

B.1.5 XSAVE group (AVX and extended state)

       XGETBV                                    NEHALEM
       XSETBV                                    NEHALEM,PRIV
       XSAVE            mem                      NEHALEM
       XSAVE64          mem                      LONG,NEHALEM
       XSAVEOPT         mem                      FUTURE
       XSAVEOPT64       mem                      LONG,FUTURE
       XRSTOR           mem                      NEHALEM
       XRSTOR64         mem                      LONG,NEHALEM

B.1.6 Generic memory operations

       PREFETCHNTA      mem8                     KATMAI
       PREFETCHT0       mem8                     KATMAI
       PREFETCHT1       mem8                     KATMAI
       PREFETCHT2       mem8                     KATMAI
       SFENCE                                    KATMAI

B.1.7 New MMX instructions introduced in Katmai

       MASKMOVQ         mmxreg,mmxreg            KATMAI,MMX
       MOVNTQ           mem,mmxreg               KATMAI,MMX
       PAVGB            mmxreg,mmxrm             KATMAI,MMX
       PAVGW            mmxreg,mmxrm             KATMAI,MMX
       PEXTRW           reg32,mmxreg,imm         KATMAI,MMX
       PINSRW           mmxreg,mem,imm           KATMAI,MMX
       PINSRW           mmxreg,rm16,imm          KATMAI,MMX
       PINSRW           mmxreg,reg32,imm         KATMAI,MMX
       PMAXSW           mmxreg,mmxrm             KATMAI,MMX
       PMAXUB           mmxreg,mmxrm             KATMAI,MMX
       PMINSW           mmxreg,mmxrm             KATMAI,MMX
       PMINUB           mmxreg,mmxrm             KATMAI,MMX
       PMOVMSKB         reg32,mmxreg             KATMAI,MMX
       PMULHUW          mmxreg,mmxrm             KATMAI,MMX
       PSADBW           mmxreg,mmxrm             KATMAI,MMX
       PSHUFW           mmxreg,mmxrm,imm         KATMAI,MMX2

B.1.8 AMD Enhanced 3DNow! (Athlon) instructions

       PF2IW            mmxreg,mmxrm             PENT,3DNOW
       PFNACC           mmxreg,mmxrm             PENT,3DNOW
       PFPNACC          mmxreg,mmxrm             PENT,3DNOW
       PI2FW            mmxreg,mmxrm             PENT,3DNOW
       PSWAPD           mmxreg,mmxrm             PENT,3DNOW

B.1.9 Willamette SSE2 Cacheability Instructions

       MASKMOVDQU       xmmreg,xmmreg            WILLAMETTE,SSE2
       CLFLUSH          mem                      WILLAMETTE,SSE2
       MOVNTDQ          mem,xmmreg               WILLAMETTE,SSE2,SO
       MOVNTI           mem,reg32                WILLAMETTE,SD
       MOVNTI           mem,reg64                X64
       MOVNTPD          mem,xmmreg               WILLAMETTE,SSE2,SO
       LFENCE                                    WILLAMETTE,SSE2
       MFENCE                                    WILLAMETTE,SSE2

B.1.10 Willamette MMX instructions (SSE2 SIMD Integer Instructions)

       MOVD             mem,xmmreg               WILLAMETTE,SSE2,SD
       MOVD             xmmreg,mem               WILLAMETTE,SSE2,SD
       MOVD             xmmreg,rm32              WILLAMETTE,SSE2
       MOVD             rm32,xmmreg              WILLAMETTE,SSE2
       MOVDQA           xmmreg,xmmreg            WILLAMETTE,SSE2
       MOVDQA           mem,xmmreg               WILLAMETTE,SSE2,SO
       MOVDQA           xmmreg,mem               WILLAMETTE,SSE2,SO
       MOVDQA           xmmreg,xmmreg            WILLAMETTE,SSE2
       MOVDQU           xmmreg,xmmreg            WILLAMETTE,SSE2
       MOVDQU           mem,xmmreg               WILLAMETTE,SSE2,SO
       MOVDQU           xmmreg,mem               WILLAMETTE,SSE2,SO
       MOVDQU           xmmreg,xmmreg            WILLAMETTE,SSE2
       MOVDQ2Q          mmxreg,xmmreg            WILLAMETTE,SSE2
       MOVQ             xmmreg,xmmreg            WILLAMETTE,SSE2
       MOVQ             xmmreg,xmmreg            WILLAMETTE,SSE2
       MOVQ             mem,xmmreg               WILLAMETTE,SSE2
       MOVQ             xmmreg,mem               WILLAMETTE,SSE2
       MOVQ             xmmreg,rm64              X64,SSE2
       MOVQ             rm64,xmmreg              X64,SSE2
       MOVQ2DQ          xmmreg,mmxreg            WILLAMETTE,SSE2
       PACKSSWB         xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PACKSSDW         xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PACKUSWB         xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PADDB            xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PADDW            xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PADDD            xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PADDQ            mmxreg,mmxrm             WILLAMETTE,MMX
       PADDQ            xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PADDSB           xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PADDSW           xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PADDUSB          xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PADDUSW          xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PAND             xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PANDN            xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PAVGB            xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PAVGW            xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PCMPEQB          xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PCMPEQW          xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PCMPEQD          xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PCMPGTB          xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PCMPGTW          xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PCMPGTD          xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PEXTRW           reg32,xmmreg,imm         WILLAMETTE,SSE2
       PINSRW           xmmreg,reg16,imm         WILLAMETTE,SSE2
       PINSRW           xmmreg,reg32,imm         WILLAMETTE,SSE2,ND
       PINSRW           xmmreg,mem,imm           WILLAMETTE,SSE2
       PINSRW           xmmreg,mem16,imm         WILLAMETTE,SSE2
       PMADDWD          xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PMAXSW           xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PMAXUB           xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PMINSW           xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PMINUB           xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PMOVMSKB         reg32,xmmreg             WILLAMETTE,SSE2
       PMULHUW          xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PMULHW           xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PMULLW           xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PMULUDQ          mmxreg,mmxrm             WILLAMETTE,SSE2,SO
       PMULUDQ          xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       POR              xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PSADBW           xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PSHUFD           xmmreg,xmmreg,imm        WILLAMETTE,SSE2
       PSHUFD           xmmreg,mem,imm           WILLAMETTE,SSE22
       PSHUFHW          xmmreg,xmmreg,imm        WILLAMETTE,SSE2
       PSHUFHW          xmmreg,mem,imm           WILLAMETTE,SSE22
       PSHUFLW          xmmreg,xmmreg,imm        WILLAMETTE,SSE2
       PSHUFLW          xmmreg,mem,imm           WILLAMETTE,SSE22
       PSLLDQ           xmmreg,imm               WILLAMETTE,SSE2,AR1
       PSLLW            xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PSLLW            xmmreg,imm               WILLAMETTE,SSE2,AR1
       PSLLD            xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PSLLD            xmmreg,imm               WILLAMETTE,SSE2,AR1
       PSLLQ            xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PSLLQ            xmmreg,imm               WILLAMETTE,SSE2,AR1
       PSRAW            xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PSRAW            xmmreg,imm               WILLAMETTE,SSE2,AR1
       PSRAD            xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PSRAD            xmmreg,imm               WILLAMETTE,SSE2,AR1
       PSRLDQ           xmmreg,imm               WILLAMETTE,SSE2,AR1
       PSRLW            xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PSRLW            xmmreg,imm               WILLAMETTE,SSE2,AR1
       PSRLD            xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PSRLD            xmmreg,imm               WILLAMETTE,SSE2,AR1
       PSRLQ            xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PSRLQ            xmmreg,imm               WILLAMETTE,SSE2,AR1
       PSUBB            xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PSUBW            xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PSUBD            xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PSUBQ            mmxreg,mmxrm             WILLAMETTE,SSE2,SO
       PSUBQ            xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PSUBSB           xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PSUBSW           xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PSUBUSB          xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PSUBUSW          xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PUNPCKHBW        xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PUNPCKHWD        xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PUNPCKHDQ        xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PUNPCKHQDQ       xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PUNPCKLBW        xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PUNPCKLWD        xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PUNPCKLDQ        xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PUNPCKLQDQ       xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       PXOR             xmmreg,xmmrm             WILLAMETTE,SSE2,SO

B.1.11 Willamette Streaming SIMD instructions (SSE2)

       ADDPD            xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       ADDSD            xmmreg,xmmrm             WILLAMETTE,SSE2
       ANDNPD           xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       ANDPD            xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       CMPEQPD          xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       CMPEQSD          xmmreg,xmmrm             WILLAMETTE,SSE2
       CMPLEPD          xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       CMPLESD          xmmreg,xmmrm             WILLAMETTE,SSE2
       CMPLTPD          xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       CMPLTSD          xmmreg,xmmrm             WILLAMETTE,SSE2
       CMPNEQPD         xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       CMPNEQSD         xmmreg,xmmrm             WILLAMETTE,SSE2
       CMPNLEPD         xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       CMPNLESD         xmmreg,xmmrm             WILLAMETTE,SSE2
       CMPNLTPD         xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       CMPNLTSD         xmmreg,xmmrm             WILLAMETTE,SSE2
       CMPORDPD         xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       CMPORDSD         xmmreg,xmmrm             WILLAMETTE,SSE2
       CMPUNORDPD       xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       CMPUNORDSD       xmmreg,xmmrm             WILLAMETTE,SSE2
       CMPPD            xmmreg,xmmrm128,imm8     WILLAMETTE,SSE2
       CMPSD            xmmreg,xmmrm128,imm8     WILLAMETTE,SSE2
       COMISD           xmmreg,xmmrm             WILLAMETTE,SSE2
       CVTDQ2PD         xmmreg,xmmrm             WILLAMETTE,SSE2
       CVTDQ2PS         xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       CVTPD2DQ         xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       CVTPD2PI         mmxreg,xmmrm             WILLAMETTE,SSE2,SO
       CVTPD2PS         xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       CVTPI2PD         xmmreg,mmxrm             WILLAMETTE,SSE2
       CVTPS2DQ         xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       CVTPS2PD         xmmreg,xmmrm             WILLAMETTE,SSE2
       CVTSD2SI         reg32,xmmreg             WILLAMETTE,SSE2,AR1
       CVTSD2SI         reg32,mem                WILLAMETTE,SSE2,AR1
       CVTSD2SI         reg64,xmmreg             X64,SSE2,AR1
       CVTSD2SI         reg64,mem                X64,SSE2,AR1
       CVTSD2SS         xmmreg,xmmrm             WILLAMETTE,SSE2
       CVTSI2SD         xmmreg,mem               WILLAMETTE,SSE2,SD,AR1,ND
       CVTSI2SD         xmmreg,rm32              WILLAMETTE,SSE2,SD,AR1
       CVTSI2SD         xmmreg,rm64              X64,SSE2,AR1
       CVTSS2SD         xmmreg,xmmrm             WILLAMETTE,SSE2,SD
       CVTTPD2PI        mmxreg,xmmrm             WILLAMETTE,SSE2,SO
       CVTTPD2DQ        xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       CVTTPS2DQ        xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       CVTTSD2SI        reg32,xmmreg             WILLAMETTE,SSE2,AR1
       CVTTSD2SI        reg32,mem                WILLAMETTE,SSE2,AR1
       CVTTSD2SI        reg64,xmmreg             X64,SSE2,AR1
       CVTTSD2SI        reg64,mem                X64,SSE2,AR1
       DIVPD            xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       DIVSD            xmmreg,xmmrm             WILLAMETTE,SSE2
       MAXPD            xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       MAXSD            xmmreg,xmmrm             WILLAMETTE,SSE2
       MINPD            xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       MINSD            xmmreg,xmmrm             WILLAMETTE,SSE2
       MOVAPD           xmmreg,xmmreg            WILLAMETTE,SSE2
       MOVAPD           xmmreg,xmmreg            WILLAMETTE,SSE2
       MOVAPD           mem,xmmreg               WILLAMETTE,SSE2,SO
       MOVAPD           xmmreg,mem               WILLAMETTE,SSE2,SO
       MOVHPD           mem,xmmreg               WILLAMETTE,SSE2
       MOVHPD           xmmreg,mem               WILLAMETTE,SSE2
       MOVLPD           mem64,xmmreg             WILLAMETTE,SSE2
       MOVLPD           xmmreg,mem64             WILLAMETTE,SSE2
       MOVMSKPD         reg32,xmmreg             WILLAMETTE,SSE2
       MOVMSKPD         reg64,xmmreg             X64,SSE2
       MOVSD            xmmreg,xmmreg            WILLAMETTE,SSE2
       MOVSD            xmmreg,xmmreg            WILLAMETTE,SSE2
       MOVSD            mem64,xmmreg             WILLAMETTE,SSE2
       MOVSD            xmmreg,mem64             WILLAMETTE,SSE2
       MOVUPD           xmmreg,xmmreg            WILLAMETTE,SSE2
       MOVUPD           xmmreg,xmmreg            WILLAMETTE,SSE2
       MOVUPD           mem,xmmreg               WILLAMETTE,SSE2,SO
       MOVUPD           xmmreg,mem               WILLAMETTE,SSE2,SO
       MULPD            xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       MULSD            xmmreg,xmmrm             WILLAMETTE,SSE2
       ORPD             xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       SHUFPD           xmmreg,xmmreg,imm        WILLAMETTE,SSE2
       SHUFPD           xmmreg,mem,imm           WILLAMETTE,SSE2
       SQRTPD           xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       SQRTSD           xmmreg,xmmrm             WILLAMETTE,SSE2
       SUBPD            xmmreg,xmmrm             WILLAMETTE,SSE2,SO
       SUBSD            xmmreg,xmmrm             WILLAMETTE,SSE2
       UCOMISD          xmmreg,xmmrm             WILLAMETTE,SSE2
       UNPCKHPD         xmmreg,xmmrm128          WILLAMETTE,SSE2
       UNPCKLPD         xmmreg,xmmrm128          WILLAMETTE,SSE2
       XORPD            xmmreg,xmmrm128          WILLAMETTE,SSE2

B.1.12 Prescott New Instructions (SSE3)

       ADDSUBPD         xmmreg,xmmrm             PRESCOTT,SSE3,SO
       ADDSUBPS         xmmreg,xmmrm             PRESCOTT,SSE3,SO
       HADDPD           xmmreg,xmmrm             PRESCOTT,SSE3,SO
       HADDPS           xmmreg,xmmrm             PRESCOTT,SSE3,SO
       HSUBPD           xmmreg,xmmrm             PRESCOTT,SSE3,SO
       HSUBPS           xmmreg,xmmrm             PRESCOTT,SSE3,SO
       LDDQU            xmmreg,mem               PRESCOTT,SSE3,SO
       MOVDDUP          xmmreg,xmmrm             PRESCOTT,SSE3
       MOVSHDUP         xmmreg,xmmrm             PRESCOTT,SSE3
       MOVSLDUP         xmmreg,xmmrm             PRESCOTT,SSE3

B.1.13 VMX/SVM Instructions

       CLGI                                      VMX,AMD
       STGI                                      VMX,AMD
       VMCALL                                    VMX
       VMCLEAR          mem                      VMX
       VMFUNC                                    VMX
       VMLAUNCH                                  VMX
       VMLOAD                                    VMX,AMD
       VMMCALL                                   VMX,AMD
       VMPTRLD          mem                      VMX
       VMPTRST          mem                      VMX
       VMREAD           rm32,reg32               VMX,NOLONG,SD
       VMREAD           rm64,reg64               X64,VMX
       VMRESUME                                  VMX
       VMRUN                                     VMX,AMD
       VMSAVE                                    VMX,AMD
       VMWRITE          reg32,rm32               VMX,NOLONG,SD
       VMWRITE          reg64,rm64               X64,VMX
       VMXOFF                                    VMX
       VMXON            mem                      VMX

B.1.14 Extended Page Tables VMX instructions

       INVEPT           reg32,mem                VMX,SO,NOLONG
       INVEPT           reg64,mem                VMX,SO,LONG
       INVVPID          reg32,mem                VMX,SO,NOLONG
       INVVPID          reg64,mem                VMX,SO,LONG

B.1.15 Tejas New Instructions (SSSE3)

       PABSB            mmxreg,mmxrm             SSSE3,MMX
       PABSB            xmmreg,xmmrm             SSSE3
       PABSW            mmxreg,mmxrm             SSSE3,MMX
       PABSW            xmmreg,xmmrm             SSSE3
       PABSD            mmxreg,mmxrm             SSSE3,MMX
       PABSD            xmmreg,xmmrm             SSSE3
       PALIGNR          mmxreg,mmxrm,imm         SSSE3,MMX
       PALIGNR          xmmreg,xmmrm,imm         SSSE3
       PHADDW           mmxreg,mmxrm             SSSE3,MMX
       PHADDW           xmmreg,xmmrm             SSSE3
       PHADDD           mmxreg,mmxrm             SSSE3,MMX
       PHADDD           xmmreg,xmmrm             SSSE3
       PHADDSW          mmxreg,mmxrm             SSSE3,MMX
       PHADDSW          xmmreg,xmmrm             SSSE3
       PHSUBW           mmxreg,mmxrm             SSSE3,MMX
       PHSUBW           xmmreg,xmmrm             SSSE3
       PHSUBD           mmxreg,mmxrm             SSSE3,MMX
       PHSUBD           xmmreg,xmmrm             SSSE3
       PHSUBSW          mmxreg,mmxrm             SSSE3,MMX
       PHSUBSW          xmmreg,xmmrm             SSSE3
       PMADDUBSW        mmxreg,mmxrm             SSSE3,MMX
       PMADDUBSW        xmmreg,xmmrm             SSSE3
       PMULHRSW         mmxreg,mmxrm             SSSE3,MMX
       PMULHRSW         xmmreg,xmmrm             SSSE3
       PSHUFB           mmxreg,mmxrm             SSSE3,MMX
       PSHUFB           xmmreg,xmmrm             SSSE3
       PSIGNB           mmxreg,mmxrm             SSSE3,MMX
       PSIGNB           xmmreg,xmmrm             SSSE3
       PSIGNW           mmxreg,mmxrm             SSSE3,MMX
       PSIGNW           xmmreg,xmmrm             SSSE3
       PSIGND           mmxreg,mmxrm             SSSE3,MMX
       PSIGND           xmmreg,xmmrm             SSSE3

B.1.16 AMD SSE4A

       EXTRQ            xmmreg,imm,imm           SSE4A,AMD
       EXTRQ            xmmreg,xmmreg            SSE4A,AMD
       INSERTQ          xmmreg,xmmreg,imm,imm    SSE4A,AMD
       INSERTQ          xmmreg,xmmreg            SSE4A,AMD
       MOVNTSD          mem,xmmreg               SSE4A,AMD
       MOVNTSS          mem,xmmreg               SSE4A,AMD,SD

B.1.17 New instructions in Barcelona

       LZCNT            reg16,rm16               P6,AMD
       LZCNT            reg32,rm32               P6,AMD
       LZCNT            reg64,rm64               X64,AMD

B.1.18 Penryn New Instructions (SSE4.1)

       BLENDPD          xmmreg,xmmrm,imm         SSE41
       BLENDPS          xmmreg,xmmrm,imm         SSE41
       BLENDVPD         xmmreg,xmmrm,xmm0        SSE41
       BLENDVPS         xmmreg,xmmrm,xmm0        SSE41
       DPPD             xmmreg,xmmrm,imm         SSE41
       DPPS             xmmreg,xmmrm,imm         SSE41
       EXTRACTPS        rm32,xmmreg,imm          SSE41
       EXTRACTPS        reg64,xmmreg,imm         SSE41,X64
       INSERTPS         xmmreg,xmmrm,imm         SSE41,SD
       MOVNTDQA         xmmreg,mem128            SSE41
       MPSADBW          xmmreg,xmmrm,imm         SSE41
       PACKUSDW         xmmreg,xmmrm             SSE41
       PBLENDVB         xmmreg,xmmrm,xmm0        SSE41
       PBLENDW          xmmreg,xmmrm,imm         SSE41
       PCMPEQQ          xmmreg,xmmrm             SSE41
       PEXTRB           reg32,xmmreg,imm         SSE41
       PEXTRB           mem8,xmmreg,imm          SSE41
       PEXTRB           reg64,xmmreg,imm         SSE41,X64
       PEXTRD           rm32,xmmreg,imm          SSE41
       PEXTRQ           rm64,xmmreg,imm          SSE41,X64
       PEXTRW           reg32,xmmreg,imm         SSE41
       PEXTRW           mem16,xmmreg,imm         SSE41
       PEXTRW           reg64,xmmreg,imm         SSE41,X64
       PHMINPOSUW       xmmreg,xmmrm             SSE41
       PINSRB           xmmreg,mem,imm           SSE41
       PINSRB           xmmreg,rm8,imm           SSE41
       PINSRB           xmmreg,reg32,imm         SSE41
       PINSRD           xmmreg,mem,imm           SSE41
       PINSRD           xmmreg,rm32,imm          SSE41
       PINSRQ           xmmreg,mem,imm           SSE41,X64
       PINSRQ           xmmreg,rm64,imm          SSE41,X64
       PMAXSB           xmmreg,xmmrm             SSE41
       PMAXSD           xmmreg,xmmrm             SSE41
       PMAXUD           xmmreg,xmmrm             SSE41
       PMAXUW           xmmreg,xmmrm             SSE41
       PMINSB           xmmreg,xmmrm             SSE41
       PMINSD           xmmreg,xmmrm             SSE41
       PMINUD           xmmreg,xmmrm             SSE41
       PMINUW           xmmreg,xmmrm             SSE41
       PMOVSXBW         xmmreg,xmmrm             SSE41
       PMOVSXBD         xmmreg,xmmrm             SSE41,SD
       PMOVSXBQ         xmmreg,xmmrm             SSE41,SW
       PMOVSXWD         xmmreg,xmmrm             SSE41
       PMOVSXWQ         xmmreg,xmmrm             SSE41,SD
       PMOVSXDQ         xmmreg,xmmrm             SSE41
       PMOVZXBW         xmmreg,xmmrm             SSE41
       PMOVZXBD         xmmreg,xmmrm             SSE41,SD
       PMOVZXBQ         xmmreg,xmmrm             SSE41,SW
       PMOVZXWD         xmmreg,xmmrm             SSE41
       PMOVZXWQ         xmmreg,xmmrm             SSE41,SD
       PMOVZXDQ         xmmreg,xmmrm             SSE41
       PMULDQ           xmmreg,xmmrm             SSE41
       PMULLD           xmmreg,xmmrm             SSE41
       PTEST            xmmreg,xmmrm             SSE41
       ROUNDPD          xmmreg,xmmrm,imm         SSE41
       ROUNDPS          xmmreg,xmmrm,imm         SSE41
       ROUNDSD          xmmreg,xmmrm,imm         SSE41
       ROUNDSS          xmmreg,xmmrm,imm         SSE41

B.1.19 Nehalem New Instructions (SSE4.2)

       CRC32            reg32,rm8                SSE42
       CRC32            reg32,rm16               SSE42
       CRC32            reg32,rm32               SSE42
       CRC32            reg64,rm8                SSE42,X64
       CRC32            reg64,rm64               SSE42,X64
       PCMPESTRI        xmmreg,xmmrm,imm         SSE42
       PCMPESTRM        xmmreg,xmmrm,imm         SSE42
       PCMPISTRI        xmmreg,xmmrm,imm         SSE42
       PCMPISTRM        xmmreg,xmmrm,imm         SSE42
       PCMPGTQ          xmmreg,xmmrm             SSE42
       POPCNT           reg16,rm16               NEHALEM,SW
       POPCNT           reg32,rm32               NEHALEM,SD
       POPCNT           reg64,rm64               NEHALEM,X64

B.1.20 Intel SMX

       GETSEC                                    KATMAI

B.1.21 Geode (Cyrix) 3DNow! additions

       PFRCPV           mmxreg,mmxrm             PENT,3DNOW,CYRIX
       PFRSQRTV         mmxreg,mmxrm             PENT,3DNOW,CYRIX

B.1.22 Intel new instructions in ???

       MOVBE            reg16,mem16              NEHALEM
       MOVBE            reg32,mem32              NEHALEM
       MOVBE            reg64,mem64              NEHALEM
       MOVBE            mem16,reg16              NEHALEM
       MOVBE            mem32,reg32              NEHALEM
       MOVBE            mem64,reg64              NEHALEM

B.1.23 Intel AES instructions

       AESENC           xmmreg,xmmrm128          SSE,WESTMERE
       AESENCLAST       xmmreg,xmmrm128          SSE,WESTMERE
       AESDEC           xmmreg,xmmrm128          SSE,WESTMERE
       AESDECLAST       xmmreg,xmmrm128          SSE,WESTMERE
       AESIMC           xmmreg,xmmrm128          SSE,WESTMERE
       AESKEYGENASSIST  xmmreg,xmmrm128,imm8     SSE,WESTMERE

B.1.24 Intel AVX AES instructions

       VAESENC          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VAESENCLAST      xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VAESDEC          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VAESDECLAST      xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VAESIMC          xmmreg,xmmrm128          AVX,SANDYBRIDGE
       VAESKEYGENASSIST xmmreg,xmmrm128,imm8     AVX,SANDYBRIDGE

B.1.25 Intel AVX instructions

       VADDPD           xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VADDPD           ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VADDPS           xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VADDPS           ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VADDSD           xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VADDSS           xmmreg,xmmreg*,xmmrm32   AVX,SANDYBRIDGE
       VADDSUBPD        xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VADDSUBPD        ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VADDSUBPS        xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VADDSUBPS        ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VANDPD           xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VANDPD           ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VANDPS           xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VANDPS           ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VANDNPD          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VANDNPD          ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VANDNPS          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VANDNPS          ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VBLENDPD         xmmreg,xmmreg*,xmmrm128,imm8 AVX,SANDYBRIDGE
       VBLENDPD         ymmreg,ymmreg*,ymmrm256,imm8 AVX,SANDYBRIDGE
       VBLENDPS         xmmreg,xmmreg*,xmmrm128,imm8 AVX,SANDYBRIDGE
       VBLENDPS         ymmreg,ymmreg*,ymmrm256,imm8 AVX,SANDYBRIDGE
       VBLENDVPD        xmmreg,xmmreg*,xmmrm128,xmmreg AVX,SANDYBRIDGE
       VBLENDVPD        ymmreg,ymmreg*,ymmrm256,ymmreg AVX,SANDYBRIDGE
       VBLENDVPS        xmmreg,xmmreg*,xmmrm128,xmmreg AVX,SANDYBRIDGE
       VBLENDVPS        ymmreg,ymmreg*,ymmrm256,ymmreg AVX,SANDYBRIDGE
       VBROADCASTSS     xmmreg,mem32             AVX,SANDYBRIDGE
       VBROADCASTSS     ymmreg,mem32             AVX,SANDYBRIDGE
       VBROADCASTSD     ymmreg,mem64             AVX,SANDYBRIDGE
       VBROADCASTF128   ymmreg,mem128            AVX,SANDYBRIDGE
       VCMPEQ_OSPD      xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPEQ_OSPD      ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPEQPD         xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPEQPD         ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPLT_OSPD      xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPLT_OSPD      ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPLTPD         xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPLTPD         ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPLE_OSPD      xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPLE_OSPD      ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPLEPD         xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPLEPD         ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPUNORD_QPD    xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPUNORD_QPD    ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPUNORDPD      xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPUNORDPD      ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPNEQ_UQPD     xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPNEQ_UQPD     ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPNEQPD        xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPNEQPD        ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPNLT_USPD     xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPNLT_USPD     ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPNLTPD        xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPNLTPD        ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPNLE_USPD     xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPNLE_USPD     ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPNLEPD        xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPNLEPD        ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPORD_QPD      xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPORD_QPD      ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPORDPD        xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPORDPD        ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPEQ_UQPD      xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPEQ_UQPD      ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPNGE_USPD     xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPNGE_USPD     ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPNGEPD        xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPNGEPD        ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPNGT_USPD     xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPNGT_USPD     ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPNGTPD        xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPNGTPD        ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPFALSE_OQPD   xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPFALSE_OQPD   ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPFALSEPD      xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPFALSEPD      ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPNEQ_OQPD     xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPNEQ_OQPD     ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPGE_OSPD      xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPGE_OSPD      ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPGEPD         xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPGEPD         ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPGT_OSPD      xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPGT_OSPD      ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPGTPD         xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPGTPD         ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPTRUE_UQPD    xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPTRUE_UQPD    ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPTRUEPD       xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPTRUEPD       ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPEQ_OSPD      xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPEQ_OSPD      ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPLT_OQPD      xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPLT_OQPD      ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPLE_OQPD      xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPLE_OQPD      ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPUNORD_SPD    xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPUNORD_SPD    ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPNEQ_USPD     xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPNEQ_USPD     ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPNLT_UQPD     xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPNLT_UQPD     ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPNLE_UQPD     xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPNLE_UQPD     ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPORD_SPD      xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPORD_SPD      ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPEQ_USPD      xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPEQ_USPD      ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPNGE_UQPD     xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPNGE_UQPD     ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPNGT_UQPD     xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPNGT_UQPD     ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPFALSE_OSPD   xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPFALSE_OSPD   ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPNEQ_OSPD     xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPNEQ_OSPD     ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPGE_OQPD      xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPGE_OQPD      ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPGT_OQPD      xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPGT_OQPD      ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPTRUE_USPD    xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPTRUE_USPD    ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPPD           xmmreg,xmmreg*,xmmrm128,imm8 AVX,SANDYBRIDGE
       VCMPPD           ymmreg,ymmreg*,ymmrm256,imm8 AVX,SANDYBRIDGE
       VCMPEQ_OSPS      xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPEQ_OSPS      ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPEQPS         xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPEQPS         ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPLT_OSPS      xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPLT_OSPS      ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPLTPS         xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPLTPS         ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPLE_OSPS      xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPLE_OSPS      ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPLEPS         xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPLEPS         ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPUNORD_QPS    xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPUNORD_QPS    ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPUNORDPS      xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPUNORDPS      ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPNEQ_UQPS     xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPNEQ_UQPS     ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPNEQPS        xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPNEQPS        ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPNLT_USPS     xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPNLT_USPS     ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPNLTPS        xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPNLTPS        ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPNLE_USPS     xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPNLE_USPS     ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPNLEPS        xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPNLEPS        ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPORD_QPS      xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPORD_QPS      ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPORDPS        xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPORDPS        ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPEQ_UQPS      xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPEQ_UQPS      ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPNGE_USPS     xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPNGE_USPS     ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPNGEPS        xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPNGEPS        ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPNGT_USPS     xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPNGT_USPS     ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPNGTPS        xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPNGTPS        ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPFALSE_OQPS   xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPFALSE_OQPS   ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPFALSEPS      xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPFALSEPS      ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPNEQ_OQPS     xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPNEQ_OQPS     ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPGE_OSPS      xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPGE_OSPS      ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPGEPS         xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPGEPS         ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPGT_OSPS      xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPGT_OSPS      ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPGTPS         xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPGTPS         ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPTRUE_UQPS    xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPTRUE_UQPS    ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPTRUEPS       xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPTRUEPS       ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPEQ_OSPS      xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPEQ_OSPS      ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPLT_OQPS      xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPLT_OQPS      ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPLE_OQPS      xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPLE_OQPS      ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPUNORD_SPS    xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPUNORD_SPS    ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPNEQ_USPS     xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPNEQ_USPS     ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPNLT_UQPS     xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPNLT_UQPS     ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPNLE_UQPS     xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPNLE_UQPS     ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPORD_SPS      xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPORD_SPS      ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPEQ_USPS      xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPEQ_USPS      ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPNGE_UQPS     xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPNGE_UQPS     ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPNGT_UQPS     xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPNGT_UQPS     ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPFALSE_OSPS   xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPFALSE_OSPS   ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPNEQ_OSPS     xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPNEQ_OSPS     ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPGE_OQPS      xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPGE_OQPS      ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPGT_OQPS      xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPGT_OQPS      ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPTRUE_USPS    xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VCMPTRUE_USPS    ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VCMPPS           xmmreg,xmmreg*,xmmrm128,imm8 AVX,SANDYBRIDGE
       VCMPPS           ymmreg,ymmreg*,ymmrm256,imm8 AVX,SANDYBRIDGE
       VCMPEQ_OSSD      xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPEQSD         xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPLT_OSSD      xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPLTSD         xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPLE_OSSD      xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPLESD         xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPUNORD_QSD    xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPUNORDSD      xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPNEQ_UQSD     xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPNEQSD        xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPNLT_USSD     xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPNLTSD        xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPNLE_USSD     xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPNLESD        xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPORD_QSD      xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPORDSD        xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPEQ_UQSD      xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPNGE_USSD     xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPNGESD        xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPNGT_USSD     xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPNGTSD        xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPFALSE_OQSD   xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPFALSESD      xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPNEQ_OQSD     xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPGE_OSSD      xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPGESD         xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPGT_OSSD      xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPGTSD         xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPTRUE_UQSD    xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPTRUESD       xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPEQ_OSSD      xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPLT_OQSD      xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPLE_OQSD      xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPUNORD_SSD    xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPNEQ_USSD     xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPNLT_UQSD     xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPNLE_UQSD     xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPORD_SSD      xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPEQ_USSD      xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPNGE_UQSD     xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPNGT_UQSD     xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPFALSE_OSSD   xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPNEQ_OSSD     xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPGE_OQSD      xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPGT_OQSD      xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPTRUE_USSD    xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPSD           xmmreg,xmmreg*,xmmrm64,imm8 AVX,SANDYBRIDGE
       VCMPEQ_OSSS      xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPEQSS         xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPLT_OSSS      xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPLTSS         xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPLE_OSSS      xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPLESS         xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPUNORD_QSS    xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPUNORDSS      xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPNEQ_UQSS     xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPNEQSS        xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPNLT_USSS     xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPNLTSS        xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPNLE_USSS     xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPNLESS        xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPORD_QSS      xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPORDSS        xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPEQ_UQSS      xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPNGE_USSS     xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPNGESS        xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPNGT_USSS     xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPNGTSS        xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPFALSE_OQSS   xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPFALSESS      xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPNEQ_OQSS     xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPGE_OSSS      xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPGESS         xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPGT_OSSS      xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPGTSS         xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPTRUE_UQSS    xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPTRUESS       xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPEQ_OSSS      xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPLT_OQSS      xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPLE_OQSS      xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPUNORD_SSS    xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPNEQ_USSS     xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPNLT_UQSS     xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPNLE_UQSS     xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPORD_SSS      xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPEQ_USSS      xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPNGE_UQSS     xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPNGT_UQSS     xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPFALSE_OSSS   xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPNEQ_OSSS     xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPGE_OQSS      xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPGT_OQSS      xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPTRUE_USSS    xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCMPSS           xmmreg,xmmreg*,xmmrm64,imm8 AVX,SANDYBRIDGE
       VCOMISD          xmmreg,xmmrm64           AVX,SANDYBRIDGE
       VCOMISS          xmmreg,xmmrm32           AVX,SANDYBRIDGE
       VCVTDQ2PD        xmmreg,xmmrm64           AVX,SANDYBRIDGE
       VCVTDQ2PD        ymmreg,xmmrm128          AVX,SANDYBRIDGE
       VCVTDQ2PS        xmmreg,xmmrm128          AVX,SANDYBRIDGE
       VCVTDQ2PS        ymmreg,ymmrm256          AVX,SANDYBRIDGE
       VCVTPD2DQ        xmmreg,xmmreg            AVX,SANDYBRIDGE
       VCVTPD2DQ        xmmreg,mem128            AVX,SANDYBRIDGE,SO
       VCVTPD2DQ        xmmreg,ymmreg            AVX,SANDYBRIDGE
       VCVTPD2DQ        xmmreg,mem256            AVX,SANDYBRIDGE,SY
       VCVTPD2PS        xmmreg,xmmreg            AVX,SANDYBRIDGE
       VCVTPD2PS        xmmreg,mem128            AVX,SANDYBRIDGE,SO
       VCVTPD2PS        xmmreg,ymmreg            AVX,SANDYBRIDGE
       VCVTPD2PS        xmmreg,mem256            AVX,SANDYBRIDGE,SY
       VCVTPS2DQ        xmmreg,xmmrm128          AVX,SANDYBRIDGE
       VCVTPS2DQ        ymmreg,ymmrm256          AVX,SANDYBRIDGE
       VCVTPS2PD        xmmreg,xmmrm64           AVX,SANDYBRIDGE
       VCVTPS2PD        ymmreg,xmmrm128          AVX,SANDYBRIDGE
       VCVTSD2SI        reg32,xmmrm64            AVX,SANDYBRIDGE
       VCVTSD2SI        reg64,xmmrm64            AVX,SANDYBRIDGE,LONG
       VCVTSD2SS        xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VCVTSI2SD        xmmreg,xmmreg*,rm32      AVX,SANDYBRIDGE,SD
       VCVTSI2SD        xmmreg,xmmreg*,mem32     AVX,SANDYBRIDGE,ND,SD
       VCVTSI2SD        xmmreg,xmmreg*,rm64      AVX,SANDYBRIDGE,LONG
       VCVTSI2SS        xmmreg,xmmreg*,rm32      AVX,SANDYBRIDGE,SD
       VCVTSI2SS        xmmreg,xmmreg*,mem32     AVX,SANDYBRIDGE,ND,SD
       VCVTSI2SS        xmmreg,xmmreg*,rm64      AVX,SANDYBRIDGE,LONG
       VCVTSS2SD        xmmreg,xmmreg*,xmmrm32   AVX,SANDYBRIDGE
       VCVTSS2SI        reg32,xmmrm32            AVX,SANDYBRIDGE
       VCVTSS2SI        reg64,xmmrm32            AVX,SANDYBRIDGE,LONG
       VCVTTPD2DQ       xmmreg,xmmreg            AVX,SANDYBRIDGE
       VCVTTPD2DQ       xmmreg,mem128            AVX,SANDYBRIDGE,SO
       VCVTTPD2DQ       xmmreg,ymmreg            AVX,SANDYBRIDGE
       VCVTTPD2DQ       xmmreg,mem256            AVX,SANDYBRIDGE,SY
       VCVTTPS2DQ       xmmreg,xmmrm128          AVX,SANDYBRIDGE
       VCVTTPS2DQ       ymmreg,ymmrm256          AVX,SANDYBRIDGE
       VCVTTSD2SI       reg32,xmmrm64            AVX,SANDYBRIDGE
       VCVTTSD2SI       reg64,xmmrm64            AVX,SANDYBRIDGE,LONG
       VCVTTSS2SI       reg32,xmmrm32            AVX,SANDYBRIDGE
       VCVTTSS2SI       reg64,xmmrm32            AVX,SANDYBRIDGE,LONG
       VDIVPD           xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VDIVPD           ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VDIVPS           xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VDIVPS           ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VDIVSD           xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VDIVSS           xmmreg,xmmreg*,xmmrm32   AVX,SANDYBRIDGE
       VDPPD            xmmreg,xmmreg*,xmmrm128,imm8 AVX,SANDYBRIDGE
       VDPPS            xmmreg,xmmreg*,xmmrm128,imm8 AVX,SANDYBRIDGE
       VDPPS            ymmreg,ymmreg*,ymmrm256,imm8 AVX,SANDYBRIDGE
       VEXTRACTF128     xmmrm128,ymmreg,imm8     AVX,SANDYBRIDGE
       VEXTRACTPS       rm32,xmmreg,imm8         AVX,SANDYBRIDGE
       VHADDPD          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VHADDPD          ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VHADDPS          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VHADDPS          ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VHSUBPD          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VHSUBPD          ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VHSUBPS          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VHSUBPS          ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VINSERTF128      ymmreg,ymmreg,xmmrm128,imm8 AVX,SANDYBRIDGE
       VINSERTPS        xmmreg,xmmreg*,xmmrm32,imm8 AVX,SANDYBRIDGE
       VLDDQU           xmmreg,mem128            AVX,SANDYBRIDGE
       VLDQQU           ymmreg,mem256            AVX,SANDYBRIDGE
       VLDDQU           ymmreg,mem256            AVX,SANDYBRIDGE
       VLDMXCSR         mem32                    AVX,SANDYBRIDGE
       VMASKMOVDQU      xmmreg,xmmreg            AVX,SANDYBRIDGE
       VMASKMOVPS       xmmreg,xmmreg,mem128     AVX,SANDYBRIDGE
       VMASKMOVPS       ymmreg,ymmreg,mem256     AVX,SANDYBRIDGE
       VMASKMOVPS       mem128,xmmreg,xmmreg     AVX,SANDYBRIDGE,SO
       VMASKMOVPS       mem256,ymmreg,ymmreg     AVX,SANDYBRIDGE,SY
       VMASKMOVPD       xmmreg,xmmreg,mem128     AVX,SANDYBRIDGE
       VMASKMOVPD       ymmreg,ymmreg,mem256     AVX,SANDYBRIDGE
       VMASKMOVPD       mem128,xmmreg,xmmreg     AVX,SANDYBRIDGE
       VMASKMOVPD       mem256,ymmreg,ymmreg     AVX,SANDYBRIDGE
       VMAXPD           xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VMAXPD           ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VMAXPS           xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VMAXPS           ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VMAXSD           xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VMAXSS           xmmreg,xmmreg*,xmmrm32   AVX,SANDYBRIDGE
       VMINPD           xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VMINPD           ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VMINPS           xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VMINPS           ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VMINSD           xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VMINSS           xmmreg,xmmreg*,xmmrm32   AVX,SANDYBRIDGE
       VMOVAPD          xmmreg,xmmrm128          AVX,SANDYBRIDGE
       VMOVAPD          xmmrm128,xmmreg          AVX,SANDYBRIDGE
       VMOVAPD          ymmreg,ymmrm256          AVX,SANDYBRIDGE
       VMOVAPD          ymmrm256,ymmreg          AVX,SANDYBRIDGE
       VMOVAPS          xmmreg,xmmrm128          AVX,SANDYBRIDGE
       VMOVAPS          xmmrm128,xmmreg          AVX,SANDYBRIDGE
       VMOVAPS          ymmreg,ymmrm256          AVX,SANDYBRIDGE
       VMOVAPS          ymmrm256,ymmreg          AVX,SANDYBRIDGE
       VMOVD            xmmreg,rm32              AVX,SANDYBRIDGE
       VMOVD            rm32,xmmreg              AVX,SANDYBRIDGE
       VMOVQ            xmmreg,xmmrm64           AVX,SANDYBRIDGE
       VMOVQ            xmmrm64,xmmreg           AVX,SANDYBRIDGE
       VMOVQ            xmmreg,rm64              AVX,SANDYBRIDGE,LONG
       VMOVQ            rm64,xmmreg              AVX,SANDYBRIDGE,LONG
       VMOVDDUP         xmmreg,xmmrm64           AVX,SANDYBRIDGE
       VMOVDDUP         ymmreg,ymmrm256          AVX,SANDYBRIDGE
       VMOVDQA          xmmreg,xmmrm128          AVX,SANDYBRIDGE
       VMOVDQA          xmmrm128,xmmreg          AVX,SANDYBRIDGE
       VMOVQQA          ymmreg,ymmrm256          AVX,SANDYBRIDGE
       VMOVQQA          ymmrm256,ymmreg          AVX,SANDYBRIDGE
       VMOVDQA          ymmreg,ymmrm             AVX,SANDYBRIDGE
       VMOVDQA          ymmrm256,ymmreg          AVX,SANDYBRIDGE
       VMOVDQU          xmmreg,xmmrm128          AVX,SANDYBRIDGE
       VMOVDQU          xmmrm128,xmmreg          AVX,SANDYBRIDGE
       VMOVQQU          ymmreg,ymmrm256          AVX,SANDYBRIDGE
       VMOVQQU          ymmrm256,ymmreg          AVX,SANDYBRIDGE
       VMOVDQU          ymmreg,ymmrm256          AVX,SANDYBRIDGE
       VMOVDQU          ymmrm256,ymmreg          AVX,SANDYBRIDGE
       VMOVHLPS         xmmreg,xmmreg*,xmmreg    AVX,SANDYBRIDGE
       VMOVHPD          xmmreg,xmmreg*,mem64     AVX,SANDYBRIDGE
       VMOVHPD          mem64,xmmreg             AVX,SANDYBRIDGE
       VMOVHPS          xmmreg,xmmreg*,mem64     AVX,SANDYBRIDGE
       VMOVHPS          mem64,xmmreg             AVX,SANDYBRIDGE
       VMOVLHPS         xmmreg,xmmreg*,xmmreg    AVX,SANDYBRIDGE
       VMOVLPD          xmmreg,xmmreg*,mem64     AVX,SANDYBRIDGE
       VMOVLPD          mem64,xmmreg             AVX,SANDYBRIDGE
       VMOVLPS          xmmreg,xmmreg*,mem64     AVX,SANDYBRIDGE
       VMOVLPS          mem64,xmmreg             AVX,SANDYBRIDGE
       VMOVMSKPD        reg64,xmmreg             AVX,SANDYBRIDGE,LONG
       VMOVMSKPD        reg32,xmmreg             AVX,SANDYBRIDGE
       VMOVMSKPD        reg64,ymmreg             AVX,SANDYBRIDGE,LONG
       VMOVMSKPD        reg32,ymmreg             AVX,SANDYBRIDGE
       VMOVMSKPS        reg64,xmmreg             AVX,SANDYBRIDGE,LONG
       VMOVMSKPS        reg32,xmmreg             AVX,SANDYBRIDGE
       VMOVMSKPS        reg64,ymmreg             AVX,SANDYBRIDGE,LONG
       VMOVMSKPS        reg32,ymmreg             AVX,SANDYBRIDGE
       VMOVNTDQ         mem128,xmmreg            AVX,SANDYBRIDGE
       VMOVNTQQ         mem256,ymmreg            AVX,SANDYBRIDGE
       VMOVNTDQ         mem256,ymmreg            AVX,SANDYBRIDGE
       VMOVNTDQA        xmmreg,mem128            AVX,SANDYBRIDGE
       VMOVNTPD         mem128,xmmreg            AVX,SANDYBRIDGE
       VMOVNTPD         mem256,ymmreg            AVX,SANDYBRIDGE
       VMOVNTPS         mem128,xmmreg            AVX,SANDYBRIDGE
       VMOVNTPS         mem128,ymmreg            AVX,SANDYBRIDGE
       VMOVSD           xmmreg,xmmreg*,xmmreg    AVX,SANDYBRIDGE
       VMOVSD           xmmreg,mem64             AVX,SANDYBRIDGE
       VMOVSD           xmmreg,xmmreg*,xmmreg    AVX,SANDYBRIDGE
       VMOVSD           mem64,xmmreg             AVX,SANDYBRIDGE
       VMOVSHDUP        xmmreg,xmmrm128          AVX,SANDYBRIDGE
       VMOVSHDUP        ymmreg,ymmrm256          AVX,SANDYBRIDGE
       VMOVSLDUP        xmmreg,xmmrm128          AVX,SANDYBRIDGE
       VMOVSLDUP        ymmreg,ymmrm256          AVX,SANDYBRIDGE
       VMOVSS           xmmreg,xmmreg*,xmmreg    AVX,SANDYBRIDGE
       VMOVSS           xmmreg,mem32             AVX,SANDYBRIDGE
       VMOVSS           xmmreg,xmmreg*,xmmreg    AVX,SANDYBRIDGE
       VMOVSS           mem32,xmmreg             AVX,SANDYBRIDGE
       VMOVUPD          xmmreg,xmmrm128          AVX,SANDYBRIDGE
       VMOVUPD          xmmrm128,xmmreg          AVX,SANDYBRIDGE
       VMOVUPD          ymmreg,ymmrm256          AVX,SANDYBRIDGE
       VMOVUPD          ymmrm256,ymmreg          AVX,SANDYBRIDGE
       VMOVUPS          xmmreg,xmmrm128          AVX,SANDYBRIDGE
       VMOVUPS          xmmrm128,xmmreg          AVX,SANDYBRIDGE
       VMOVUPS          ymmreg,ymmrm256          AVX,SANDYBRIDGE
       VMOVUPS          ymmrm256,ymmreg          AVX,SANDYBRIDGE
       VMPSADBW         xmmreg,xmmreg*,xmmrm128,imm8 AVX,SANDYBRIDGE
       VMULPD           xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VMULPD           ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VMULPS           xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VMULPS           ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VMULSD           xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VMULSS           xmmreg,xmmreg*,xmmrm32   AVX,SANDYBRIDGE
       VORPD            xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VORPD            ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VORPS            xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VORPS            ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VPABSB           xmmreg,xmmrm128          AVX,SANDYBRIDGE
       VPABSW           xmmreg,xmmrm128          AVX,SANDYBRIDGE
       VPABSD           xmmreg,xmmrm128          AVX,SANDYBRIDGE
       VPACKSSWB        xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPACKSSDW        xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPACKUSWB        xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPACKUSDW        xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPADDB           xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPADDW           xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPADDD           xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPADDQ           xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPADDSB          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPADDSW          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPADDUSB         xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPADDUSW         xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPALIGNR         xmmreg,xmmreg*,xmmrm128,imm8 AVX,SANDYBRIDGE
       VPAND            xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPANDN           xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPAVGB           xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPAVGW           xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPBLENDVB        xmmreg,xmmreg*,xmmrm128,xmmreg AVX,SANDYBRIDGE
       VPBLENDW         xmmreg,xmmreg*,xmmrm128,imm8 AVX,SANDYBRIDGE
       VPCMPESTRI       xmmreg,xmmrm128,imm8     AVX,SANDYBRIDGE
       VPCMPESTRM       xmmreg,xmmrm128,imm8     AVX,SANDYBRIDGE
       VPCMPISTRI       xmmreg,xmmrm128,imm8     AVX,SANDYBRIDGE
       VPCMPISTRM       xmmreg,xmmrm128,imm8     AVX,SANDYBRIDGE
       VPCMPEQB         xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPCMPEQW         xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPCMPEQD         xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPCMPEQQ         xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPCMPGTB         xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPCMPGTW         xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPCMPGTD         xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPCMPGTQ         xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPERMILPD        xmmreg,xmmreg,xmmrm128   AVX,SANDYBRIDGE
       VPERMILPD        ymmreg,ymmreg,ymmrm256   AVX,SANDYBRIDGE
       VPERMILPD        xmmreg,xmmrm128,imm8     AVX,SANDYBRIDGE
       VPERMILPD        ymmreg,ymmrm256,imm8     AVX,SANDYBRIDGE
       VPERMILPS        xmmreg,xmmreg,xmmrm128   AVX,SANDYBRIDGE
       VPERMILPS        ymmreg,ymmreg,ymmrm256   AVX,SANDYBRIDGE
       VPERMILPS        xmmreg,xmmrm128,imm8     AVX,SANDYBRIDGE
       VPERMILPS        ymmreg,ymmrm256,imm8     AVX,SANDYBRIDGE
       VPERM2F128       ymmreg,ymmreg,ymmrm256,imm8 AVX,SANDYBRIDGE
       VPEXTRB          reg64,xmmreg,imm8        AVX,SANDYBRIDGE,LONG
       VPEXTRB          reg32,xmmreg,imm8        AVX,SANDYBRIDGE
       VPEXTRB          mem8,xmmreg,imm8         AVX,SANDYBRIDGE
       VPEXTRW          reg64,xmmreg,imm8        AVX,SANDYBRIDGE,LONG
       VPEXTRW          reg32,xmmreg,imm8        AVX,SANDYBRIDGE
       VPEXTRW          reg64,xmmreg,imm8        AVX,SANDYBRIDGE,LONG
       VPEXTRW          reg32,xmmreg,imm8        AVX,SANDYBRIDGE
       VPEXTRW          mem16,xmmreg,imm8        AVX,SANDYBRIDGE
       VPEXTRD          reg64,xmmreg,imm8        AVX,SANDYBRIDGE,LONG
       VPEXTRD          rm32,xmmreg,imm8         AVX,SANDYBRIDGE
       VPEXTRQ          rm64,xmmreg,imm8         AVX,SANDYBRIDGE,LONG
       VPHADDW          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPHADDD          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPHADDSW         xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPHMINPOSUW      xmmreg,xmmrm128          AVX,SANDYBRIDGE
       VPHSUBW          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPHSUBD          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPHSUBSW         xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPINSRB          xmmreg,xmmreg*,mem8,imm8 AVX,SANDYBRIDGE
       VPINSRB          xmmreg,xmmreg*,rm8,imm8  AVX,SANDYBRIDGE
       VPINSRB          xmmreg,xmmreg*,reg32,imm8 AVX,SANDYBRIDGE
       VPINSRW          xmmreg,xmmreg*,mem16,imm8 AVX,SANDYBRIDGE
       VPINSRW          xmmreg,xmmreg*,rm16,imm8 AVX,SANDYBRIDGE
       VPINSRW          xmmreg,xmmreg*,reg32,imm8 AVX,SANDYBRIDGE
       VPINSRD          xmmreg,xmmreg*,mem32,imm8 AVX,SANDYBRIDGE
       VPINSRD          xmmreg,xmmreg*,rm32,imm8 AVX,SANDYBRIDGE
       VPINSRQ          xmmreg,xmmreg*,mem64,imm8 AVX,SANDYBRIDGE,LONG
       VPINSRQ          xmmreg,xmmreg*,rm64,imm8 AVX,SANDYBRIDGE,LONG
       VPMADDWD         xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPMADDUBSW       xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPMAXSB          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPMAXSW          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPMAXSD          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPMAXUB          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPMAXUW          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPMAXUD          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPMINSB          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPMINSW          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPMINSD          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPMINUB          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPMINUW          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPMINUD          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPMOVMSKB        reg64,xmmreg             AVX,SANDYBRIDGE,LONG
       VPMOVMSKB        reg32,xmmreg             AVX,SANDYBRIDGE
       VPMOVSXBW        xmmreg,xmmrm64           AVX,SANDYBRIDGE
       VPMOVSXBD        xmmreg,xmmrm32           AVX,SANDYBRIDGE
       VPMOVSXBQ        xmmreg,xmmrm16           AVX,SANDYBRIDGE
       VPMOVSXWD        xmmreg,xmmrm64           AVX,SANDYBRIDGE
       VPMOVSXWQ        xmmreg,xmmrm32           AVX,SANDYBRIDGE
       VPMOVSXDQ        xmmreg,xmmrm64           AVX,SANDYBRIDGE
       VPMOVZXBW        xmmreg,xmmrm64           AVX,SANDYBRIDGE
       VPMOVZXBD        xmmreg,xmmrm32           AVX,SANDYBRIDGE
       VPMOVZXBQ        xmmreg,xmmrm16           AVX,SANDYBRIDGE
       VPMOVZXWD        xmmreg,xmmrm64           AVX,SANDYBRIDGE
       VPMOVZXWQ        xmmreg,xmmrm32           AVX,SANDYBRIDGE
       VPMOVZXDQ        xmmreg,xmmrm64           AVX,SANDYBRIDGE
       VPMULHUW         xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPMULHRSW        xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPMULHW          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPMULLW          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPMULLD          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPMULUDQ         xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPMULDQ          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPOR             xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPSADBW          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPSHUFB          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPSHUFD          xmmreg,xmmrm128,imm8     AVX,SANDYBRIDGE
       VPSHUFHW         xmmreg,xmmrm128,imm8     AVX,SANDYBRIDGE
       VPSHUFLW         xmmreg,xmmrm128,imm8     AVX,SANDYBRIDGE
       VPSIGNB          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPSIGNW          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPSIGND          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPSLLDQ          xmmreg,xmmreg*,imm8      AVX,SANDYBRIDGE
       VPSRLDQ          xmmreg,xmmreg*,imm8      AVX,SANDYBRIDGE
       VPSLLW           xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPSLLW           xmmreg,xmmreg*,imm8      AVX,SANDYBRIDGE
       VPSLLD           xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPSLLD           xmmreg,xmmreg*,imm8      AVX,SANDYBRIDGE
       VPSLLQ           xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPSLLQ           xmmreg,xmmreg*,imm8      AVX,SANDYBRIDGE
       VPSRAW           xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPSRAW           xmmreg,xmmreg*,imm8      AVX,SANDYBRIDGE
       VPSRAD           xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPSRAD           xmmreg,xmmreg*,imm8      AVX,SANDYBRIDGE
       VPSRLW           xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPSRLW           xmmreg,xmmreg*,imm8      AVX,SANDYBRIDGE
       VPSRLD           xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPSRLD           xmmreg,xmmreg*,imm8      AVX,SANDYBRIDGE
       VPSRLQ           xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPSRLQ           xmmreg,xmmreg*,imm8      AVX,SANDYBRIDGE
       VPTEST           xmmreg,xmmrm128          AVX,SANDYBRIDGE
       VPTEST           ymmreg,ymmrm256          AVX,SANDYBRIDGE
       VPSUBB           xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPSUBW           xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPSUBD           xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPSUBQ           xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPSUBSB          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPSUBSW          xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPSUBUSB         xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPSUBUSW         xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPUNPCKHBW       xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPUNPCKHWD       xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPUNPCKHDQ       xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPUNPCKHQDQ      xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPUNPCKLBW       xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPUNPCKLWD       xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPUNPCKLDQ       xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPUNPCKLQDQ      xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPXOR            xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VRCPPS           xmmreg,xmmrm128          AVX,SANDYBRIDGE
       VRCPPS           ymmreg,ymmrm256          AVX,SANDYBRIDGE
       VRCPSS           xmmreg,xmmreg*,xmmrm32   AVX,SANDYBRIDGE
       VRSQRTPS         xmmreg,xmmrm128          AVX,SANDYBRIDGE
       VRSQRTPS         ymmreg,ymmrm256          AVX,SANDYBRIDGE
       VRSQRTSS         xmmreg,xmmreg*,xmmrm32   AVX,SANDYBRIDGE
       VROUNDPD         xmmreg,xmmrm128,imm8     AVX,SANDYBRIDGE
       VROUNDPD         ymmreg,ymmrm256,imm8     AVX,SANDYBRIDGE
       VROUNDPS         xmmreg,xmmrm128,imm8     AVX,SANDYBRIDGE
       VROUNDPS         ymmreg,ymmrm256,imm8     AVX,SANDYBRIDGE
       VROUNDSD         xmmreg,xmmreg*,xmmrm64,imm8 AVX,SANDYBRIDGE
       VROUNDSS         xmmreg,xmmreg*,xmmrm32,imm8 AVX,SANDYBRIDGE
       VSHUFPD          xmmreg,xmmreg*,xmmrm128,imm8 AVX,SANDYBRIDGE
       VSHUFPD          ymmreg,ymmreg*,ymmrm256,imm8 AVX,SANDYBRIDGE
       VSHUFPS          xmmreg,xmmreg*,xmmrm128,imm8 AVX,SANDYBRIDGE
       VSHUFPS          ymmreg,ymmreg*,ymmrm256,imm8 AVX,SANDYBRIDGE
       VSQRTPD          xmmreg,xmmrm128          AVX,SANDYBRIDGE
       VSQRTPD          ymmreg,ymmrm256          AVX,SANDYBRIDGE
       VSQRTPS          xmmreg,xmmrm128          AVX,SANDYBRIDGE
       VSQRTPS          ymmreg,ymmrm256          AVX,SANDYBRIDGE
       VSQRTSD          xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VSQRTSS          xmmreg,xmmreg*,xmmrm32   AVX,SANDYBRIDGE
       VSTMXCSR         mem32                    AVX,SANDYBRIDGE
       VSUBPD           xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VSUBPD           ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VSUBPS           xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VSUBPS           ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VSUBSD           xmmreg,xmmreg*,xmmrm64   AVX,SANDYBRIDGE
       VSUBSS           xmmreg,xmmreg*,xmmrm32   AVX,SANDYBRIDGE
       VTESTPS          xmmreg,xmmrm128          AVX,SANDYBRIDGE
       VTESTPS          ymmreg,ymmrm256          AVX,SANDYBRIDGE
       VTESTPD          xmmreg,xmmrm128          AVX,SANDYBRIDGE
       VTESTPD          ymmreg,ymmrm256          AVX,SANDYBRIDGE
       VUCOMISD         xmmreg,xmmrm64           AVX,SANDYBRIDGE
       VUCOMISS         xmmreg,xmmrm32           AVX,SANDYBRIDGE
       VUNPCKHPD        xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VUNPCKHPD        ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VUNPCKHPS        xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VUNPCKHPS        ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VUNPCKLPD        xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VUNPCKLPD        ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VUNPCKLPS        xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VUNPCKLPS        ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VXORPD           xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VXORPD           ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VXORPS           xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VXORPS           ymmreg,ymmreg*,ymmrm256  AVX,SANDYBRIDGE
       VZEROALL                                  AVX,SANDYBRIDGE
       VZEROUPPER                                AVX,SANDYBRIDGE

B.1.26 Intel Carry-Less Multiplication instructions (CLMUL)

       PCLMULLQLQDQ     xmmreg,xmmrm128          SSE,WESTMERE
       PCLMULHQLQDQ     xmmreg,xmmrm128          SSE,WESTMERE
       PCLMULLQHQDQ     xmmreg,xmmrm128          SSE,WESTMERE
       PCLMULHQHQDQ     xmmreg,xmmrm128          SSE,WESTMERE
       PCLMULQDQ        xmmreg,xmmrm128,imm8     SSE,WESTMERE

B.1.27 Intel AVX Carry-Less Multiplication instructions (CLMUL)

       VPCLMULLQLQDQ    xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPCLMULHQLQDQ    xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPCLMULLQHQDQ    xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPCLMULHQHQDQ    xmmreg,xmmreg*,xmmrm128  AVX,SANDYBRIDGE
       VPCLMULQDQ       xmmreg,xmmreg*,xmmrm128,imm8 AVX,SANDYBRIDGE

B.1.28 Intel Fused Multiply-Add instructions (FMA)

       VFMADD132PS      xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMADD132PS      ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMADD132PD      xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMADD132PD      ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMADD312PS      xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMADD312PS      ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMADD312PD      xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMADD312PD      ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMADD213PS      xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMADD213PS      ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMADD213PD      xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMADD213PD      ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMADD123PS      xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMADD123PS      ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMADD123PD      xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMADD123PD      ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMADD231PS      xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMADD231PS      ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMADD231PD      xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMADD231PD      ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMADD321PS      xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMADD321PS      ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMADD321PD      xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMADD321PD      ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMADDSUB132PS   xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMADDSUB132PS   ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMADDSUB132PD   xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMADDSUB132PD   ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMADDSUB312PS   xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMADDSUB312PS   ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMADDSUB312PD   xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMADDSUB312PD   ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMADDSUB213PS   xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMADDSUB213PS   ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMADDSUB213PD   xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMADDSUB213PD   ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMADDSUB123PS   xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMADDSUB123PS   ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMADDSUB123PD   xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMADDSUB123PD   ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMADDSUB231PS   xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMADDSUB231PS   ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMADDSUB231PD   xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMADDSUB231PD   ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMADDSUB321PS   xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMADDSUB321PS   ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMADDSUB321PD   xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMADDSUB321PD   ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMSUB132PS      xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMSUB132PS      ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMSUB132PD      xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMSUB132PD      ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMSUB312PS      xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMSUB312PS      ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMSUB312PD      xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMSUB312PD      ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMSUB213PS      xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMSUB213PS      ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMSUB213PD      xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMSUB213PD      ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMSUB123PS      xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMSUB123PS      ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMSUB123PD      xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMSUB123PD      ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMSUB231PS      xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMSUB231PS      ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMSUB231PD      xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMSUB231PD      ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMSUB321PS      xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMSUB321PS      ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMSUB321PD      xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMSUB321PD      ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMSUBADD132PS   xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMSUBADD132PS   ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMSUBADD132PD   xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMSUBADD132PD   ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMSUBADD312PS   xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMSUBADD312PS   ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMSUBADD312PD   xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMSUBADD312PD   ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMSUBADD213PS   xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMSUBADD213PS   ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMSUBADD213PD   xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMSUBADD213PD   ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMSUBADD123PS   xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMSUBADD123PS   ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMSUBADD123PD   xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMSUBADD123PD   ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMSUBADD231PS   xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMSUBADD231PS   ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMSUBADD231PD   xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMSUBADD231PD   ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMSUBADD321PS   xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMSUBADD321PS   ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMSUBADD321PD   xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFMSUBADD321PD   ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFNMADD132PS     xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFNMADD132PS     ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFNMADD132PD     xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFNMADD132PD     ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFNMADD312PS     xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFNMADD312PS     ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFNMADD312PD     xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFNMADD312PD     ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFNMADD213PS     xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFNMADD213PS     ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFNMADD213PD     xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFNMADD213PD     ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFNMADD123PS     xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFNMADD123PS     ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFNMADD123PD     xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFNMADD123PD     ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFNMADD231PS     xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFNMADD231PS     ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFNMADD231PD     xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFNMADD231PD     ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFNMADD321PS     xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFNMADD321PS     ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFNMADD321PD     xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFNMADD321PD     ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFNMSUB132PS     xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFNMSUB132PS     ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFNMSUB132PD     xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFNMSUB132PD     ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFNMSUB312PS     xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFNMSUB312PS     ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFNMSUB312PD     xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFNMSUB312PD     ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFNMSUB213PS     xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFNMSUB213PS     ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFNMSUB213PD     xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFNMSUB213PD     ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFNMSUB123PS     xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFNMSUB123PS     ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFNMSUB123PD     xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFNMSUB123PD     ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFNMSUB231PS     xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFNMSUB231PS     ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFNMSUB231PD     xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFNMSUB231PD     ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFNMSUB321PS     xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFNMSUB321PS     ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFNMSUB321PD     xmmreg,xmmreg,xmmrm128   FMA,FUTURE
       VFNMSUB321PD     ymmreg,ymmreg,ymmrm256   FMA,FUTURE
       VFMADD132SS      xmmreg,xmmreg,xmmrm32    FMA,FUTURE
       VFMADD132SD      xmmreg,xmmreg,xmmrm64    FMA,FUTURE
       VFMADD312SS      xmmreg,xmmreg,xmmrm32    FMA,FUTURE
       VFMADD312SD      xmmreg,xmmreg,xmmrm64    FMA,FUTURE
       VFMADD213SS      xmmreg,xmmreg,xmmrm32    FMA,FUTURE
       VFMADD213SD      xmmreg,xmmreg,xmmrm64    FMA,FUTURE
       VFMADD123SS      xmmreg,xmmreg,xmmrm32    FMA,FUTURE
       VFMADD123SD      xmmreg,xmmreg,xmmrm64    FMA,FUTURE
       VFMADD231SS      xmmreg,xmmreg,xmmrm32    FMA,FUTURE
       VFMADD231SD      xmmreg,xmmreg,xmmrm64    FMA,FUTURE
       VFMADD321SS      xmmreg,xmmreg,xmmrm32    FMA,FUTURE
       VFMADD321SD      xmmreg,xmmreg,xmmrm64    FMA,FUTURE
       VFMSUB132SS      xmmreg,xmmreg,xmmrm32    FMA,FUTURE
       VFMSUB132SD      xmmreg,xmmreg,xmmrm64    FMA,FUTURE
       VFMSUB312SS      xmmreg,xmmreg,xmmrm32    FMA,FUTURE
       VFMSUB312SD      xmmreg,xmmreg,xmmrm64    FMA,FUTURE
       VFMSUB213SS      xmmreg,xmmreg,xmmrm32    FMA,FUTURE
       VFMSUB213SD      xmmreg,xmmreg,xmmrm64    FMA,FUTURE
       VFMSUB123SS      xmmreg,xmmreg,xmmrm32    FMA,FUTURE
       VFMSUB123SD      xmmreg,xmmreg,xmmrm64    FMA,FUTURE
       VFMSUB231SS      xmmreg,xmmreg,xmmrm32    FMA,FUTURE
       VFMSUB231SD      xmmreg,xmmreg,xmmrm64    FMA,FUTURE
       VFMSUB321SS      xmmreg,xmmreg,xmmrm32    FMA,FUTURE
       VFMSUB321SD      xmmreg,xmmreg,xmmrm64    FMA,FUTURE
       VFNMADD132SS     xmmreg,xmmreg,xmmrm32    FMA,FUTURE
       VFNMADD132SD     xmmreg,xmmreg,xmmrm64    FMA,FUTURE
       VFNMADD312SS     xmmreg,xmmreg,xmmrm32    FMA,FUTURE
       VFNMADD312SD     xmmreg,xmmreg,xmmrm64    FMA,FUTURE
       VFNMADD213SS     xmmreg,xmmreg,xmmrm32    FMA,FUTURE
       VFNMADD213SD     xmmreg,xmmreg,xmmrm64    FMA,FUTURE
       VFNMADD123SS     xmmreg,xmmreg,xmmrm32    FMA,FUTURE
       VFNMADD123SD     xmmreg,xmmreg,xmmrm64    FMA,FUTURE
       VFNMADD231SS     xmmreg,xmmreg,xmmrm32    FMA,FUTURE
       VFNMADD231SD     xmmreg,xmmreg,xmmrm64    FMA,FUTURE
       VFNMADD321SS     xmmreg,xmmreg,xmmrm32    FMA,FUTURE
       VFNMADD321SD     xmmreg,xmmreg,xmmrm64    FMA,FUTURE
       VFNMSUB132SS     xmmreg,xmmreg,xmmrm32    FMA,FUTURE
       VFNMSUB132SD     xmmreg,xmmreg,xmmrm64    FMA,FUTURE
       VFNMSUB312SS     xmmreg,xmmreg,xmmrm32    FMA,FUTURE
       VFNMSUB312SD     xmmreg,xmmreg,xmmrm64    FMA,FUTURE
       VFNMSUB213SS     xmmreg,xmmreg,xmmrm32    FMA,FUTURE
       VFNMSUB213SD     xmmreg,xmmreg,xmmrm64    FMA,FUTURE
       VFNMSUB123SS     xmmreg,xmmreg,xmmrm32    FMA,FUTURE
       VFNMSUB123SD     xmmreg,xmmreg,xmmrm64    FMA,FUTURE
       VFNMSUB231SS     xmmreg,xmmreg,xmmrm32    FMA,FUTURE
       VFNMSUB231SD     xmmreg,xmmreg,xmmrm64    FMA,FUTURE
       VFNMSUB321SS     xmmreg,xmmreg,xmmrm32    FMA,FUTURE
       VFNMSUB321SD     xmmreg,xmmreg,xmmrm64    FMA,FUTURE

B.1.29 Intel post-32 nm processor instructions

       RDFSBASE         reg32                    LONG,FUTURE
       RDFSBASE         reg64                    LONG,FUTURE
       RDGSBASE         reg32                    LONG,FUTURE
       RDGSBASE         reg64                    LONG,FUTURE
       RDRAND           reg16                    FUTURE
       RDRAND           reg32                    FUTURE
       RDRAND           reg64                    LONG,FUTURE
       WRFSBASE         reg32                    LONG,FUTURE
       WRFSBASE         reg64                    LONG,FUTURE
       WRGSBASE         reg32                    LONG,FUTURE
       WRGSBASE         reg64                    LONG,FUTURE
       VCVTPH2PS        ymmreg,xmmrm128          AVX,FUTURE
       VCVTPH2PS        xmmreg,xmmrm64           AVX,FUTURE
       VCVTPS2PH        xmmrm128,ymmreg,imm8     AVX,FUTURE
       VCVTPS2PH        xmmrm64,xmmreg,imm8      AVX,FUTURE
       ADCX             reg32,rm32               FUTURE
       ADCX             reg64,rm64               LONG,FUTURE
       ADOX             reg32,rm32               FUTURE
       ADOX             reg64,rm64               LONG,FUTURE
       RDSEED           reg16                    FUTURE
       RDSEED           reg32                    FUTURE
       RDSEED           reg64                    LONG,FUTURE
       CLAC                                      PRIV,FUTURE
       STAC                                      PRIV,FUTURE

B.1.30 VIA (Centaur) security instructions

       XSTORE                                    PENT,CYRIX
       XCRYPTECB                                 PENT,CYRIX
       XCRYPTCBC                                 PENT,CYRIX
       XCRYPTCTR                                 PENT,CYRIX
       XCRYPTCFB                                 PENT,CYRIX
       XCRYPTOFB                                 PENT,CYRIX
       MONTMUL                                   PENT,CYRIX
       XSHA1                                     PENT,CYRIX
       XSHA256                                   PENT,CYRIX

B.1.31 AMD Lightweight Profiling (LWP) instructions

       LLWPCB           reg32                    AMD,386
       LLWPCB           reg64                    AMD,X64
       SLWPCB           reg32                    AMD,386
       SLWPCB           reg64                    AMD,X64
       LWPVAL           reg32,rm32,imm32         AMD,386
       LWPVAL           reg64,rm32,imm32         AMD,X64
       LWPINS           reg32,rm32,imm32         AMD,386
       LWPINS           reg64,rm32,imm32         AMD,X64

B.1.32 AMD XOP and FMA4 instructions (SSE5)

       VFMADDPD         xmmreg,xmmreg*,xmmrm128,xmmreg AMD,SSE5
       VFMADDPD         ymmreg,ymmreg*,ymmrm256,ymmreg AMD,SSE5
       VFMADDPD         xmmreg,xmmreg*,xmmreg,xmmrm128 AMD,SSE5
       VFMADDPD         ymmreg,ymmreg*,ymmreg,ymmrm256 AMD,SSE5
       VFMADDPS         xmmreg,xmmreg*,xmmrm128,xmmreg AMD,SSE5
       VFMADDPS         ymmreg,ymmreg*,ymmrm256,ymmreg AMD,SSE5
       VFMADDPS         xmmreg,xmmreg*,xmmreg,xmmrm128 AMD,SSE5
       VFMADDPS         ymmreg,ymmreg*,ymmreg,ymmrm256 AMD,SSE5
       VFMADDSD         xmmreg,xmmreg*,xmmrm64,xmmreg AMD,SSE5
       VFMADDSD         xmmreg,xmmreg*,xmmreg,xmmrm64 AMD,SSE5
       VFMADDSS         xmmreg,xmmreg*,xmmrm32,xmmreg AMD,SSE5
       VFMADDSS         xmmreg,xmmreg*,xmmreg,xmmrm32 AMD,SSE5
       VFMADDSUBPD      xmmreg,xmmreg*,xmmrm128,xmmreg AMD,SSE5
       VFMADDSUBPD      ymmreg,ymmreg*,ymmrm256,ymmreg AMD,SSE5
       VFMADDSUBPD      xmmreg,xmmreg*,xmmreg,xmmrm128 AMD,SSE5
       VFMADDSUBPD      ymmreg,ymmreg*,ymmreg,ymmrm256 AMD,SSE5
       VFMADDSUBPS      xmmreg,xmmreg*,xmmrm128,xmmreg AMD,SSE5
       VFMADDSUBPS      ymmreg,ymmreg*,ymmrm256,ymmreg AMD,SSE5
       VFMADDSUBPS      xmmreg,xmmreg*,xmmreg,xmmrm128 AMD,SSE5
       VFMADDSUBPS      ymmreg,ymmreg*,ymmreg,ymmrm256 AMD,SSE5
       VFMSUBADDPD      xmmreg,xmmreg*,xmmrm128,xmmreg AMD,SSE5
       VFMSUBADDPD      ymmreg,ymmreg*,ymmrm256,ymmreg AMD,SSE5
       VFMSUBADDPD      xmmreg,xmmreg*,xmmreg,xmmrm128 AMD,SSE5
       VFMSUBADDPD      ymmreg,ymmreg*,ymmreg,ymmrm256 AMD,SSE5
       VFMSUBADDPS      xmmreg,xmmreg*,xmmrm128,xmmreg AMD,SSE5
       VFMSUBADDPS      ymmreg,ymmreg*,ymmrm256,ymmreg AMD,SSE5
       VFMSUBADDPS      xmmreg,xmmreg*,xmmreg,xmmrm128 AMD,SSE5
       VFMSUBADDPS      ymmreg,ymmreg*,ymmreg,ymmrm256 AMD,SSE5
       VFMSUBPD         xmmreg,xmmreg*,xmmrm128,xmmreg AMD,SSE5
       VFMSUBPD         ymmreg,ymmreg*,ymmrm256,ymmreg AMD,SSE5
       VFMSUBPD         xmmreg,xmmreg*,xmmreg,xmmrm128 AMD,SSE5
       VFMSUBPD         ymmreg,ymmreg*,ymmreg,ymmrm256 AMD,SSE5
       VFMSUBPS         xmmreg,xmmreg*,xmmrm128,xmmreg AMD,SSE5
       VFMSUBPS         ymmreg,ymmreg*,ymmrm256,ymmreg AMD,SSE5
       VFMSUBPS         xmmreg,xmmreg*,xmmreg,xmmrm128 AMD,SSE5
       VFMSUBPS         ymmreg,ymmreg*,ymmreg,ymmrm256 AMD,SSE5
       VFMSUBSD         xmmreg,xmmreg*,xmmrm64,xmmreg AMD,SSE5
       VFMSUBSD         xmmreg,xmmreg*,xmmreg,xmmrm64 AMD,SSE5
       VFMSUBSS         xmmreg,xmmreg*,xmmrm32,xmmreg AMD,SSE5
       VFMSUBSS         xmmreg,xmmreg*,xmmreg,xmmrm32 AMD,SSE5
       VFNMADDPD        xmmreg,xmmreg*,xmmrm128,xmmreg AMD,SSE5
       VFNMADDPD        ymmreg,ymmreg*,ymmrm256,ymmreg AMD,SSE5
       VFNMADDPD        xmmreg,xmmreg*,xmmreg,xmmrm128 AMD,SSE5
       VFNMADDPD        ymmreg,ymmreg*,ymmreg,ymmrm256 AMD,SSE5
       VFNMADDPS        xmmreg,xmmreg*,xmmrm128,xmmreg AMD,SSE5
       VFNMADDPS        ymmreg,ymmreg*,ymmrm256,ymmreg AMD,SSE5
       VFNMADDPS        xmmreg,xmmreg*,xmmreg,xmmrm128 AMD,SSE5
       VFNMADDPS        ymmreg,ymmreg*,ymmreg,ymmrm256 AMD,SSE5
       VFNMADDSD        xmmreg,xmmreg*,xmmrm64,xmmreg AMD,SSE5
       VFNMADDSD        xmmreg,xmmreg*,xmmreg,xmmrm64 AMD,SSE5
       VFNMADDSS        xmmreg,xmmreg*,xmmrm32,xmmreg AMD,SSE5
       VFNMADDSS        xmmreg,xmmreg*,xmmreg,xmmrm32 AMD,SSE5
       VFNMSUBPD        xmmreg,xmmreg*,xmmrm128,xmmreg AMD,SSE5
       VFNMSUBPD        ymmreg,ymmreg*,ymmrm256,ymmreg AMD,SSE5
       VFNMSUBPD        xmmreg,xmmreg*,xmmreg,xmmrm128 AMD,SSE5
       VFNMSUBPD        ymmreg,ymmreg*,ymmreg,ymmrm256 AMD,SSE5
       VFNMSUBPS        xmmreg,xmmreg*,xmmrm128,xmmreg AMD,SSE5
       VFNMSUBPS        ymmreg,ymmreg*,ymmrm256,ymmreg AMD,SSE5
       VFNMSUBPS        xmmreg,xmmreg*,xmmreg,xmmrm128 AMD,SSE5
       VFNMSUBPS        ymmreg,ymmreg*,ymmreg,ymmrm256 AMD,SSE5
       VFNMSUBSD        xmmreg,xmmreg*,xmmrm64,xmmreg AMD,SSE5
       VFNMSUBSD        xmmreg,xmmreg*,xmmreg,xmmrm64 AMD,SSE5
       VFNMSUBSS        xmmreg,xmmreg*,xmmrm32,xmmreg AMD,SSE5
       VFNMSUBSS        xmmreg,xmmreg*,xmmreg,xmmrm32 AMD,SSE5
       VFRCZPD          xmmreg,xmmrm128*         AMD,SSE5
       VFRCZPD          ymmreg,ymmrm256*         AMD,SSE5
       VFRCZPS          xmmreg,xmmrm128*         AMD,SSE5
       VFRCZPS          ymmreg,ymmrm256*         AMD,SSE5
       VFRCZSD          xmmreg,xmmrm64*          AMD,SSE5
       VFRCZSS          xmmreg,xmmrm32*          AMD,SSE5
       VPCMOV           xmmreg,xmmreg*,xmmrm128,xmmreg AMD,SSE5
       VPCMOV           ymmreg,ymmreg*,ymmrm256,ymmreg AMD,SSE5
       VPCMOV           xmmreg,xmmreg*,xmmreg,xmmrm128 AMD,SSE5
       VPCMOV           ymmreg,ymmreg*,ymmreg,ymmrm256 AMD,SSE5
       VPCOMB           xmmreg,xmmreg*,xmmrm128,imm8 AMD,SSE5
       VPCOMD           xmmreg,xmmreg*,xmmrm128,imm8 AMD,SSE5
       VPCOMQ           xmmreg,xmmreg*,xmmrm128,imm8 AMD,SSE5
       VPCOMUB          xmmreg,xmmreg*,xmmrm128,imm8 AMD,SSE5
       VPCOMUD          xmmreg,xmmreg*,xmmrm128,imm8 AMD,SSE5
       VPCOMUQ          xmmreg,xmmreg*,xmmrm128,imm8 AMD,SSE5
       VPCOMUW          xmmreg,xmmreg*,xmmrm128,imm8 AMD,SSE5
       VPCOMW           xmmreg,xmmreg*,xmmrm128,imm8 AMD,SSE5
       VPHADDBD         xmmreg,xmmrm128*         AMD,SSE5
       VPHADDBQ         xmmreg,xmmrm128*         AMD,SSE5
       VPHADDBW         xmmreg,xmmrm128*         AMD,SSE5
       VPHADDDQ         xmmreg,xmmrm128*         AMD,SSE5
       VPHADDUBD        xmmreg,xmmrm128*         AMD,SSE5
       VPHADDUBQ        xmmreg,xmmrm128*         AMD,SSE5
       VPHADDUBW        xmmreg,xmmrm128*         AMD,SSE5
       VPHADDUDQ        xmmreg,xmmrm128*         AMD,SSE5
       VPHADDUWD        xmmreg,xmmrm128*         AMD,SSE5
       VPHADDUWQ        xmmreg,xmmrm128*         AMD,SSE5
       VPHADDWD         xmmreg,xmmrm128*         AMD,SSE5
       VPHADDWQ         xmmreg,xmmrm128*         AMD,SSE5
       VPHSUBBW         xmmreg,xmmrm128*         AMD,SSE5
       VPHSUBDQ         xmmreg,xmmrm128*         AMD,SSE5
       VPHSUBWD         xmmreg,xmmrm128*         AMD,SSE5
       VPMACSDD         xmmreg,xmmreg*,xmmrm128,xmmreg AMD,SSE5
       VPMACSDQH        xmmreg,xmmreg*,xmmrm128,xmmreg AMD,SSE5
       VPMACSDQL        xmmreg,xmmreg*,xmmrm128,xmmreg AMD,SSE5
       VPMACSSDD        xmmreg,xmmreg*,xmmrm128,xmmreg AMD,SSE5
       VPMACSSDQH       xmmreg,xmmreg*,xmmrm128,xmmreg AMD,SSE5
       VPMACSSDQL       xmmreg,xmmreg*,xmmrm128,xmmreg AMD,SSE5
       VPMACSSWD        xmmreg,xmmreg*,xmmrm128,xmmreg AMD,SSE5
       VPMACSSWW        xmmreg,xmmreg*,xmmrm128,xmmreg AMD,SSE5
       VPMACSWD         xmmreg,xmmreg*,xmmrm128,xmmreg AMD,SSE5
       VPMACSWW         xmmreg,xmmreg*,xmmrm128,xmmreg AMD,SSE5
       VPMADCSSWD       xmmreg,xmmreg*,xmmrm128,xmmreg AMD,SSE5
       VPMADCSWD        xmmreg,xmmreg*,xmmrm128,xmmreg AMD,SSE5
       VPPERM           xmmreg,xmmreg*,xmmreg,xmmrm128 AMD,SSE5
       VPPERM           xmmreg,xmmreg*,xmmrm128,xmmreg AMD,SSE5
       VPROTB           xmmreg,xmmrm128*,xmmreg  AMD,SSE5
       VPROTB           xmmreg,xmmreg*,xmmrm128  AMD,SSE5
       VPROTB           xmmreg,xmmrm128*,imm8    AMD,SSE5
       VPROTD           xmmreg,xmmrm128*,xmmreg  AMD,SSE5
       VPROTD           xmmreg,xmmreg*,xmmrm128  AMD,SSE5
       VPROTD           xmmreg,xmmrm128*,imm8    AMD,SSE5
       VPROTQ           xmmreg,xmmrm128*,xmmreg  AMD,SSE5
       VPROTQ           xmmreg,xmmreg*,xmmrm128  AMD,SSE5
       VPROTQ           xmmreg,xmmrm128*,imm8    AMD,SSE5
       VPROTW           xmmreg,xmmrm128*,xmmreg  AMD,SSE5
       VPROTW           xmmreg,xmmreg*,xmmrm128  AMD,SSE5
       VPROTW           xmmreg,xmmrm128*,imm8    AMD,SSE5
       VPSHAB           xmmreg,xmmrm128*,xmmreg  AMD,SSE5
       VPSHAB           xmmreg,xmmreg*,xmmrm128  AMD,SSE5
       VPSHAD           xmmreg,xmmrm128*,xmmreg  AMD,SSE5
       VPSHAD           xmmreg,xmmreg*,xmmrm128  AMD,SSE5
       VPSHAQ           xmmreg,xmmrm128*,xmmreg  AMD,SSE5
       VPSHAQ           xmmreg,xmmreg*,xmmrm128  AMD,SSE5
       VPSHAW           xmmreg,xmmrm128*,xmmreg  AMD,SSE5
       VPSHAW           xmmreg,xmmreg*,xmmrm128  AMD,SSE5
       VPSHLB           xmmreg,xmmrm128*,xmmreg  AMD,SSE5
       VPSHLB           xmmreg,xmmreg*,xmmrm128  AMD,SSE5
       VPSHLD           xmmreg,xmmrm128*,xmmreg  AMD,SSE5
       VPSHLD           xmmreg,xmmreg*,xmmrm128  AMD,SSE5
       VPSHLQ           xmmreg,xmmrm128*,xmmreg  AMD,SSE5
       VPSHLQ           xmmreg,xmmreg*,xmmrm128  AMD,SSE5
       VPSHLW           xmmreg,xmmrm128*,xmmreg  AMD,SSE5
       VPSHLW           xmmreg,xmmreg*,xmmrm128  AMD,SSE5

B.1.33 Intel AVX2 instructions

       VMPSADBW         ymmreg,ymmreg*,ymmrm256,imm8 FUTURE,AVX2
       VPABSB           ymmreg,ymmrm256          FUTURE,AVX2
       VPABSW           ymmreg,ymmrm256          FUTURE,AVX2
       VPABSD           ymmreg,ymmrm256          FUTURE,AVX2
       VPACKSSWB        ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPACKSSDW        ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPACKUSDW        ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPACKUSWB        ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPADDB           ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPADDW           ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPADDD           ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPADDQ           ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPADDSB          ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPADDSW          ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPADDUSB         ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPADDUSW         ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPALIGNR         ymmreg,ymmreg*,ymmrm256,imm8 FUTURE,AVX2
       VPAND            ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPANDN           ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPAVGB           ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPAVGW           ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPBLENDVB        ymmreg,ymmreg*,ymmrm256,ymmreg FUTURE,AVX2
       VPBLENDW         ymmreg,ymmreg*,ymmrm256,imm8 FUTURE,AVX2
       VPCMPEQB         ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPCMPEQW         ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPCMPEQD         ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPCMPEQQ         ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPCMPGTB         ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPCMPGTW         ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPCMPGTD         ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPCMPGTQ         ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPHADDW          ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPHADDD          ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPHADDSW         ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPHSUBW          ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPHSUBD          ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPHSUBSW         ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPMADDUBSW       ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPMADDWD         ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPMAXSB          ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPMAXSW          ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPMAXSD          ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPMAXUB          ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPMAXUW          ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPMAXUD          ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPMINSB          ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPMINSW          ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPMINSD          ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPMINUB          ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPMINUW          ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPMINUD          ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPMOVMSKB        reg32,ymmreg             FUTURE,AVX2
       VPMOVMSKB        reg64,ymmreg             FUTURE,AVX2
       VPMOVSXBW        ymmreg,xmmrm128          FUTURE,AVX2
       VPMOVSXBD        ymmreg,mem64             FUTURE,AVX2
       VPMOVSXBD        ymmreg,xmmreg            FUTURE,AVX2
       VPMOVSXBQ        ymmreg,mem32             FUTURE,AVX2
       VPMOVSXBQ        ymmreg,xmmreg            FUTURE,AVX2
       VPMOVSXWD        ymmreg,xmmrm128          FUTURE,AVX2
       VPMOVSXWQ        ymmreg,mem64             FUTURE,AVX2
       VPMOVSXWQ        ymmreg,xmmreg            FUTURE,AVX2
       VPMOVSXDQ        ymmreg,xmmrm128          FUTURE,AVX2
       VPMOVZXBW        ymmreg,xmmrm128          FUTURE,AVX2
       VPMOVZXBD        ymmreg,mem64             FUTURE,AVX2
       VPMOVZXBD        ymmreg,xmmreg            FUTURE,AVX2
       VPMOVZXBQ        ymmreg,mem32             FUTURE,AVX2
       VPMOVZXBQ        ymmreg,xmmreg            FUTURE,AVX2
       VPMOVZXWD        ymmreg,xmmrm128          FUTURE,AVX2
       VPMOVZXWQ        ymmreg,mem64             FUTURE,AVX2
       VPMOVZXWQ        ymmreg,xmmreg            FUTURE,AVX2
       VPMOVZXDQ        ymmreg,xmmrm128          FUTURE,AVX2
       VPMULDQ          ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPMULHRSW        ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPMULHUW         ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPMULHW          ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPMULLW          ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPMULLD          ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPMULUDQ         ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPOR             ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPSADBW          ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPSHUFB          ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPSHUFD          ymmreg,ymmrm256,imm8     FUTURE,AVX2
       VPSHUFHW         ymmreg,ymmrm256,imm8     FUTURE,AVX2
       VPSHUFLW         ymmreg,ymmrm256,imm8     FUTURE,AVX2
       VPSIGNB          ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPSIGNW          ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPSIGND          ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPSLLDQ          ymmreg,ymmreg*,imm8      FUTURE,AVX2
       VPSLLW           ymmreg,ymmreg*,xmmrm128  FUTURE,AVX2
       VPSLLW           ymmreg,ymmreg*,imm8      FUTURE,AVX2
       VPSLLD           ymmreg,ymmreg*,xmmrm128  FUTURE,AVX2
       VPSLLD           ymmreg,ymmreg*,imm8      FUTURE,AVX2
       VPSLLQ           ymmreg,ymmreg*,xmmrm128  FUTURE,AVX2
       VPSLLQ           ymmreg,ymmreg*,imm8      FUTURE,AVX2
       VPSRAW           ymmreg,ymmreg*,xmmrm128  FUTURE,AVX2
       VPSRAW           ymmreg,ymmreg*,imm8      FUTURE,AVX2
       VPSRAD           ymmreg,ymmreg*,xmmrm128  FUTURE,AVX2
       VPSRAD           ymmreg,ymmreg*,imm8      FUTURE,AVX2
       VPSRLDQ          ymmreg,ymmreg*,imm8      FUTURE,AVX2
       VPSRLW           ymmreg,ymmreg*,xmmrm128  FUTURE,AVX2
       VPSRLW           ymmreg,ymmreg*,imm8      FUTURE,AVX2
       VPSRLD           ymmreg,ymmreg*,xmmrm128  FUTURE,AVX2
       VPSRLD           ymmreg,ymmreg*,imm8      FUTURE,AVX2
       VPSRLQ           ymmreg,ymmreg*,xmmrm128  FUTURE,AVX2
       VPSRLQ           ymmreg,ymmreg*,imm8      FUTURE,AVX2
       VPSUBB           ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPSUBW           ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPSUBD           ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPSUBQ           ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPSUBSB          ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPSUBSW          ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPSUBUSB         ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPSUBUSW         ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPUNPCKHBW       ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPUNPCKHWD       ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPUNPCKHDQ       ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPUNPCKHQDQ      ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPUNPCKLBW       ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPUNPCKLWD       ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPUNPCKLDQ       ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPUNPCKLQDQ      ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPXOR            ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VMOVNTDQA        ymmreg,mem256            FUTURE,AVX2
       VBROADCASTSS     xmmreg,xmmreg            FUTURE,AVX2
       VBROADCASTSS     ymmreg,xmmreg            FUTURE,AVX2
       VBROADCASTSD     ymmreg,xmmreg            FUTURE,AVX2
       VBROADCASTI128   ymmreg,mem128            FUTURE,AVX2
       VPBLENDD         xmmreg,xmmreg*,xmmrm128,imm8 FUTURE,AVX2
       VPBLENDD         ymmreg,ymmreg*,ymmrm256,imm8 FUTURE,AVX2
       VPBROADCASTB     xmmreg,mem8              FUTURE,AVX2
       VPBROADCASTB     xmmreg,xmmreg            FUTURE,AVX2
       VPBROADCASTB     ymmreg,mem8              FUTURE,AVX2
       VPBROADCASTB     ymmreg,xmmreg            FUTURE,AVX2
       VPBROADCASTW     xmmreg,mem16             FUTURE,AVX2
       VPBROADCASTW     xmmreg,xmmreg            FUTURE,AVX2
       VPBROADCASTW     ymmreg,mem16             FUTURE,AVX2
       VPBROADCASTW     ymmreg,xmmreg            FUTURE,AVX2
       VPBROADCASTD     xmmreg,mem32             FUTURE,AVX2
       VPBROADCASTD     xmmreg,xmmreg            FUTURE,AVX2
       VPBROADCASTD     ymmreg,mem32             FUTURE,AVX2
       VPBROADCASTD     ymmreg,xmmreg            FUTURE,AVX2
       VPBROADCASTQ     xmmreg,mem64             FUTURE,AVX2
       VPBROADCASTQ     xmmreg,xmmreg            FUTURE,AVX2
       VPBROADCASTQ     ymmreg,mem64             FUTURE,AVX2
       VPBROADCASTQ     ymmreg,xmmreg            FUTURE,AVX2
       VPERMD           ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPERMPD          ymmreg,ymmrm256,imm8     FUTURE,AVX2
       VPERMPS          ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPERMQ           ymmreg,ymmrm256,imm8     FUTURE,AVX2
       VPERM2I128       ymmreg,ymmreg,ymmrm256,imm8 FUTURE,AVX2
       VEXTRACTI128     xmmrm128,ymmreg,imm8     FUTURE,AVX2
       VINSERTI128      ymmreg,ymmreg*,xmmrm128,imm8 FUTURE,AVX2
       VPMASKMOVD       xmmreg,xmmreg*,mem128    FUTURE,AVX2
       VPMASKMOVD       ymmreg,ymmreg*,mem256    FUTURE,AVX2
       VPMASKMOVQ       xmmreg,xmmreg*,mem128    FUTURE,AVX2
       VPMASKMOVQ       ymmreg,ymmreg*,mem256    FUTURE,AVX2
       VPMASKMOVD       mem128,xmmreg*,xmmreg    FUTURE,AVX2
       VPMASKMOVD       mem256,ymmreg*,ymmreg    FUTURE,AVX2
       VPMASKMOVQ       mem128,xmmreg*,xmmreg    FUTURE,AVX2
       VPMASKMOVQ       mem256,ymmreg*,ymmreg    FUTURE,AVX2
       VPSLLVD          xmmreg,xmmreg*,xmmrm128  FUTURE,AVX2
       VPSLLVQ          xmmreg,xmmreg*,xmmrm128  FUTURE,AVX2
       VPSLLVD          ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPSLLVQ          ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPSRAVD          xmmreg,xmmreg*,xmmrm128  FUTURE,AVX2
       VPSRAVD          ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPSRLVD          xmmreg,xmmreg*,xmmrm128  FUTURE,AVX2
       VPSRLVQ          xmmreg,xmmreg*,xmmrm128  FUTURE,AVX2
       VPSRLVD          ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VPSRLVQ          ymmreg,ymmreg*,ymmrm256  FUTURE,AVX2
       VGATHERDPD       xmmreg,xmem64,xmmreg     FUTURE,AVX2
       VGATHERQPD       xmmreg,xmem64,xmmreg     FUTURE,AVX2
       VGATHERDPD       ymmreg,xmem64,ymmreg     FUTURE,AVX2
       VGATHERQPD       ymmreg,ymem64,ymmreg     FUTURE,AVX2
       VGATHERDPS       xmmreg,xmem32,xmmreg     FUTURE,AVX2
       VGATHERQPS       xmmreg,xmem32,xmmreg     FUTURE,AVX2
       VGATHERDPS       ymmreg,ymem32,ymmreg     FUTURE,AVX2
       VGATHERQPS       xmmreg,ymem32,xmmreg     FUTURE,AVX2
       VPGATHERDD       xmmreg,xmem32,xmmreg     FUTURE,AVX2
       VPGATHERQD       xmmreg,xmem32,xmmreg     FUTURE,AVX2
       VPGATHERDD       ymmreg,ymem32,ymmreg     FUTURE,AVX2
       VPGATHERQD       xmmreg,ymem32,xmmreg     FUTURE,AVX2
       VPGATHERDQ       xmmreg,xmem64,xmmreg     FUTURE,AVX2
       VPGATHERQQ       xmmreg,xmem64,xmmreg     FUTURE,AVX2
       VPGATHERDQ       ymmreg,xmem64,ymmreg     FUTURE,AVX2
       VPGATHERQQ       ymmreg,ymem64,ymmreg     FUTURE,AVX2

B.1.34 Transactional Synchronization Extensions (TSX)

       XABORT           imm                      FUTURE,RTM
       XABORT           imm8                     FUTURE,RTM
       XBEGIN           imm                      FUTURE,RTM
       XBEGIN           imm|near                 FUTURE,RTM,ND
       XBEGIN           imm16                    FUTURE,RTM,NOLONG
       XBEGIN           imm16|near               FUTURE,RTM,NOLONG,ND
       XBEGIN           imm32                    FUTURE,RTM,NOLONG
       XBEGIN           imm32|near               FUTURE,RTM,NOLONG,ND
       XBEGIN           imm64                    FUTURE,RTM,LONG
       XBEGIN           imm64|near               FUTURE,RTM,LONG,ND
       XEND                                      FUTURE,RTM
       XTEST                                     FUTURE,HLE,RTM

B.1.35 Intel BMI1 and BMI2 instructions, AMD TBM instructions

       ANDN             reg32,reg32,rm32         FUTURE,BMI1
       ANDN             reg64,reg64,rm64         LONG,FUTURE,BMI1
       BEXTR            reg32,rm32,reg32         FUTURE,BMI1
       BEXTR            reg64,rm64,reg64         LONG,FUTURE,BMI1
       BEXTR            reg32,rm32,imm32         FUTURE,TBM
       BEXTR            reg64,rm64,imm32         LONG,FUTURE,TBM
       BLCI             reg32,rm32               FUTURE,TBM
       BLCI             reg64,rm64               LONG,FUTURE,TBM
       BLCIC            reg32,rm32               FUTURE,TBM
       BLCIC            reg64,rm64               LONG,FUTURE,TBM
       BLSI             reg32,rm32               FUTURE,BMI1
       BLSI             reg64,rm64               LONG,FUTURE,BMI1
       BLSIC            reg32,rm32               FUTURE,TBM
       BLSIC            reg64,rm64               LONG,FUTURE,TBM
       BLCFILL          reg32,rm32               FUTURE,TBM
       BLCFILL          reg64,rm64               LONG,FUTURE,TBM
       BLSFILL          reg32,rm32               FUTURE,TBM
       BLSFILL          reg64,rm64               LONG,FUTURE,TBM
       BLCMSK           reg32,rm32               FUTURE,TBM
       BLCMSK           reg64,rm64               LONG,FUTURE,TBM
       BLSMSK           reg32,rm32               FUTURE,BMI1
       BLSMSK           reg64,rm64               LONG,FUTURE,BMI1
       BLSR             reg32,rm32               FUTURE,BMI1
       BLSR             reg64,rm64               LONG,FUTURE,BMI1
       BLCS             reg32,rm32               FUTURE,TBM
       BLCS             reg64,rm64               LONG,FUTURE,TBM
       BZHI             reg32,rm32,reg32         FUTURE,BMI2
       BZHI             reg64,rm64,reg64         LONG,FUTURE,BMI2
       MULX             reg32,reg32,rm32         FUTURE,BMI2
       MULX             reg64,reg64,rm64         LONG,FUTURE,BMI2
       PDEP             reg32,reg32,rm32         FUTURE,BMI2
       PDEP             reg64,reg64,rm64         LONG,FUTURE,BMI2
       PEXT             reg32,reg32,rm32         FUTURE,BMI2
       PEXT             reg64,reg64,rm64         LONG,FUTURE,BMI2
       RORX             reg32,rm32,imm8          FUTURE,BMI2
       RORX             reg64,rm64,imm8          LONG,FUTURE,BMI2
       SARX             reg32,rm32,reg32         FUTURE,BMI2
       SARX             reg64,rm64,reg64         LONG,FUTURE,BMI2
       SHLX             reg32,rm32,reg32         FUTURE,BMI2
       SHLX             reg64,rm64,reg64         LONG,FUTURE,BMI2
       SHRX             reg32,rm32,reg32         FUTURE,BMI2
       SHRX             reg64,rm64,reg64         LONG,FUTURE,BMI2
       TZCNT            reg16,rm16               FUTURE,BMI1
       TZCNT            reg32,rm32               FUTURE,BMI1
       TZCNT            reg64,rm64               LONG,FUTURE,BMI1
       TZMSK            reg32,rm32               FUTURE,TBM
       TZMSK            reg64,rm64               LONG,FUTURE,TBM
       T1MSKC           reg32,rm32               FUTURE,TBM
       T1MSKC           reg64,rm64               LONG,FUTURE,TBM

B.1.36 Intel AVX-512 instructions

       VADDPD           zmmreg|mask|z,zmmreg,zmmrm512|b64|er AVX512,FUTURE
       VADDPS           zmmreg|mask|z,zmmreg,zmmrm512|b32|er AVX512,FUTURE
       VADDSD           xmmreg|mask|z,xmmreg,xmmrm64|er AVX512,FUTURE
       VADDSS           xmmreg|mask|z,xmmreg,xmmrm32|er AVX512,FUTURE
       VALIGND          zmmreg|mask|z,zmmreg,zmmrm512|b32,imm8 AVX512,FUTURE
       VALIGNQ          zmmreg|mask|z,zmmreg,zmmrm512|b64,imm8 AVX512,FUTURE
       VBLENDMPD        zmmreg|mask|z,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VBLENDMPS        zmmreg|mask|z,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VBROADCASTF32X4  zmmreg|mask|z,mem128     AVX512,FUTURE
       VBROADCASTF64X4  zmmreg|mask|z,mem256     AVX512,FUTURE
       VBROADCASTI32X4  zmmreg|mask|z,mem128     AVX512,FUTURE
       VBROADCASTI64X4  zmmreg|mask|z,mem256     AVX512,FUTURE
       VBROADCASTSD     zmmreg|mask|z,mem64      AVX512,FUTURE
       VBROADCASTSD     zmmreg|mask|z,xmmreg     AVX512,FUTURE
       VBROADCASTSS     zmmreg|mask|z,xmmreg     AVX512,FUTURE
       VBROADCASTSS     zmmreg|mask|z,mem32      AVX512,FUTURE
       VCMPEQPD         kreg|mask,zmmreg,zmmrm512|b64|sae AVX512,FUTURE
       VCMPLTPD         kreg|mask,zmmreg,zmmrm512|b64|sae AVX512,FUTURE
       VCMPLEPD         kreg|mask,zmmreg,zmmrm512|b64|sae AVX512,FUTURE
       VCMPUNORDPD      kreg|mask,zmmreg,zmmrm512|b64|sae AVX512,FUTURE
       VCMPNEQPD        kreg|mask,zmmreg,zmmrm512|b64|sae AVX512,FUTURE
       VCMPNLTPD        kreg|mask,zmmreg,zmmrm512|b64|sae AVX512,FUTURE
       VCMPNLEPD        kreg|mask,zmmreg,zmmrm512|b64|sae AVX512,FUTURE
       VCMPORDPD        kreg|mask,zmmreg,zmmrm512|b64|sae AVX512,FUTURE
       VCMPEQ_UQPD      kreg|mask,zmmreg,zmmrm512|b64|sae AVX512,FUTURE
       VCMPNGEPD        kreg|mask,zmmreg,zmmrm512|b64|sae AVX512,FUTURE
       VCMPNGTPD        kreg|mask,zmmreg,zmmrm512|b64|sae AVX512,FUTURE
       VCMPFALSEPD      kreg|mask,zmmreg,zmmrm512|b64|sae AVX512,FUTURE
       VCMPNEQ_OQPD     kreg|mask,zmmreg,zmmrm512|b64|sae AVX512,FUTURE
       VCMPGEPD         kreg|mask,zmmreg,zmmrm512|b64|sae AVX512,FUTURE
       VCMPGTPD         kreg|mask,zmmreg,zmmrm512|b64|sae AVX512,FUTURE
       VCMPTRUEPD       kreg|mask,zmmreg,zmmrm512|b64|sae AVX512,FUTURE
       VCMPEQ_OSPD      kreg|mask,zmmreg,zmmrm512|b64|sae AVX512,FUTURE
       VCMPLT_OQPD      kreg|mask,zmmreg,zmmrm512|b64|sae AVX512,FUTURE
       VCMPLE_OQPD      kreg|mask,zmmreg,zmmrm512|b64|sae AVX512,FUTURE
       VCMPUNORD_SPD    kreg|mask,zmmreg,zmmrm512|b64|sae AVX512,FUTURE
       VCMPNEQ_USPD     kreg|mask,zmmreg,zmmrm512|b64|sae AVX512,FUTURE
       VCMPNLT_UQPD     kreg|mask,zmmreg,zmmrm512|b64|sae AVX512,FUTURE
       VCMPNLE_UQPD     kreg|mask,zmmreg,zmmrm512|b64|sae AVX512,FUTURE
       VCMPORD_SPD      kreg|mask,zmmreg,zmmrm512|b64|sae AVX512,FUTURE
       VCMPEQ_USPD      kreg|mask,zmmreg,zmmrm512|b64|sae AVX512,FUTURE
       VCMPNGE_UQPD     kreg|mask,zmmreg,zmmrm512|b64|sae AVX512,FUTURE
       VCMPNGT_UQPD     kreg|mask,zmmreg,zmmrm512|b64|sae AVX512,FUTURE
       VCMPFALSE_OSPD   kreg|mask,zmmreg,zmmrm512|b64|sae AVX512,FUTURE
       VCMPNEQ_OSPD     kreg|mask,zmmreg,zmmrm512|b64|sae AVX512,FUTURE
       VCMPGE_OQPD      kreg|mask,zmmreg,zmmrm512|b64|sae AVX512,FUTURE
       VCMPGT_OQPD      kreg|mask,zmmreg,zmmrm512|b64|sae AVX512,FUTURE
       VCMPTRUE_USPD    kreg|mask,zmmreg,zmmrm512|b64|sae AVX512,FUTURE
       VCMPPD           kreg|mask,zmmreg,zmmrm512|b64|sae,imm8 AVX512,FUTURE
       VCMPEQPS         kreg|mask,zmmreg,zmmrm512|b32|sae AVX512,FUTURE
       VCMPLTPS         kreg|mask,zmmreg,zmmrm512|b32|sae AVX512,FUTURE
       VCMPLEPS         kreg|mask,zmmreg,zmmrm512|b32|sae AVX512,FUTURE
       VCMPUNORDPS      kreg|mask,zmmreg,zmmrm512|b32|sae AVX512,FUTURE
       VCMPNEQPS        kreg|mask,zmmreg,zmmrm512|b32|sae AVX512,FUTURE
       VCMPNLTPS        kreg|mask,zmmreg,zmmrm512|b32|sae AVX512,FUTURE
       VCMPNLEPS        kreg|mask,zmmreg,zmmrm512|b32|sae AVX512,FUTURE
       VCMPORDPS        kreg|mask,zmmreg,zmmrm512|b32|sae AVX512,FUTURE
       VCMPEQ_UQPS      kreg|mask,zmmreg,zmmrm512|b32|sae AVX512,FUTURE
       VCMPNGEPS        kreg|mask,zmmreg,zmmrm512|b32|sae AVX512,FUTURE
       VCMPNGTPS        kreg|mask,zmmreg,zmmrm512|b32|sae AVX512,FUTURE
       VCMPFALSEPS      kreg|mask,zmmreg,zmmrm512|b32|sae AVX512,FUTURE
       VCMPNEQ_OQPS     kreg|mask,zmmreg,zmmrm512|b32|sae AVX512,FUTURE
       VCMPGEPS         kreg|mask,zmmreg,zmmrm512|b32|sae AVX512,FUTURE
       VCMPGTPS         kreg|mask,zmmreg,zmmrm512|b32|sae AVX512,FUTURE
       VCMPTRUEPS       kreg|mask,zmmreg,zmmrm512|b32|sae AVX512,FUTURE
       VCMPEQ_OSPS      kreg|mask,zmmreg,zmmrm512|b32|sae AVX512,FUTURE
       VCMPLT_OQPS      kreg|mask,zmmreg,zmmrm512|b32|sae AVX512,FUTURE
       VCMPLE_OQPS      kreg|mask,zmmreg,zmmrm512|b32|sae AVX512,FUTURE
       VCMPUNORD_SPS    kreg|mask,zmmreg,zmmrm512|b32|sae AVX512,FUTURE
       VCMPNEQ_USPS     kreg|mask,zmmreg,zmmrm512|b32|sae AVX512,FUTURE
       VCMPNLT_UQPS     kreg|mask,zmmreg,zmmrm512|b32|sae AVX512,FUTURE
       VCMPNLE_UQPS     kreg|mask,zmmreg,zmmrm512|b32|sae AVX512,FUTURE
       VCMPORD_SPS      kreg|mask,zmmreg,zmmrm512|b32|sae AVX512,FUTURE
       VCMPEQ_USPS      kreg|mask,zmmreg,zmmrm512|b32|sae AVX512,FUTURE
       VCMPNGE_UQPS     kreg|mask,zmmreg,zmmrm512|b32|sae AVX512,FUTURE
       VCMPNGT_UQPS     kreg|mask,zmmreg,zmmrm512|b32|sae AVX512,FUTURE
       VCMPFALSE_OSPS   kreg|mask,zmmreg,zmmrm512|b32|sae AVX512,FUTURE
       VCMPNEQ_OSPS     kreg|mask,zmmreg,zmmrm512|b32|sae AVX512,FUTURE
       VCMPGE_OQPS      kreg|mask,zmmreg,zmmrm512|b32|sae AVX512,FUTURE
       VCMPGT_OQPS      kreg|mask,zmmreg,zmmrm512|b32|sae AVX512,FUTURE
       VCMPTRUE_USPS    kreg|mask,zmmreg,zmmrm512|b32|sae AVX512,FUTURE
       VCMPPS           kreg|mask,zmmreg,zmmrm512|b32|sae,imm8 AVX512,FUTURE
       VCMPEQSD         kreg|mask,xmmreg,xmmrm64|sae AVX512,FUTURE
       VCMPLTSD         kreg|mask,xmmreg,xmmrm64|sae AVX512,FUTURE
       VCMPLESD         kreg|mask,xmmreg,xmmrm64|sae AVX512,FUTURE
       VCMPUNORDSD      kreg|mask,xmmreg,xmmrm64|sae AVX512,FUTURE
       VCMPNEQSD        kreg|mask,xmmreg,xmmrm64|sae AVX512,FUTURE
       VCMPNLTSD        kreg|mask,xmmreg,xmmrm64|sae AVX512,FUTURE
       VCMPNLESD        kreg|mask,xmmreg,xmmrm64|sae AVX512,FUTURE
       VCMPORDSD        kreg|mask,xmmreg,xmmrm64|sae AVX512,FUTURE
       VCMPEQ_UQSD      kreg|mask,xmmreg,xmmrm64|sae AVX512,FUTURE
       VCMPNGESD        kreg|mask,xmmreg,xmmrm64|sae AVX512,FUTURE
       VCMPNGTSD        kreg|mask,xmmreg,xmmrm64|sae AVX512,FUTURE
       VCMPFALSESD      kreg|mask,xmmreg,xmmrm64|sae AVX512,FUTURE
       VCMPNEQ_OQSD     kreg|mask,xmmreg,xmmrm64|sae AVX512,FUTURE
       VCMPGESD         kreg|mask,xmmreg,xmmrm64|sae AVX512,FUTURE
       VCMPGTSD         kreg|mask,xmmreg,xmmrm64|sae AVX512,FUTURE
       VCMPTRUESD       kreg|mask,xmmreg,xmmrm64|sae AVX512,FUTURE
       VCMPEQ_OSSD      kreg|mask,xmmreg,xmmrm64|sae AVX512,FUTURE
       VCMPLT_OQSD      kreg|mask,xmmreg,xmmrm64|sae AVX512,FUTURE
       VCMPLE_OQSD      kreg|mask,xmmreg,xmmrm64|sae AVX512,FUTURE
       VCMPUNORD_SSD    kreg|mask,xmmreg,xmmrm64|sae AVX512,FUTURE
       VCMPNEQ_USSD     kreg|mask,xmmreg,xmmrm64|sae AVX512,FUTURE
       VCMPNLT_UQSD     kreg|mask,xmmreg,xmmrm64|sae AVX512,FUTURE
       VCMPNLE_UQSD     kreg|mask,xmmreg,xmmrm64|sae AVX512,FUTURE
       VCMPORD_SSD      kreg|mask,xmmreg,xmmrm64|sae AVX512,FUTURE
       VCMPEQ_USSD      kreg|mask,xmmreg,xmmrm64|sae AVX512,FUTURE
       VCMPNGE_UQSD     kreg|mask,xmmreg,xmmrm64|sae AVX512,FUTURE
       VCMPNGT_UQSD     kreg|mask,xmmreg,xmmrm64|sae AVX512,FUTURE
       VCMPFALSE_OSSD   kreg|mask,xmmreg,xmmrm64|sae AVX512,FUTURE
       VCMPNEQ_OSSD     kreg|mask,xmmreg,xmmrm64|sae AVX512,FUTURE
       VCMPGE_OQSD      kreg|mask,xmmreg,xmmrm64|sae AVX512,FUTURE
       VCMPGT_OQSD      kreg|mask,xmmreg,xmmrm64|sae AVX512,FUTURE
       VCMPTRUE_USSD    kreg|mask,xmmreg,xmmrm64|sae AVX512,FUTURE
       VCMPSD           kreg|mask,xmmreg,xmmrm64|sae,imm8 AVX512,FUTURE
       VCMPEQSS         kreg|mask,xmmreg,xmmrm32|sae AVX512,FUTURE
       VCMPLTSS         kreg|mask,xmmreg,xmmrm32|sae AVX512,FUTURE
       VCMPLESS         kreg|mask,xmmreg,xmmrm32|sae AVX512,FUTURE
       VCMPUNORDSS      kreg|mask,xmmreg,xmmrm32|sae AVX512,FUTURE
       VCMPNEQSS        kreg|mask,xmmreg,xmmrm32|sae AVX512,FUTURE
       VCMPNLTSS        kreg|mask,xmmreg,xmmrm32|sae AVX512,FUTURE
       VCMPNLESS        kreg|mask,xmmreg,xmmrm32|sae AVX512,FUTURE
       VCMPORDSS        kreg|mask,xmmreg,xmmrm32|sae AVX512,FUTURE
       VCMPEQ_UQSS      kreg|mask,xmmreg,xmmrm32|sae AVX512,FUTURE
       VCMPNGESS        kreg|mask,xmmreg,xmmrm32|sae AVX512,FUTURE
       VCMPNGTSS        kreg|mask,xmmreg,xmmrm32|sae AVX512,FUTURE
       VCMPFALSESS      kreg|mask,xmmreg,xmmrm32|sae AVX512,FUTURE
       VCMPNEQ_OQSS     kreg|mask,xmmreg,xmmrm32|sae AVX512,FUTURE
       VCMPGESS         kreg|mask,xmmreg,xmmrm32|sae AVX512,FUTURE
       VCMPGTSS         kreg|mask,xmmreg,xmmrm32|sae AVX512,FUTURE
       VCMPTRUESS       kreg|mask,xmmreg,xmmrm32|sae AVX512,FUTURE
       VCMPEQ_OSSS      kreg|mask,xmmreg,xmmrm32|sae AVX512,FUTURE
       VCMPLT_OQSS      kreg|mask,xmmreg,xmmrm32|sae AVX512,FUTURE
       VCMPLE_OQSS      kreg|mask,xmmreg,xmmrm32|sae AVX512,FUTURE
       VCMPUNORD_SSS    kreg|mask,xmmreg,xmmrm32|sae AVX512,FUTURE
       VCMPNEQ_USSS     kreg|mask,xmmreg,xmmrm32|sae AVX512,FUTURE
       VCMPNLT_UQSS     kreg|mask,xmmreg,xmmrm32|sae AVX512,FUTURE
       VCMPNLE_UQSS     kreg|mask,xmmreg,xmmrm32|sae AVX512,FUTURE
       VCMPORD_SSS      kreg|mask,xmmreg,xmmrm32|sae AVX512,FUTURE
       VCMPEQ_USSS      kreg|mask,xmmreg,xmmrm32|sae AVX512,FUTURE
       VCMPNGE_UQSS     kreg|mask,xmmreg,xmmrm32|sae AVX512,FUTURE
       VCMPNGT_UQSS     kreg|mask,xmmreg,xmmrm32|sae AVX512,FUTURE
       VCMPFALSE_OSSS   kreg|mask,xmmreg,xmmrm32|sae AVX512,FUTURE
       VCMPNEQ_OSSS     kreg|mask,xmmreg,xmmrm32|sae AVX512,FUTURE
       VCMPGE_OQSS      kreg|mask,xmmreg,xmmrm32|sae AVX512,FUTURE
       VCMPGT_OQSS      kreg|mask,xmmreg,xmmrm32|sae AVX512,FUTURE
       VCMPTRUE_USSS    kreg|mask,xmmreg,xmmrm32|sae AVX512,FUTURE
       VCMPSS           kreg|mask,xmmreg,xmmrm32|sae,imm8 AVX512,FUTURE
       VCOMISD          xmmreg,xmmrm64|sae       AVX512,FUTURE
       VCOMISS          xmmreg,xmmrm32|sae       AVX512,FUTURE
       VCOMPRESSPD      mem512|mask,zmmreg       AVX512,FUTURE
       VCOMPRESSPD      zmmreg|mask|z,zmmreg     AVX512,FUTURE
       VCOMPRESSPS      mem512|mask,zmmreg       AVX512,FUTURE
       VCOMPRESSPS      zmmreg|mask|z,zmmreg     AVX512,FUTURE
       VCVTDQ2PD        zmmreg|mask|z,ymmrm256|b32|er AVX512,FUTURE
       VCVTDQ2PS        zmmreg|mask|z,zmmrm512|b32|er AVX512,FUTURE
       VCVTPD2DQ        ymmreg|mask|z,zmmrm512|b64|er AVX512,FUTURE
       VCVTPD2PS        ymmreg|mask|z,zmmrm512|b64|er AVX512,FUTURE
       VCVTPD2UDQ       ymmreg|mask|z,zmmrm512|b64|er AVX512,FUTURE
       VCVTPH2PS        zmmreg|mask|z,ymmrm256|sae AVX512,FUTURE
       VCVTPS2DQ        zmmreg|mask|z,zmmrm512|b32|er AVX512,FUTURE
       VCVTPS2PD        zmmreg|mask|z,ymmrm256|b32|sae AVX512,FUTURE
       VCVTPS2PH        ymmreg|mask|z,zmmreg|sae,imm8 AVX512,FUTURE
       VCVTPS2PH        mem256|mask,zmmreg|sae,imm8 AVX512,FUTURE
       VCVTPS2UDQ       zmmreg|mask|z,zmmrm512|b32|er AVX512,FUTURE
       VCVTSD2SI        reg64,xmmrm64|er         AVX512,FUTURE
       VCVTSD2SI        reg32,xmmrm64|er         AVX512,FUTURE
       VCVTSD2SS        xmmreg|mask|z,xmmreg,xmmrm64|er AVX512,FUTURE
       VCVTSD2USI       reg64,xmmrm64|er         AVX512,FUTURE
       VCVTSD2USI       reg32,xmmrm64|er         AVX512,FUTURE
       VCVTSI2SD        xmmreg,xmmreg|er,rm64    AVX512,FUTURE
       VCVTSI2SD        xmmreg,xmmreg|er,rm32    AVX512,FUTURE
       VCVTSI2SS        xmmreg,xmmreg|er,rm32    AVX512,FUTURE
       VCVTSI2SS        xmmreg,xmmreg|er,rm64    AVX512,FUTURE
       VCVTSS2SD        xmmreg|mask|z,xmmreg,xmmrm32|sae AVX512,FUTURE
       VCVTSS2SI        reg32,xmmrm32|er         AVX512,FUTURE
       VCVTSS2SI        reg64,xmmrm32|er         AVX512,FUTURE
       VCVTSS2USI       reg32,xmmrm32|er         AVX512,FUTURE
       VCVTSS2USI       reg64,xmmrm32|er         AVX512,FUTURE
       VCVTTPD2DQ       ymmreg|mask|z,zmmrm512|b64|sae AVX512,FUTURE
       VCVTTPD2UDQ      ymmreg|mask|z,zmmrm512|b64|sae AVX512,FUTURE
       VCVTTPS2DQ       zmmreg|mask|z,zmmrm512|b32|sae AVX512,FUTURE
       VCVTTPS2UDQ      zmmreg|mask|z,zmmrm512|b32|sae AVX512,FUTURE
       VCVTTSD2SI       reg64,xmmrm64|sae        AVX512,FUTURE
       VCVTTSD2SI       reg32,xmmrm64|sae        AVX512,FUTURE
       VCVTTSD2USI      reg32,xmmrm64|sae        AVX512,FUTURE
       VCVTTSD2USI      reg64,xmmrm64|sae        AVX512,FUTURE
       VCVTTSS2SI       reg64,xmmrm32|sae        AVX512,FUTURE
       VCVTTSS2SI       reg32,xmmrm32|sae        AVX512,FUTURE
       VCVTTSS2USI      reg32,xmmrm32|sae        AVX512,FUTURE
       VCVTTSS2USI      reg64,xmmrm32|sae        AVX512,FUTURE
       VCVTUDQ2PD       zmmreg|mask|z,ymmrm256|b32|er AVX512,FUTURE
       VCVTUDQ2PS       zmmreg|mask|z,zmmrm512|b32|er AVX512,FUTURE
       VCVTUSI2SD       xmmreg,xmmreg|er,rm32    AVX512,FUTURE
       VCVTUSI2SD       xmmreg,xmmreg|er,rm64    AVX512,FUTURE
       VCVTUSI2SS       xmmreg,xmmreg|er,rm64    AVX512,FUTURE
       VCVTUSI2SS       xmmreg,xmmreg|er,rm32    AVX512,FUTURE
       VDIVPD           zmmreg|mask|z,zmmreg,zmmrm512|b64|er AVX512,FUTURE
       VDIVPS           zmmreg|mask|z,zmmreg,zmmrm512|b32|er AVX512,FUTURE
       VDIVSD           xmmreg|mask|z,xmmreg,xmmrm64|er AVX512,FUTURE
       VDIVSS           xmmreg|mask|z,xmmreg,xmmrm32|er AVX512,FUTURE
       VEXPANDPD        zmmreg|mask|z,mem512     AVX512,FUTURE
       VEXPANDPD        zmmreg|mask|z,zmmreg     AVX512,FUTURE
       VEXPANDPS        zmmreg|mask|z,mem512     AVX512,FUTURE
       VEXPANDPS        zmmreg|mask|z,zmmreg     AVX512,FUTURE
       VEXTRACTF32X4    mem128|mask,zmmreg,imm8  AVX512,FUTURE
       VEXTRACTF32X4    xmmreg|mask|z,zmmreg,imm8 AVX512,FUTURE
       VEXTRACTF64X4    mem256|mask,zmmreg,imm8  AVX512,FUTURE
       VEXTRACTF64X4    ymmreg|mask|z,zmmreg,imm8 AVX512,FUTURE
       VEXTRACTI32X4    mem128|mask,zmmreg,imm8  AVX512,FUTURE
       VEXTRACTI32X4    xmmreg|mask|z,zmmreg,imm8 AVX512,FUTURE
       VEXTRACTI64X4    ymmreg|mask|z,zmmreg,imm8 AVX512,FUTURE
       VEXTRACTI64X4    mem256|mask,zmmreg,imm8  AVX512,FUTURE
       VEXTRACTPS       rm64,xmmreg,imm8         AVX512,FUTURE
       VEXTRACTPS       rm32,xmmreg,imm8         AVX512,FUTURE
       VFIXUPIMMPD      zmmreg|mask|z,zmmreg,zmmrm512|b64|sae,imm8 AVX512,FUTURE
       VFIXUPIMMPS      zmmreg|mask|z,zmmreg,zmmrm512|b32|sae,imm8 AVX512,FUTURE
       VFIXUPIMMSD      xmmreg|mask|z,xmmreg,xmmrm64|sae,imm8 AVX512,FUTURE
       VFIXUPIMMSS      xmmreg|mask|z,xmmreg,xmmrm32|sae,imm8 AVX512,FUTURE
       VFMADD132PD      zmmreg|mask|z,zmmreg,zmmrm512|b64|er AVX512,FUTURE
       VFMADD132PS      zmmreg|mask|z,zmmreg,zmmrm512|b32|er AVX512,FUTURE
       VFMADD132SD      xmmreg|mask|z,xmmreg,xmmrm64|er AVX512,FUTURE
       VFMADD132SS      xmmreg|mask|z,xmmreg,xmmrm32|er AVX512,FUTURE
       VFMADD213PD      zmmreg|mask|z,zmmreg,zmmrm512|b64|er AVX512,FUTURE
       VFMADD213PS      zmmreg|mask|z,zmmreg,zmmrm512|b32|er AVX512,FUTURE
       VFMADD213SD      xmmreg|mask|z,xmmreg,xmmrm64|er AVX512,FUTURE
       VFMADD213SS      xmmreg|mask|z,xmmreg,xmmrm32|er AVX512,FUTURE
       VFMADD231PD      zmmreg|mask|z,zmmreg,zmmrm512|b64|er AVX512,FUTURE
       VFMADD231PS      zmmreg|mask|z,zmmreg,zmmrm512|b32|er AVX512,FUTURE
       VFMADD231SD      xmmreg|mask|z,xmmreg,xmmrm64|er AVX512,FUTURE
       VFMADD231SS      xmmreg|mask|z,xmmreg,xmmrm32|er AVX512,FUTURE
       VFMADDSUB132PD   zmmreg|mask|z,zmmreg,zmmrm512|b64|er AVX512,FUTURE
       VFMADDSUB132PS   zmmreg|mask|z,zmmreg,zmmrm512|b32|er AVX512,FUTURE
       VFMADDSUB213PD   zmmreg|mask|z,zmmreg,zmmrm512|b64|er AVX512,FUTURE
       VFMADDSUB213PS   zmmreg|mask|z,zmmreg,zmmrm512|b32|er AVX512,FUTURE
       VFMADDSUB231PD   zmmreg|mask|z,zmmreg,zmmrm512|b64|er AVX512,FUTURE
       VFMADDSUB231PS   zmmreg|mask|z,zmmreg,zmmrm512|b32|er AVX512,FUTURE
       VFMSUB132PD      zmmreg|mask|z,zmmreg,zmmrm512|b64|er AVX512,FUTURE
       VFMSUB132PS      zmmreg|mask|z,zmmreg,zmmrm512|b32|er AVX512,FUTURE
       VFMSUB132SD      xmmreg|mask|z,xmmreg,xmmrm64|er AVX512,FUTURE
       VFMSUB132SS      xmmreg|mask|z,xmmreg,xmmrm32|er AVX512,FUTURE
       VFMSUB213PD      zmmreg|mask|z,zmmreg,zmmrm512|b64|er AVX512,FUTURE
       VFMSUB213PS      zmmreg|mask|z,zmmreg,zmmrm512|b32|er AVX512,FUTURE
       VFMSUB213SD      xmmreg|mask|z,xmmreg,xmmrm64|er AVX512,FUTURE
       VFMSUB213SS      xmmreg|mask|z,xmmreg,xmmrm32|er AVX512,FUTURE
       VFMSUB231PD      zmmreg|mask|z,zmmreg,zmmrm512|b64|er AVX512,FUTURE
       VFMSUB231PS      zmmreg|mask|z,zmmreg,zmmrm512|b32|er AVX512,FUTURE
       VFMSUB231SD      xmmreg|mask|z,xmmreg,xmmrm64|er AVX512,FUTURE
       VFMSUB231SS      xmmreg|mask|z,xmmreg,xmmrm32|er AVX512,FUTURE
       VFMSUBADD132PD   zmmreg|mask|z,zmmreg,zmmrm512|b64|er AVX512,FUTURE
       VFMSUBADD132PS   zmmreg|mask|z,zmmreg,zmmrm512|b32|er AVX512,FUTURE
       VFMSUBADD213PD   zmmreg|mask|z,zmmreg,zmmrm512|b64|er AVX512,FUTURE
       VFMSUBADD213PS   zmmreg|mask|z,zmmreg,zmmrm512|b32|er AVX512,FUTURE
       VFMSUBADD231PD   zmmreg|mask|z,zmmreg,zmmrm512|b64|er AVX512,FUTURE
       VFMSUBADD231PS   zmmreg|mask|z,zmmreg,zmmrm512|b32|er AVX512,FUTURE
       VFNMADD132PD     zmmreg|mask|z,zmmreg,zmmrm512|b64|er AVX512,FUTURE
       VFNMADD132PS     zmmreg|mask|z,zmmreg,zmmrm512|b32|er AVX512,FUTURE
       VFNMADD132SD     xmmreg|mask|z,xmmreg,xmmrm64|er AVX512,FUTURE
       VFNMADD132SS     xmmreg|mask|z,xmmreg,xmmrm32|er AVX512,FUTURE
       VFNMADD213PD     zmmreg|mask|z,zmmreg,zmmrm512|b64|er AVX512,FUTURE
       VFNMADD213PS     zmmreg|mask|z,zmmreg,zmmrm512|b32|er AVX512,FUTURE
       VFNMADD213SD     xmmreg|mask|z,xmmreg,xmmrm64|er AVX512,FUTURE
       VFNMADD213SS     xmmreg|mask|z,xmmreg,xmmrm32|er AVX512,FUTURE
       VFNMADD231PD     zmmreg|mask|z,zmmreg,zmmrm512|b64|er AVX512,FUTURE
       VFNMADD231PS     zmmreg|mask|z,zmmreg,zmmrm512|b32|er AVX512,FUTURE
       VFNMADD231SD     xmmreg|mask|z,xmmreg,xmmrm64|er AVX512,FUTURE
       VFNMADD231SS     xmmreg|mask|z,xmmreg,xmmrm32|er AVX512,FUTURE
       VFNMSUB132PD     zmmreg|mask|z,zmmreg,zmmrm512|b64|er AVX512,FUTURE
       VFNMSUB132PS     zmmreg|mask|z,zmmreg,zmmrm512|b32|er AVX512,FUTURE
       VFNMSUB132SD     xmmreg|mask|z,xmmreg,xmmrm64|er AVX512,FUTURE
       VFNMSUB132SS     xmmreg|mask|z,xmmreg,xmmrm32|er AVX512,FUTURE
       VFNMSUB213PD     zmmreg|mask|z,zmmreg,zmmrm512|b64|er AVX512,FUTURE
       VFNMSUB213PS     zmmreg|mask|z,zmmreg,zmmrm512|b32|er AVX512,FUTURE
       VFNMSUB213SD     xmmreg|mask|z,xmmreg,xmmrm64|er AVX512,FUTURE
       VFNMSUB213SS     xmmreg|mask|z,xmmreg,xmmrm32|er AVX512,FUTURE
       VFNMSUB231PD     zmmreg|mask|z,zmmreg,zmmrm512|b64|er AVX512,FUTURE
       VFNMSUB231PS     zmmreg|mask|z,zmmreg,zmmrm512|b32|er AVX512,FUTURE
       VFNMSUB231SD     xmmreg|mask|z,xmmreg,xmmrm64|er AVX512,FUTURE
       VFNMSUB231SS     xmmreg|mask|z,xmmreg,xmmrm32|er AVX512,FUTURE
       VGATHERDPD       zmmreg|mask,ymem64       AVX512,FUTURE
       VGATHERDPS       zmmreg|mask,zmem32       AVX512,FUTURE
       VGATHERQPD       zmmreg|mask,zmem64       AVX512,FUTURE
       VGATHERQPS       ymmreg|mask,zmem32       AVX512,FUTURE
       VGETEXPPD        zmmreg|mask|z,zmmrm512|b64|sae AVX512,FUTURE
       VGETEXPPS        zmmreg|mask|z,zmmrm512|b32|sae AVX512,FUTURE
       VGETEXPSD        xmmreg|mask|z,xmmreg,xmmrm64|sae AVX512,FUTURE
       VGETEXPSS        xmmreg|mask|z,xmmreg,xmmrm32|sae AVX512,FUTURE
       VGETMANTPD       zmmreg|mask|z,zmmrm512|b64|sae,imm8 AVX512,FUTURE
       VGETMANTPS       zmmreg|mask|z,zmmrm512|b32|sae,imm8 AVX512,FUTURE
       VGETMANTSD       xmmreg|mask|z,xmmreg,xmmrm64|sae,imm8 AVX512,FUTURE
       VGETMANTSS       xmmreg|mask|z,xmmreg,xmmrm32|sae,imm8 AVX512,FUTURE
       VINSERTF32X4     zmmreg|mask|z,zmmreg,xmmrm128,imm8 AVX512,FUTURE
       VINSERTF64X4     zmmreg|mask|z,zmmreg,ymmrm256,imm8 AVX512,FUTURE
       VINSERTI32X4     zmmreg|mask|z,zmmreg,xmmrm128,imm8 AVX512,FUTURE
       VINSERTI64X4     zmmreg|mask|z,zmmreg,ymmrm256,imm8 AVX512,FUTURE
       VINSERTPS        xmmreg,xmmreg,xmmrm32,imm8 AVX512,FUTURE
       VMAXPD           zmmreg|mask|z,zmmreg,zmmrm512|b64|sae AVX512,FUTURE
       VMAXPS           zmmreg|mask|z,zmmreg,zmmrm512|b32|sae AVX512,FUTURE
       VMAXSD           xmmreg|mask|z,xmmreg,xmmrm64|sae AVX512,FUTURE
       VMAXSS           xmmreg|mask|z,xmmreg,xmmrm32|sae AVX512,FUTURE
       VMINPD           zmmreg|mask|z,zmmreg,zmmrm512|b64|sae AVX512,FUTURE
       VMINPS           zmmreg|mask|z,zmmreg,zmmrm512|b32|sae AVX512,FUTURE
       VMINSD           xmmreg|mask|z,xmmreg,xmmrm64|sae AVX512,FUTURE
       VMINSS           xmmreg|mask|z,xmmreg,xmmrm32|sae AVX512,FUTURE
       VMOVAPD          zmmreg|mask|z,zmmrm512   AVX512,FUTURE
       VMOVAPD          mem512|mask,zmmreg       AVX512,FUTURE
       VMOVAPD          zmmreg|mask|z,zmmreg     AVX512,FUTURE
       VMOVAPS          zmmreg|mask|z,zmmrm512   AVX512,FUTURE
       VMOVAPS          zmmreg|mask|z,zmmreg     AVX512,FUTURE
       VMOVAPS          mem512|mask,zmmreg       AVX512,FUTURE
       VMOVD            xmmreg,rm32              AVX512,FUTURE
       VMOVD            rm32,xmmreg              AVX512,FUTURE
       VMOVDDUP         zmmreg|mask|z,zmmrm512   AVX512,FUTURE
       VMOVDQA32        zmmreg|mask|z,zmmrm512   AVX512,FUTURE
       VMOVDQA32        mem512|mask,zmmreg       AVX512,FUTURE
       VMOVDQA32        zmmreg|mask|z,zmmreg     AVX512,FUTURE
       VMOVDQA64        zmmreg|mask|z,zmmrm512   AVX512,FUTURE
       VMOVDQA64        zmmreg|mask|z,zmmreg     AVX512,FUTURE
       VMOVDQA64        mem512|mask,zmmreg       AVX512,FUTURE
       VMOVDQU32        zmmreg|mask|z,zmmrm512   AVX512,FUTURE
       VMOVDQU32        mem512|mask,zmmreg       AVX512,FUTURE
       VMOVDQU32        zmmreg|mask|z,zmmreg     AVX512,FUTURE
       VMOVDQU64        zmmreg|mask|z,zmmrm512   AVX512,FUTURE
       VMOVDQU64        mem512|mask,zmmreg       AVX512,FUTURE
       VMOVDQU64        zmmreg|mask|z,zmmreg     AVX512,FUTURE
       VMOVHLPS         xmmreg,xmmreg,xmmreg     AVX512,FUTURE
       VMOVHPD          xmmreg,xmmreg,mem64      AVX512,FUTURE
       VMOVHPD          mem64,xmmreg             AVX512,FUTURE
       VMOVHPS          xmmreg,xmmreg,mem64      AVX512,FUTURE
       VMOVHPS          mem64,xmmreg             AVX512,FUTURE
       VMOVLHPS         xmmreg,xmmreg,xmmreg     AVX512,FUTURE
       VMOVLPD          xmmreg,xmmreg,mem64      AVX512,FUTURE
       VMOVLPD          mem64,xmmreg             AVX512,FUTURE
       VMOVLPS          xmmreg,xmmreg,mem64      AVX512,FUTURE
       VMOVLPS          mem64,xmmreg             AVX512,FUTURE
       VMOVNTDQ         mem512,zmmreg            AVX512,FUTURE
       VMOVNTDQA        zmmreg,mem512            AVX512,FUTURE
       VMOVNTPD         mem512,zmmreg            AVX512,FUTURE
       VMOVNTPS         mem512,zmmreg            AVX512,FUTURE
       VMOVQ            xmmreg,rm64              AVX512,FUTURE
       VMOVQ            rm64,xmmreg              AVX512,FUTURE
       VMOVQ            xmmreg,xmmrm64           AVX512,FUTURE
       VMOVQ            xmmrm64,xmmreg           AVX512,FUTURE
       VMOVSD           xmmreg|mask|z,xmmreg,xmmreg AVX512,FUTURE
       VMOVSD           xmmreg|mask|z,mem64      AVX512,FUTURE
       VMOVSD           mem64|mask,xmmreg        AVX512,FUTURE
       VMOVSD           xmmreg|mask|z,xmmreg,xmmreg AVX512,FUTURE
       VMOVSHDUP        zmmreg|mask|z,zmmrm512   AVX512,FUTURE
       VMOVSLDUP        zmmreg|mask|z,zmmrm512   AVX512,FUTURE
       VMOVSS           xmmreg|mask|z,mem32      AVX512,FUTURE
       VMOVSS           xmmreg|mask|z,xmmreg,xmmreg AVX512,FUTURE
       VMOVSS           mem32|mask,xmmreg        AVX512,FUTURE
       VMOVSS           xmmreg|mask|z,xmmreg,xmmreg AVX512,FUTURE
       VMOVUPD          zmmreg|mask|z,zmmrm512   AVX512,FUTURE
       VMOVUPD          mem512|mask,zmmreg       AVX512,FUTURE
       VMOVUPD          zmmreg|mask|z,zmmreg     AVX512,FUTURE
       VMOVUPS          zmmreg|mask|z,zmmrm512   AVX512,FUTURE
       VMOVUPS          zmmreg|mask|z,zmmreg     AVX512,FUTURE
       VMOVUPS          mem512|mask,zmmreg       AVX512,FUTURE
       VMULPD           zmmreg|mask|z,zmmreg,zmmrm512|b64|er AVX512,FUTURE
       VMULPS           zmmreg|mask|z,zmmreg,zmmrm512|b32|er AVX512,FUTURE
       VMULSD           xmmreg|mask|z,xmmreg,xmmrm64|er AVX512,FUTURE
       VMULSS           xmmreg|mask|z,xmmreg,xmmrm32|er AVX512,FUTURE
       VPABSD           zmmreg|mask|z,zmmrm512|b32 AVX512,FUTURE
       VPABSQ           zmmreg|mask|z,zmmrm512|b64 AVX512,FUTURE
       VPADDD           zmmreg|mask|z,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPADDQ           zmmreg|mask|z,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPANDD           zmmreg|mask|z,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPANDND          zmmreg|mask|z,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPANDNQ          zmmreg|mask|z,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPANDQ           zmmreg|mask|z,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPBLENDMD        zmmreg|mask|z,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPBLENDMQ        zmmreg|mask|z,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPBROADCASTD     zmmreg|mask|z,xmmreg     AVX512,FUTURE
       VPBROADCASTD     zmmreg|mask|z,mem32      AVX512,FUTURE
       VPBROADCASTD     zmmreg|mask|z,reg32      AVX512,FUTURE
       VPBROADCASTQ     zmmreg|mask|z,xmmreg     AVX512,FUTURE
       VPBROADCASTQ     zmmreg|mask|z,mem64      AVX512,FUTURE
       VPBROADCASTQ     zmmreg|mask|z,reg64      AVX512,FUTURE
       VPCMPLTD         kreg|mask,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPCMPLED         kreg|mask,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPCMPNEQD        kreg|mask,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPCMPNLTD        kreg|mask,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPCMPNLED        kreg|mask,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPCMPD           kreg|mask,zmmreg,zmmrm512|b32,imm8 AVX512,FUTURE
       VPCMPEQD         kreg|mask,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPCMPEQQ         kreg|mask,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPCMPGTD         kreg|mask,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPCMPGTQ         kreg|mask,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPCMPLTQ         kreg|mask,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPCMPLEQ         kreg|mask,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPCMPNEQQ        kreg|mask,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPCMPNLTQ        kreg|mask,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPCMPNLEQ        kreg|mask,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPCMPQ           kreg|mask,zmmreg,zmmrm512|b64,imm8 AVX512,FUTURE
       VPCMPEQUD        kreg|mask,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPCMPLTUD        kreg|mask,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPCMPLEUD        kreg|mask,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPCMPNEQUD       kreg|mask,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPCMPNLTUD       kreg|mask,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPCMPNLEUD       kreg|mask,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPCMPUD          kreg|mask,zmmreg,zmmrm512|b32,imm8 AVX512,FUTURE
       VPCMPEQUQ        kreg|mask,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPCMPLTUQ        kreg|mask,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPCMPLEUQ        kreg|mask,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPCMPNEQUQ       kreg|mask,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPCMPNLTUQ       kreg|mask,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPCMPNLEUQ       kreg|mask,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPCMPUQ          kreg|mask,zmmreg,zmmrm512|b64,imm8 AVX512,FUTURE
       VPCOMPRESSD      zmmreg|mask|z,zmmreg     AVX512,FUTURE
       VPCOMPRESSD      mem512|mask,zmmreg       AVX512,FUTURE
       VPCOMPRESSQ      zmmreg|mask|z,zmmreg     AVX512,FUTURE
       VPCOMPRESSQ      mem512|mask,zmmreg       AVX512,FUTURE
       VPERMD           zmmreg|mask|z,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPERMI2D         zmmreg|mask|z,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPERMI2PD        zmmreg|mask|z,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPERMI2PS        zmmreg|mask|z,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPERMI2Q         zmmreg|mask|z,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPERMILPD        zmmreg|mask|z,zmmrm512|b64,imm8 AVX512,FUTURE
       VPERMILPD        zmmreg|mask|z,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPERMILPS        zmmreg|mask|z,zmmrm512|b32,imm8 AVX512,FUTURE
       VPERMILPS        zmmreg|mask|z,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPERMPD          zmmreg|mask|z,zmmrm512|b64,imm8 AVX512,FUTURE
       VPERMPD          zmmreg|mask|z,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPERMPS          zmmreg|mask|z,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPERMQ           zmmreg|mask|z,zmmrm512|b64,imm8 AVX512,FUTURE
       VPERMQ           zmmreg|mask|z,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPERMT2D         zmmreg|mask|z,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPERMT2PD        zmmreg|mask|z,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPERMT2PS        zmmreg|mask|z,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPERMT2Q         zmmreg|mask|z,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPEXPANDD        zmmreg|mask|z,zmmreg     AVX512,FUTURE
       VPEXPANDD        zmmreg|mask|z,mem512     AVX512,FUTURE
       VPEXPANDQ        zmmreg|mask|z,mem512     AVX512,FUTURE
       VPEXPANDQ        zmmreg|mask|z,zmmreg     AVX512,FUTURE
       VPGATHERDD       zmmreg|mask,zmem32       AVX512,FUTURE
       VPGATHERDQ       zmmreg|mask,ymem64       AVX512,FUTURE
       VPGATHERQD       ymmreg|mask,zmem32       AVX512,FUTURE
       VPGATHERQQ       zmmreg|mask,zmem64       AVX512,FUTURE
       VPMAXSD          zmmreg|mask|z,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPMAXSQ          zmmreg|mask|z,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPMAXUD          zmmreg|mask|z,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPMAXUQ          zmmreg|mask|z,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPMINSD          zmmreg|mask|z,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPMINSQ          zmmreg|mask|z,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPMINUD          zmmreg|mask|z,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPMINUQ          zmmreg|mask|z,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPMOVDB          xmmreg|mask|z,zmmreg     AVX512,FUTURE
       VPMOVDB          mem128|mask,zmmreg       AVX512,FUTURE
       VPMOVDW          ymmreg|mask|z,zmmreg     AVX512,FUTURE
       VPMOVDW          mem256|mask,zmmreg       AVX512,FUTURE
       VPMOVQB          xmmreg|mask|z,zmmreg     AVX512,FUTURE
       VPMOVQB          mem64|mask,zmmreg        AVX512,FUTURE
       VPMOVQD          ymmreg|mask|z,zmmreg     AVX512,FUTURE
       VPMOVQD          mem256|mask,zmmreg       AVX512,FUTURE
       VPMOVQW          xmmreg|mask|z,zmmreg     AVX512,FUTURE
       VPMOVQW          mem128|mask,zmmreg       AVX512,FUTURE
       VPMOVSDB         mem128|mask,zmmreg       AVX512,FUTURE
       VPMOVSDB         xmmreg|mask|z,zmmreg     AVX512,FUTURE
       VPMOVSDW         mem256|mask,zmmreg       AVX512,FUTURE
       VPMOVSDW         ymmreg|mask|z,zmmreg     AVX512,FUTURE
       VPMOVSQB         mem64|mask,zmmreg        AVX512,FUTURE
       VPMOVSQB         xmmreg|mask|z,zmmreg     AVX512,FUTURE
       VPMOVSQD         mem256|mask,zmmreg       AVX512,FUTURE
       VPMOVSQD         ymmreg|mask|z,zmmreg     AVX512,FUTURE
       VPMOVSQW         xmmreg|mask|z,zmmreg     AVX512,FUTURE
       VPMOVSQW         mem128|mask,zmmreg       AVX512,FUTURE
       VPMOVSXBD        zmmreg|mask|z,xmmrm128   AVX512,FUTURE
       VPMOVSXBQ        zmmreg|mask|z,xmmrm64    AVX512,FUTURE
       VPMOVSXDQ        zmmreg|mask|z,ymmrm256   AVX512,FUTURE
       VPMOVSXWD        zmmreg|mask|z,ymmrm256   AVX512,FUTURE
       VPMOVSXWQ        zmmreg|mask|z,xmmrm128   AVX512,FUTURE
       VPMOVUSDB        xmmreg|mask|z,zmmreg     AVX512,FUTURE
       VPMOVUSDB        mem128|mask,zmmreg       AVX512,FUTURE
       VPMOVUSDW        ymmreg|mask|z,zmmreg     AVX512,FUTURE
       VPMOVUSDW        mem256|mask,zmmreg       AVX512,FUTURE
       VPMOVUSQB        xmmreg|mask|z,zmmreg     AVX512,FUTURE
       VPMOVUSQB        mem64|mask,zmmreg        AVX512,FUTURE
       VPMOVUSQD        ymmreg|mask|z,zmmreg     AVX512,FUTURE
       VPMOVUSQD        mem256|mask,zmmreg       AVX512,FUTURE
       VPMOVUSQW        xmmreg|mask|z,zmmreg     AVX512,FUTURE
       VPMOVUSQW        mem128|mask,zmmreg       AVX512,FUTURE
       VPMOVZXBD        zmmreg|mask|z,xmmrm128   AVX512,FUTURE
       VPMOVZXBQ        zmmreg|mask|z,xmmrm64    AVX512,FUTURE
       VPMOVZXDQ        zmmreg|mask|z,ymmrm256   AVX512,FUTURE
       VPMOVZXWD        zmmreg|mask|z,ymmrm256   AVX512,FUTURE
       VPMOVZXWQ        zmmreg|mask|z,xmmrm128   AVX512,FUTURE
       VPMULDQ          zmmreg|mask|z,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPMULLD          zmmreg|mask|z,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPMULUDQ         zmmreg|mask|z,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPORD            zmmreg|mask|z,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPORQ            zmmreg|mask|z,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPROLD           zmmreg|mask|z,zmmrm512|b32,imm8 AVX512,FUTURE
       VPROLQ           zmmreg|mask|z,zmmrm512|b64,imm8 AVX512,FUTURE
       VPROLVD          zmmreg|mask|z,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPROLVQ          zmmreg|mask|z,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPRORD           zmmreg|mask|z,zmmrm512|b32,imm8 AVX512,FUTURE
       VPRORQ           zmmreg|mask|z,zmmrm512|b64,imm8 AVX512,FUTURE
       VPRORVD          zmmreg|mask|z,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPRORVQ          zmmreg|mask|z,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPSCATTERDD      zmem32|mask,zmmreg       AVX512,FUTURE
       VPSCATTERDQ      ymem64|mask,zmmreg       AVX512,FUTURE
       VPSCATTERQD      zmem32|mask,ymmreg       AVX512,FUTURE
       VPSCATTERQQ      zmem64|mask,zmmreg       AVX512,FUTURE
       VPSHUFD          zmmreg|mask|z,zmmrm512|b32,imm8 AVX512,FUTURE
       VPSLLD           zmmreg|mask|z,zmmrm512|b32,imm8 AVX512,FUTURE
       VPSLLD           zmmreg|mask|z,zmmreg,xmmrm128 AVX512,FUTURE
       VPSLLQ           zmmreg|mask|z,zmmrm512|b64,imm8 AVX512,FUTURE
       VPSLLQ           zmmreg|mask|z,zmmreg,xmmrm128 AVX512,FUTURE
       VPSLLVD          zmmreg|mask|z,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPSLLVQ          zmmreg|mask|z,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPSRAD           zmmreg|mask|z,zmmrm512|b32,imm8 AVX512,FUTURE
       VPSRAD           zmmreg|mask|z,zmmreg,xmmrm128 AVX512,FUTURE
       VPSRAQ           zmmreg|mask|z,zmmrm512|b64,imm8 AVX512,FUTURE
       VPSRAQ           zmmreg|mask|z,zmmreg,xmmrm128 AVX512,FUTURE
       VPSRAVD          zmmreg|mask|z,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPSRAVQ          zmmreg|mask|z,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPSRLD           zmmreg|mask|z,zmmrm512|b32,imm8 AVX512,FUTURE
       VPSRLD           zmmreg|mask|z,zmmreg,xmmrm128 AVX512,FUTURE
       VPSRLQ           zmmreg|mask|z,zmmrm512|b64,imm8 AVX512,FUTURE
       VPSRLQ           zmmreg|mask|z,zmmreg,xmmrm128 AVX512,FUTURE
       VPSRLVD          zmmreg|mask|z,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPSRLVQ          zmmreg|mask|z,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPSUBD           zmmreg|mask|z,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPSUBQ           zmmreg|mask|z,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPTERNLOGD       zmmreg|mask|z,zmmreg,zmmrm512|b32,imm8 AVX512,FUTURE
       VPTERNLOGQ       zmmreg|mask|z,zmmreg,zmmrm512|b64,imm8 AVX512,FUTURE
       VPTESTMD         kreg|mask,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPTESTMQ         kreg|mask,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPTESTNMD        kreg|mask,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPTESTNMQ        kreg|mask,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPUNPCKHDQ       zmmreg|mask|z,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPUNPCKHQDQ      zmmreg|mask|z,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPUNPCKLDQ       zmmreg|mask|z,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPUNPCKLQDQ      zmmreg|mask|z,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VPXORD           zmmreg|mask|z,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VPXORQ           zmmreg|mask|z,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VRCP14PD         zmmreg|mask|z,zmmrm512|b64 AVX512,FUTURE
       VRCP14PS         zmmreg|mask|z,zmmrm512|b32 AVX512,FUTURE
       VRCP14SD         xmmreg|mask|z,xmmreg,xmmrm64 AVX512,FUTURE
       VRCP14SS         xmmreg|mask|z,xmmreg,xmmrm32 AVX512,FUTURE
       VRNDSCALEPD      zmmreg|mask|z,zmmrm512|b64|sae,imm8 AVX512,FUTURE
       VRNDSCALEPS      zmmreg|mask|z,zmmrm512|b32|sae,imm8 AVX512,FUTURE
       VRNDSCALESD      xmmreg|mask|z,xmmreg,xmmrm64|sae,imm8 AVX512,FUTURE
       VRNDSCALESS      xmmreg|mask|z,xmmreg,xmmrm32|sae,imm8 AVX512,FUTURE
       VRSQRT14PD       zmmreg|mask|z,zmmrm512|b64 AVX512,FUTURE
       VRSQRT14PS       zmmreg|mask|z,zmmrm512|b32 AVX512,FUTURE
       VRSQRT14SD       xmmreg|mask|z,xmmreg,xmmrm64 AVX512,FUTURE
       VRSQRT14SS       xmmreg|mask|z,xmmreg,xmmrm32 AVX512,FUTURE
       VSCALEFPD        zmmreg|mask|z,zmmreg,zmmrm512|b64|er AVX512,FUTURE
       VSCALEFPS        zmmreg|mask|z,zmmreg,zmmrm512|b32|er AVX512,FUTURE
       VSCALEFSD        xmmreg|mask|z,xmmreg,xmmrm64|er AVX512,FUTURE
       VSCALEFSS        xmmreg|mask|z,xmmreg,xmmrm32|er AVX512,FUTURE
       VSCATTERDPD      ymem64|mask,zmmreg       AVX512,FUTURE
       VSCATTERDPS      zmem32|mask,zmmreg       AVX512,FUTURE
       VSCATTERQPD      zmem64|mask,zmmreg       AVX512,FUTURE
       VSCATTERQPS      zmem32|mask,ymmreg       AVX512,FUTURE
       VSHUFF32X4       zmmreg|mask|z,zmmreg,zmmrm512|b32,imm8 AVX512,FUTURE
       VSHUFF64X2       zmmreg|mask|z,zmmreg,zmmrm512|b64,imm8 AVX512,FUTURE
       VSHUFI32X4       zmmreg|mask|z,zmmreg,zmmrm512|b32,imm8 AVX512,FUTURE
       VSHUFI64X2       zmmreg|mask|z,zmmreg,zmmrm512|b64,imm8 AVX512,FUTURE
       VSHUFPD          zmmreg|mask|z,zmmreg,zmmrm512|b64,imm8 AVX512,FUTURE
       VSHUFPS          zmmreg|mask|z,zmmreg,zmmrm512|b32,imm8 AVX512,FUTURE
       VSQRTPD          zmmreg|mask|z,zmmrm512|b64|er AVX512,FUTURE
       VSQRTPS          zmmreg|mask|z,zmmrm512|b32|er AVX512,FUTURE
       VSQRTSD          xmmreg|mask|z,xmmreg,xmmrm64|er AVX512,FUTURE
       VSQRTSS          xmmreg|mask|z,xmmreg,xmmrm32|er AVX512,FUTURE
       VSUBPD           zmmreg|mask|z,zmmreg,zmmrm512|b64|er AVX512,FUTURE
       VSUBPS           zmmreg|mask|z,zmmreg,zmmrm512|b32|er AVX512,FUTURE
       VSUBSD           xmmreg|mask|z,xmmreg,xmmrm64|er AVX512,FUTURE
       VSUBSS           xmmreg|mask|z,xmmreg,xmmrm32|er AVX512,FUTURE
       VUCOMISD         xmmreg,xmmrm64|sae       AVX512,FUTURE
       VUCOMISS         xmmreg,xmmrm32|sae       AVX512,FUTURE
       VUNPCKHPD        zmmreg|mask|z,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VUNPCKHPS        zmmreg|mask|z,zmmreg,zmmrm512|b32 AVX512,FUTURE
       VUNPCKLPD        zmmreg|mask|z,zmmreg,zmmrm512|b64 AVX512,FUTURE
       VUNPCKLPS        zmmreg|mask|z,zmmreg,zmmrm512|b32 AVX512,FUTURE
       KANDNW           kreg,kreg,kreg           AVX512,FUTURE
       KANDW            kreg,kreg,kreg           AVX512,FUTURE
       KMOVW            kreg,krm16               AVX512,FUTURE
       KMOVW            kreg,reg32               AVX512,FUTURE
       KMOVW            mem16,kreg               AVX512,FUTURE
       KMOVW            reg32,kreg               AVX512,FUTURE
       KNOTW            kreg,kreg                AVX512,FUTURE
       KORTESTW         kreg,kreg                AVX512,FUTURE
       KORW             kreg,kreg,kreg           AVX512,FUTURE
       KSHIFTLW         kreg,kreg,imm8           AVX512,FUTURE
       KSHIFTRW         kreg,kreg,imm8           AVX512,FUTURE
       KUNPCKBW         kreg,kreg,kreg           AVX512,FUTURE
       KXNORW           kreg,kreg,kreg           AVX512,FUTURE
       KXORW            kreg,kreg,kreg           AVX512,FUTURE
       VPBROADCASTMB2Q  zmmreg,kreg              AVX512CD,FUTURE
       VPBROADCASTMW2D  zmmreg,kreg              AVX512CD,FUTURE
       VPCONFLICTD      zmmreg|mask|z,zmmrm512|b32 AVX512CD,FUTURE
       VPCONFLICTQ      zmmreg|mask|z,zmmrm512|b64 AVX512CD,FUTURE
       VPLZCNTD         zmmreg|mask|z,zmmrm512|b32 AVX512CD,FUTURE
       VPLZCNTQ         zmmreg|mask|z,zmmrm512|b64 AVX512CD,FUTURE
       VEXP2PD          zmmreg|mask|z,zmmrm512|b64|sae AVX512ER,FUTURE
       VEXP2PS          zmmreg|mask|z,zmmrm512|b32|sae AVX512ER,FUTURE
       VRCP28PD         zmmreg|mask|z,zmmrm512|b64|sae AVX512ER,FUTURE
       VRCP28PS         zmmreg|mask|z,zmmrm512|b32|sae AVX512ER,FUTURE
       VRCP28SD         xmmreg|mask|z,xmmreg,xmmrm64|sae AVX512ER,FUTURE
       VRCP28SS         xmmreg|mask|z,xmmreg,xmmrm32|sae AVX512ER,FUTURE
       VRSQRT28PD       zmmreg|mask|z,zmmrm512|b64|sae AVX512ER,FUTURE
       VRSQRT28PS       zmmreg|mask|z,zmmrm512|b32|sae AVX512ER,FUTURE
       VRSQRT28SD       xmmreg|mask|z,xmmreg,xmmrm64|sae AVX512ER,FUTURE
       VRSQRT28SS       xmmreg|mask|z,xmmreg,xmmrm32|sae AVX512ER,FUTURE
       VGATHERPF0DPD    ymem64|mask              AVX512PF,FUTURE
       VGATHERPF0DPS    zmem32|mask              AVX512PF,FUTURE
       VGATHERPF0QPD    zmem64|mask              AVX512PF,FUTURE
       VGATHERPF0QPS    zmem32|mask              AVX512PF,FUTURE
       VGATHERPF1DPD    ymem64|mask              AVX512PF,FUTURE
       VGATHERPF1DPS    zmem32|mask              AVX512PF,FUTURE
       VGATHERPF1QPD    zmem64|mask              AVX512PF,FUTURE
       VGATHERPF1QPS    zmem32|mask              AVX512PF,FUTURE
       VSCATTERPF0DPD   ymem64|mask              AVX512PF,FUTURE
       VSCATTERPF0DPS   zmem32|mask              AVX512PF,FUTURE
       VSCATTERPF0QPD   zmem64|mask              AVX512PF,FUTURE
       VSCATTERPF0QPS   zmem32|mask              AVX512PF,FUTURE
       VSCATTERPF1DPD   ymem64|mask              AVX512PF,FUTURE
       VSCATTERPF1DPS   zmem32|mask              AVX512PF,FUTURE
       VSCATTERPF1QPD   zmem64|mask              AVX512PF,FUTURE
       VSCATTERPF1QPS   zmem32|mask              AVX512PF,FUTURE
       PREFETCHWT1      mem8                     PREFETCHWT1,FUTURE
       BNDMK            bndreg,mem               MPX,MIB,FUTURE
       BNDCL            bndreg,mem               MPX,FUTURE
       BNDCL            bndreg,reg32             MPX,NOLONG,FUTURE
       BNDCL            bndreg,reg64             MPX,LONG,FUTURE
       BNDCU            bndreg,mem               MPX,FUTURE
       BNDCU            bndreg,reg32             MPX,NOLONG,FUTURE
       BNDCU            bndreg,reg64             MPX,LONG,FUTURE
       BNDCN            bndreg,mem               MPX,FUTURE
       BNDCN            bndreg,reg32             MPX,NOLONG,FUTURE
       BNDCN            bndreg,reg64             MPX,LONG,FUTURE
       BNDMOV           bndreg,bndreg            MPX,FUTURE
       BNDMOV           bndreg,mem               MPX,FUTURE
       BNDMOV           bndreg,bndreg            MPX,FUTURE
       BNDMOV           mem,bndreg               MPX,FUTURE
       BNDLDX           bndreg,mem               MPX,MIB,FUTURE
       BNDLDX           bndreg,mem,reg32         MPX,MIB,NOLONG,FUTURE
       BNDLDX           bndreg,mem,reg64         MPX,MIB,LONG,FUTURE
       BNDSTX           mem,bndreg               MPX,MIB,FUTURE
       BNDSTX           mem,reg32,bndreg         MPX,MIB,NOLONG,FUTURE
       BNDSTX           mem,reg64,bndreg         MPX,MIB,LONG,FUTURE
       BNDSTX           mem,bndreg,reg32         MPX,MIB,NOLONG,FUTURE
       BNDSTX           mem,bndreg,reg64         MPX,MIB,LONG,FUTURE
       SHA1RNDS4        xmmreg,xmmrm128,imm8     SHA,FUTURE
       SHA1NEXTE        xmmreg,xmmrm128          SHA,FUTURE
       SHA1MSG1         xmmreg,xmmrm128          SHA,FUTURE
       SHA1MSG2         xmmreg,xmmrm128          SHA,FUTURE
       SHA256RNDS2      xmmreg,xmmrm128,xmm0     SHA,FUTURE
       SHA256MSG1       xmmreg,xmmrm128          SHA,FUTURE
       SHA256MSG2       xmmreg,xmmrm128          SHA,FUTURE

B.1.37 Systematic names for the hinting nop instructions

       HINT_NOP0        rm16                     P6,UNDOC
       HINT_NOP0        rm32                     P6,UNDOC
       HINT_NOP0        rm64                     X64,UNDOC
       HINT_NOP1        rm16                     P6,UNDOC
       HINT_NOP1        rm32                     P6,UNDOC
       HINT_NOP1        rm64                     X64,UNDOC
       HINT_NOP2        rm16                     P6,UNDOC
       HINT_NOP2        rm32                     P6,UNDOC
       HINT_NOP2        rm64                     X64,UNDOC
       HINT_NOP3        rm16                     P6,UNDOC
       HINT_NOP3        rm32                     P6,UNDOC
       HINT_NOP3        rm64                     X64,UNDOC
       HINT_NOP4        rm16                     P6,UNDOC
       HINT_NOP4        rm32                     P6,UNDOC
       HINT_NOP4        rm64                     X64,UNDOC
       HINT_NOP5        rm16                     P6,UNDOC
       HINT_NOP5        rm32                     P6,UNDOC
       HINT_NOP5        rm64                     X64,UNDOC
       HINT_NOP6        rm16                     P6,UNDOC
       HINT_NOP6        rm32                     P6,UNDOC
       HINT_NOP6        rm64                     X64,UNDOC
       HINT_NOP7        rm16                     P6,UNDOC
       HINT_NOP7        rm32                     P6,UNDOC
       HINT_NOP7        rm64                     X64,UNDOC
       HINT_NOP8        rm16                     P6,UNDOC
       HINT_NOP8        rm32                     P6,UNDOC
       HINT_NOP8        rm64                     X64,UNDOC
       HINT_NOP9        rm16                     P6,UNDOC
       HINT_NOP9        rm32                     P6,UNDOC
       HINT_NOP9        rm64                     X64,UNDOC
       HINT_NOP10       rm16                     P6,UNDOC
       HINT_NOP10       rm32                     P6,UNDOC
       HINT_NOP10       rm64                     X64,UNDOC
       HINT_NOP11       rm16                     P6,UNDOC
       HINT_NOP11       rm32                     P6,UNDOC
       HINT_NOP11       rm64                     X64,UNDOC
       HINT_NOP12       rm16                     P6,UNDOC
       HINT_NOP12       rm32                     P6,UNDOC
       HINT_NOP12       rm64                     X64,UNDOC
       HINT_NOP13       rm16                     P6,UNDOC
       HINT_NOP13       rm32                     P6,UNDOC
       HINT_NOP13       rm64                     X64,UNDOC
       HINT_NOP14       rm16                     P6,UNDOC
       HINT_NOP14       rm32                     P6,UNDOC
       HINT_NOP14       rm64                     X64,UNDOC
       HINT_NOP15       rm16                     P6,UNDOC
       HINT_NOP15       rm32                     P6,UNDOC
       HINT_NOP15       rm64                     X64,UNDOC
       HINT_NOP16       rm16                     P6,UNDOC
       HINT_NOP16       rm32                     P6,UNDOC
       HINT_NOP16       rm64                     X64,UNDOC
       HINT_NOP17       rm16                     P6,UNDOC
       HINT_NOP17       rm32                     P6,UNDOC
       HINT_NOP17       rm64                     X64,UNDOC
       HINT_NOP18       rm16                     P6,UNDOC
       HINT_NOP18       rm32                     P6,UNDOC
       HINT_NOP18       rm64                     X64,UNDOC
       HINT_NOP19       rm16                     P6,UNDOC
       HINT_NOP19       rm32                     P6,UNDOC
       HINT_NOP19       rm64                     X64,UNDOC
       HINT_NOP20       rm16                     P6,UNDOC
       HINT_NOP20       rm32                     P6,UNDOC
       HINT_NOP20       rm64                     X64,UNDOC
       HINT_NOP21       rm16                     P6,UNDOC
       HINT_NOP21       rm32                     P6,UNDOC
       HINT_NOP21       rm64                     X64,UNDOC
       HINT_NOP22       rm16                     P6,UNDOC
       HINT_NOP22       rm32                     P6,UNDOC
       HINT_NOP22       rm64                     X64,UNDOC
       HINT_NOP23       rm16                     P6,UNDOC
       HINT_NOP23       rm32                     P6,UNDOC
       HINT_NOP23       rm64                     X64,UNDOC
       HINT_NOP24       rm16                     P6,UNDOC
       HINT_NOP24       rm32                     P6,UNDOC
       HINT_NOP24       rm64                     X64,UNDOC
       HINT_NOP25       rm16                     P6,UNDOC
       HINT_NOP25       rm32                     P6,UNDOC
       HINT_NOP25       rm64                     X64,UNDOC
       HINT_NOP26       rm16                     P6,UNDOC
       HINT_NOP26       rm32                     P6,UNDOC
       HINT_NOP26       rm64                     X64,UNDOC
       HINT_NOP27       rm16                     P6,UNDOC
       HINT_NOP27       rm32                     P6,UNDOC
       HINT_NOP27       rm64                     X64,UNDOC
       HINT_NOP28       rm16                     P6,UNDOC
       HINT_NOP28       rm32                     P6,UNDOC
       HINT_NOP28       rm64                     X64,UNDOC
       HINT_NOP29       rm16                     P6,UNDOC
       HINT_NOP29       rm32                     P6,UNDOC
       HINT_NOP29       rm64                     X64,UNDOC
       HINT_NOP30       rm16                     P6,UNDOC
       HINT_NOP30       rm32                     P6,UNDOC
       HINT_NOP30       rm64                     X64,UNDOC
       HINT_NOP31       rm16                     P6,UNDOC
       HINT_NOP31       rm32                     P6,UNDOC
       HINT_NOP31       rm64                     X64,UNDOC
       HINT_NOP32       rm16                     P6,UNDOC
       HINT_NOP32       rm32                     P6,UNDOC
       HINT_NOP32       rm64                     X64,UNDOC
       HINT_NOP33       rm16                     P6,UNDOC
       HINT_NOP33       rm32                     P6,UNDOC
       HINT_NOP33       rm64                     X64,UNDOC
       HINT_NOP34       rm16                     P6,UNDOC
       HINT_NOP34       rm32                     P6,UNDOC
       HINT_NOP34       rm64                     X64,UNDOC
       HINT_NOP35       rm16                     P6,UNDOC
       HINT_NOP35       rm32                     P6,UNDOC
       HINT_NOP35       rm64                     X64,UNDOC
       HINT_NOP36       rm16                     P6,UNDOC
       HINT_NOP36       rm32                     P6,UNDOC
       HINT_NOP36       rm64                     X64,UNDOC
       HINT_NOP37       rm16                     P6,UNDOC
       HINT_NOP37       rm32                     P6,UNDOC
       HINT_NOP37       rm64                     X64,UNDOC
       HINT_NOP38       rm16                     P6,UNDOC
       HINT_NOP38       rm32                     P6,UNDOC
       HINT_NOP38       rm64                     X64,UNDOC
       HINT_NOP39       rm16                     P6,UNDOC
       HINT_NOP39       rm32                     P6,UNDOC
       HINT_NOP39       rm64                     X64,UNDOC
       HINT_NOP40       rm16                     P6,UNDOC
       HINT_NOP40       rm32                     P6,UNDOC
       HINT_NOP40       rm64                     X64,UNDOC
       HINT_NOP41       rm16                     P6,UNDOC
       HINT_NOP41       rm32                     P6,UNDOC
       HINT_NOP41       rm64                     X64,UNDOC
       HINT_NOP42       rm16                     P6,UNDOC
       HINT_NOP42       rm32                     P6,UNDOC
       HINT_NOP42       rm64                     X64,UNDOC
       HINT_NOP43       rm16                     P6,UNDOC
       HINT_NOP43       rm32                     P6,UNDOC
       HINT_NOP43       rm64                     X64,UNDOC
       HINT_NOP44       rm16                     P6,UNDOC
       HINT_NOP44       rm32                     P6,UNDOC
       HINT_NOP44       rm64                     X64,UNDOC
       HINT_NOP45       rm16                     P6,UNDOC
       HINT_NOP45       rm32                     P6,UNDOC
       HINT_NOP45       rm64                     X64,UNDOC
       HINT_NOP46       rm16                     P6,UNDOC
       HINT_NOP46       rm32                     P6,UNDOC
       HINT_NOP46       rm64                     X64,UNDOC
       HINT_NOP47       rm16                     P6,UNDOC
       HINT_NOP47       rm32                     P6,UNDOC
       HINT_NOP47       rm64                     X64,UNDOC
       HINT_NOP48       rm16                     P6,UNDOC
       HINT_NOP48       rm32                     P6,UNDOC
       HINT_NOP48       rm64                     X64,UNDOC
       HINT_NOP49       rm16                     P6,UNDOC
       HINT_NOP49       rm32                     P6,UNDOC
       HINT_NOP49       rm64                     X64,UNDOC
       HINT_NOP50       rm16                     P6,UNDOC
       HINT_NOP50       rm32                     P6,UNDOC
       HINT_NOP50       rm64                     X64,UNDOC
       HINT_NOP51       rm16                     P6,UNDOC
       HINT_NOP51       rm32                     P6,UNDOC
       HINT_NOP51       rm64                     X64,UNDOC
       HINT_NOP52       rm16                     P6,UNDOC
       HINT_NOP52       rm32                     P6,UNDOC
       HINT_NOP52       rm64                     X64,UNDOC
       HINT_NOP53       rm16                     P6,UNDOC
       HINT_NOP53       rm32                     P6,UNDOC
       HINT_NOP53       rm64                     X64,UNDOC
       HINT_NOP54       rm16                     P6,UNDOC
       HINT_NOP54       rm32                     P6,UNDOC
       HINT_NOP54       rm64                     X64,UNDOC
       HINT_NOP55       rm16                     P6,UNDOC
       HINT_NOP55       rm32                     P6,UNDOC
       HINT_NOP55       rm64                     X64,UNDOC
       HINT_NOP56       rm16                     P6,UNDOC
       HINT_NOP56       rm32                     P6,UNDOC
       HINT_NOP56       rm64                     X64,UNDOC
       HINT_NOP57       rm16                     P6,UNDOC
       HINT_NOP57       rm32                     P6,UNDOC
       HINT_NOP57       rm64                     X64,UNDOC
       HINT_NOP58       rm16                     P6,UNDOC
       HINT_NOP58       rm32                     P6,UNDOC
       HINT_NOP58       rm64                     X64,UNDOC
       HINT_NOP59       rm16                     P6,UNDOC
       HINT_NOP59       rm32                     P6,UNDOC
       HINT_NOP59       rm64                     X64,UNDOC
       HINT_NOP60       rm16                     P6,UNDOC
       HINT_NOP60       rm32                     P6,UNDOC
       HINT_NOP60       rm64                     X64,UNDOC
       HINT_NOP61       rm16                     P6,UNDOC
       HINT_NOP61       rm32                     P6,UNDOC
       HINT_NOP61       rm64                     X64,UNDOC
       HINT_NOP62       rm16                     P6,UNDOC
       HINT_NOP62       rm32                     P6,UNDOC
       HINT_NOP62       rm64                     X64,UNDOC
       HINT_NOP63       rm16                     P6,UNDOC
       HINT_NOP63       rm32                     P6,UNDOC
       HINT_NOP63       rm64                     X64,UNDOC

Appendix C: NASM Version History
--------------------------------

   C.1 NASM 2 Series

       The NASM 2 series supports x86-64, and is the production version of
       NASM since 2007.

C.1.1 Version 2.11

       (*) Add support for the Intel AVX-512 instruction set:

       (*) 16 new, 512-bit SIMD registers. Total 32 `(ZMM0 ~ ZMM31)'

       (*) 8 new opmask registers `(K0 ~ K7)'. One of 7 registers
           `(K1 ~ K7)' can be used as an opmask for conditional execution.

       (*) A new EVEX encoding prefix. EVEX is based on VEX and provides
           more capabilities: opmasks, broadcasting, embedded rounding and
           compressed displacements.

        - opmask
            VDIVPD zmm0{k1}{z}, zmm1, zmm3  ; conditional vector operation
                                            ; using opmask k1.
                                            ; {z} is for zero-masking
        - broadcasting
            VDIVPS zmm4, zmm5, [rbx]{1to16} ; load single-precision float and
                                            ; replicate it 16 times. 32 * 16 = 512
        - embedded rounding
            VCVTSI2SD xmm6, xmm7, {rz-sae}, rax ; round toward zero. note that it
                                             ; is used as if a separate operand.
                                             ; it comes after the last SIMD operand

       (*) Add support for `ZWORD' (512 bits), `DZ' and `RESZ'.

       (*) Add support for the MPX and SHA instruction sets.

       (*) Better handling of section redefinition.

       (*) Generate manpages when running `'make dist''.

       (*) Handle all token chains in mmacro params range.

       (*) Support split [base,index] effective address:

            mov eax,[eax+8,ecx*4]   ; eax=base, ecx=index, 4=scale, 8=disp

       This is expected to be most useful for the MPX instructions.

       (*) Support `BND' prefix for branch instructions (for MPX).

       (*) The `DEFAULT' directive can now take `BND' and `NOBND' options
           to indicate whether all relevant branches should be getting
           `BND' prefixes. This is expected to be the normal for use in MPX
           code.

       (*) Add `{evex'}, `{vex3'} and `{vex2'} instruction prefixes to have
           NASM encode the corresponding instruction, if possible, with an
           EVEX, 3-byte VEX, or 2-byte VEX prefix, respectively.

       (*) Support for section names longer than 8 bytes in Win32/Win64
           COFF.

       (*) The `NOSPLIT' directive by itself no longer forces a single
           register to become an index register, unless it has an explicit
           multiplier.

            mov eax,[nosplit eax]       ; eax as base register
            mov eax,[nosplit eax*1]     ; eax as index register

C.1.2 Version 2.10.09

       (*) Pregenerate man pages.

C.1.3 Version 2.10.08

       (*) Fix `VMOVNTDQA', `MOVNTDQA' and `MOVLPD' instructions.

       (*) Fix collision for `VGATHERQPS', `VPGATHERQD' instructions.

       (*) Fix `VPMOVSXBQ', `VGATHERQPD', `VSPLLW' instructions.

       (*) Add a bunch of AMD TBM instructions.

       (*) Fix potential stack overwrite in numbers conversion.

       (*) Allow byte size in `PREFETCHTx' instructions.

       (*) Make manual pages up to date.

       (*) Make `F3' and `F2' SSE prefixes to override `66'.

       (*) Support of AMD SVM instructions in 32 bit mode.

       (*) Fix near offsets code generation for `JMP', `CALL' instrictions
           in long mode.

       (*) Fix preprocessor parse regression when id is expanding to a
           whitespace.

C.1.4 Version 2.10.07

       (*) Fix line continuation parsing being broken in previous version.

C.1.5 Version 2.10.06

       (*) Always quote the dependency source names when using the
           automatic dependency generation options.

       (*) If no dependency target name is specified via the `-MT' or `-MQ'
           options, quote the default output name.

       (*) Fix assembly of shift operations in `CPU 8086' mode.

       (*) Fix incorrect generation of explicit immediate byte for shift by
           1 under certain circumstances.

       (*) Fix assembly of the `VPCMPGTQ' instruction.

       (*) Fix RIP-relative relocations in the `macho64' backend.

C.1.6 Version 2.10.05

       (*) Add the `CLAC' and `STAC' instructions.

C.1.7 Version 2.10.04

       (*) Add back the inadvertently deleted 256-bit version of the
           `VORPD' instruction.

       (*) Correct disassembly of instructions starting with byte `82' hex.

       (*) Fix corner cases in token pasting, for example:

          %define N 1e%++%+ 5
                  dd N, 1e+5

C.1.8 Version 2.10.03

       (*) Correct the assembly of the instruction:

       XRELEASE MOV [absolute],AL

       Previous versions would incorrectly generate `F3 A2' for this
       instruction and issue a warning; correct behavior is to emit
       `F3 88 05'.

C.1.9 Version 2.10.02

       (*) Add the `ifunc' macro package with integer functions, currently
           only integer logarithms. See section 5.4.

       (*) Add the `RDSEED', `ADCX' and `ADOX' instructions.

C.1.10 Version 2.10.01

       (*) Add missing VPMOVMSKB instruction with reg32, ymmreg operands.

C.1.11 Version 2.10

       (*) When optimization is enabled, `mov r64,imm' now optimizes to the
           shortest form possible between:

            mov r32,imm32                   ;  5 bytes
            mov r64,imm32                   ;  7 bytes
            mov r64,imm64                   ; 10 bytes

       To force a specific form, use the `STRICT' keyword, see section 3.7.

       (*) Add support for the Intel AVX2 instruction set.

       (*) Add support for Bit Manipulation Instructions 1 and 2.

       (*) Add support for Intel Transactional Synchronization Extensions
           (TSX).

       (*) Add support for x32 ELF (32-bit ELF with the CPU in 64-bit
           mode.) See section 7.9.

       (*) Add support for bigendian UTF-16 and UTF-32. See section 3.4.5.

C.1.12 Version 2.09.10

       (*) Fix up NSIS script to protect uninstaller against registry keys
           absence or corruption. It brings in a few additional questions
           to a user during deinstallation procedure but still it is better
           than unpredictable file removal.

C.1.13 Version 2.09.09

       (*) Fix initialization of section attributes of `bin' output format.

       (*) Fix `mach64' output format bug that crashes NASM due to NULL
           symbols.

C.1.14 Version 2.09.08

       (*) Fix `__OUTPUT_FORMAT__' assignment when output driver alias is
           used. For example when `-f elf' is used `__OUTPUT_FORMAT__' must
           be set to `elf', if `-f elf32' is used `__OUTPUT_FORMAT__' must
           be assigned accordingly, i.e. to `elf32'. The rule applies to
           all output driver aliases. See section 4.12.6.

C.1.15 Version 2.09.07

       (*) Fix attempts to close same file several times when `-a' option
           is used.

       (*) Fixes for VEXTRACTF128, VMASKMOVPS encoding.

C.1.16 Version 2.09.06

       (*) Fix missed section attribute initialization in `bin' output
           target.

C.1.17 Version 2.09.05

       (*) Fix arguments encoding for VPEXTRW instruction.

       (*) Remove invalid form of VPEXTRW instruction.

       (*) Add `VLDDQU' as alias for `VLDQQU' to match specification.

C.1.18 Version 2.09.04

       (*) Fix incorrect labels offset for VEX intructions.

       (*) Eliminate bogus warning on implicit operand size override.

       (*) `%if' term could not handle 64 bit numbers.

       (*) The COFF backend was limiting relocations number to 16 bits even
           if in real there were a way more relocations.

C.1.19 Version 2.09.03

       (*) Print `%macro' name inside `%rep' blocks on error.

       (*) Fix preprocessor expansion behaviour. It happened sometime too
           early and sometime simply wrong. Move behaviour back to the
           origins (down to NASM 2.05.01).

       (*) Fix unitialized data dereference on OMF output format.

       (*) Issue warning on unterminated `%{' construct.

       (*) Fix for documentation typo.

C.1.20 Version 2.09.02

       (*) Fix reversed tokens when `%deftok' produces more than one output
           token.

       (*) Fix segmentation fault on disassembling some VEX instructions.

       (*) Missing `%endif' did not always cause error.

       (*) Fix typo in documentation.

       (*) Compound context local preprocessor single line macro
           identifiers were not expanded early enough and as result lead to
           unresolved symbols.

C.1.21 Version 2.09.01

       (*) Fix NULL dereference on missed %deftok second parameter.

       (*) Fix NULL dereference on invalid %substr parameters.

C.1.22 Version 2.09

       (*) Fixed assignment the magnitude of `%rep' counter. It is limited
           to 62 bits now.

       (*) Fixed NULL dereference if argument of `%strlen' resolves to
           whitespace. For example if nonexistent macro parameter is used.

       (*) `%ifenv', `%elifenv', `%ifnenv', and `%elifnenv' directives
           introduced. See section 4.4.9.

       (*) Fixed NULL dereference if environment variable is missed.

       (*) Updates of new AVX v7 Intel instructions.

       (*) `PUSH imm32' is now officially documented.

       (*) Fix for encoding the LFS, LGS and LSS in 64-bit mode.

       (*) Fixes for compatibility with OpenWatcom compiler and DOS 8.3
           file format limitation.

       (*) Macros parameters range expansion introduced. See section 4.3.4.

       (*) Backward compatibility on expanging of local sigle macros
           restored.

       (*) 8 bit relocations for `elf' and `bin' output formats are
           introduced.

       (*) Short intersegment jumps are permitted now.

       (*) An alignment more than 64 bytes are allowed for `win32', `win64'
           output formats.

       (*) `SECTALIGN' directive introduced. See section 4.12.13.

       (*) `nojmp' option introduced in `smartalign' package. See section
           5.2.

       (*) Short aliases `win', `elf' and `macho' for output formats are
           introduced. Each stands for `win32', `elf32' and `macho32'
           accordingly.

       (*) Faster handling of missing directives implemented.

       (*) Various small improvements in documentation.

       (*) No hang anymore if unable to open malloc.log file.

       (*) The environments without vsnprintf function are able to build
           nasm again.

       (*) AMD LWP instructions updated.

       (*) Tighten EA checks. We warn a user if there overflow in EA
           addressing.

       (*) Make `-Ox' the default optimization level. For the legacy
           behavior, specify `-O0' explicitly. See section 2.1.22.

       (*) Environment variables read with `%!' or tested with `%ifenv' can
           now contain non-identifier characters if surrounded by quotes.
           See section 4.10.2.

       (*) Add a new standard macro package `%use fp' for floating-point
           convenience macros. See section 5.3.

C.1.23 Version 2.08.02

       (*) Fix crash under certain circumstances when using the `%+'
           operator.

C.1.24 Version 2.08.01

       (*) Fix the `%use' statement, which was broken in 2.08.

C.1.25 Version 2.08

       (*) A number of enhancements/fixes in macros area.

       (*) Support for converting strings to tokens. See section 4.1.9.

       (*) Fuzzy operand size logic introduced.

       (*) Fix COFF stack overrun on too long export identifiers.

       (*) Fix Macho-O alignment bug.

       (*) Fix crashes with -fwin32 on file with many exports.

       (*) Fix stack overrun for too long [DEBUG id].

       (*) Fix incorrect sbyte usage in IMUL (hit only if optimization flag
           passed).

       (*) Append ending token for `.stabs' records in the ELF output
           format.

       (*) New NSIS script which uses ModernUI and MultiUser approach.

       (*) Visual Studio 2008 NASM integration (rules file).

       (*) Warn a user if a constant is too long (and as result will be
           stripped).

       (*) The obsoleted pre-XOP AMD SSE5 instruction set which was never
           actualized was removed.

       (*) Fix stack overrun on too long error file name passed from the
           command line.

       (*) Bind symbols to the .text section by default (ie in case if
           SECTION directive was omitted) in the ELF output format.

       (*) Fix sync points array index wrapping.

       (*) A few fixes for FMA4 and XOP instruction templates.

       (*) Add AMD Lightweight Profiling (LWP) instructions.

       (*) Fix the offset for `%arg' in 64-bit mode.

       (*) An undefined local macro (`%$') no longer matches a global macro
           with the same name.

       (*) Fix NULL dereference on too long local labels.

C.1.26 Version 2.07

       (*) NASM is now under the 2-clause BSD license. See section 1.1.2.

       (*) Fix the section type for the `.strtab' section in the `elf64'
           output format.

       (*) Fix the handling of `COMMON' directives in the `obj' output
           format.

       (*) New `ith' and `srec' output formats; these are variants of the
           `bin' output format which output Intel hex and Motorola S-
           records, respectively. See section 7.2 and section 7.3.

       (*) `rdf2ihx' replaced with an enhanced `rdf2bin', which can output
           binary, COM, Intel hex or Motorola S-records.

       (*) The Windows installer now puts the NASM directory first in the
           `PATH' of the "NASM Shell".

       (*) Revert the early expansion behavior of `%+' to pre-2.06
           behavior: `%+' is only expanded late.

       (*) Yet another Mach-O alignment fix.

       (*) Don't delete the list file on errors. Also, include error and
           warning information in the list file.

       (*) Support for 64-bit Mach-O output, see section 7.8.

       (*) Fix assert failure on certain operations that involve strings
           with high-bit bytes.

C.1.27 Version 2.06

       (*) This release is dedicated to the memory of Charles A. Crayne,
           long time NASM developer as well as moderator of
           `comp.lang.asm.x86' and author of the book _Serious Assembler_.
           We miss you, Chuck.

       (*) Support for indirect macro expansion (`%[...]'). See section
           4.1.3.

       (*) `%pop' can now take an argument, see section 4.7.1.

       (*) The argument to `%use' is no longer macro-expanded. Use `%[...]'
           if macro expansion is desired.

       (*) Support for thread-local storage in ELF32 and ELF64. See section
           7.9.4.

       (*) Fix crash on `%ifmacro' without an argument.

       (*) Correct the arguments to the `POPCNT' instruction.

       (*) Fix section alignment in the Mach-O format.

       (*) Update AVX support to version 5 of the Intel specification.

       (*) Fix the handling of accesses to context-local macros from higher
           levels in the context stack.

       (*) Treat `WAIT' as a prefix rather than as an instruction, thereby
           allowing constructs like `O16 FSAVE' to work correctly.

       (*) Support for structures with a non-zero base offset. See section
           4.12.10.

       (*) Correctly handle preprocessor token concatenation (see section
           4.3.9) involving floating-point numbers.

       (*) The `PINSR' series of instructions have been corrected and
           rationalized.

       (*) Removed AMD SSE5, replaced with the new XOP/FMA4/CVT16 (rev
           3.03) spec.

       (*) The ELF backends no longer automatically generate a `.comment'
           section.

       (*) Add additional "well-known" ELF sections with default
           attributes. See section 7.9.2.

C.1.28 Version 2.05.01

       (*) Fix the `-w'/`-W' option parsing, which was broken in NASM 2.05.

C.1.29 Version 2.05

       (*) Fix redundant REX.W prefix on `JMP reg64'.

       (*) Make the behaviour of `-O0' match NASM 0.98 legacy behavior. See
           section 2.1.22.

       (*) `-w-user' can be used to suppress the output of `%warning'
           directives. See section 2.1.24.

       (*) Fix bug where `ALIGN' would issue a full alignment datum instead
           of zero bytes.

       (*) Fix offsets in list files.

       (*) Fix `%include' inside multi-line macros or loops.

       (*) Fix error where NASM would generate a spurious warning on valid
           optimizations of immediate values.

       (*) Fix arguments to a number of the `CVT' SSE instructions.

       (*) Fix RIP-relative offsets when the instruction carries an
           immediate.

       (*) Massive overhaul of the ELF64 backend for spec compliance.

       (*) Fix the Geode `PFRCPV' and `PFRSQRTV' instruction.

       (*) Fix the SSE 4.2 `CRC32' instruction.

C.1.30 Version 2.04

       (*) Sanitize macro handing in the `%error' directive.

       (*) New `%warning' directive to issue user-controlled warnings.

       (*) `%error' directives are now deferred to the final assembly
           phase.

       (*) New `%fatal' directive to immediately terminate assembly.

       (*) New `%strcat' directive to join quoted strings together.

       (*) New `%use' macro directive to support standard macro directives.
           See section 4.6.4.

       (*) Excess default parameters to `%macro' now issues a warning by
           default. See section 4.3.

       (*) Fix `%ifn' and `%elifn'.

       (*) Fix nested `%else' clauses.

       (*) Correct the handling of nested `%rep's.

       (*) New `%unmacro' directive to undeclare a multi-line macro. See
           section 4.3.12.

       (*) Builtin macro `__PASS__' which expands to the current assembly
           pass. See section 4.12.9.

       (*) `__utf16__' and `__utf32__' operators to generate UTF-16 and
           UTF-32 strings. See section 3.4.5.

       (*) Fix bug in case-insensitive matching when compiled on platforms
           that don't use the `configure' script. Of the official release
           binaries, that only affected the OS/2 binary.

       (*) Support for x87 packed BCD constants. See section 3.4.7.

       (*) Correct the `LTR' and `SLDT' instructions in 64-bit mode.

       (*) Fix unnecessary REX.W prefix on indirect jumps in 64-bit mode.

       (*) Add AVX versions of the AES instructions (`VAES'...).

       (*) Fix the 256-bit FMA instructions.

       (*) Add 256-bit AVX stores per the latest AVX spec.

       (*) VIA XCRYPT instructions can now be written either with or
           without `REP', apparently different versions of the VIA spec
           wrote them differently.

       (*) Add missing 64-bit `MOVNTI' instruction.

       (*) Fix the operand size of `VMREAD' and `VMWRITE'.

       (*) Numerous bug fixes, especially to the AES, AVX and VTX
           instructions.

       (*) The optimizer now always runs until it converges. It also runs
           even when disabled, but doesn't optimize. This allows most
           forward references to be resolved properly.

       (*) `%push' no longer needs a context identifier; omitting the
           context identifier results in an anonymous context.

C.1.31 Version 2.03.01

       (*) Fix buffer overflow in the listing module.

       (*) Fix the handling of hexadecimal escape codes in `...` strings.

       (*) The Postscript/PDF documentation has been reformatted.

       (*) The `-F' option now implies `-g'.

C.1.32 Version 2.03

       (*) Add support for Intel AVX, CLMUL and FMA instructions, including
           YMM registers.

       (*) `dy', `resy' and `yword' for 32-byte operands.

       (*) Fix some SSE5 instructions.

       (*) Intel `INVEPT', `INVVPID' and `MOVBE' instructions.

       (*) Fix checking for critical expressions when the optimizer is
           enabled.

       (*) Support the DWARF debugging format for ELF targets.

       (*) Fix optimizations of signed bytes.

       (*) Fix operation on bigendian machines.

       (*) Fix buffer overflow in the preprocessor.

       (*) `SAFESEH' support for Win32, `IMAGEREL' for Win64 (SEH).

       (*) `%?' and `%??' to refer to the name of a macro itself. In
           particular, `%idefine keyword $%?' can be used to make a keyword
           "disappear".

       (*) New options for dependency generation: `-MD', `-MF', `-MP',
           `-MT', `-MQ'.

       (*) New preprocessor directives `%pathsearch' and `%depend'; INCBIN
           reimplemented as a macro.

       (*) `%include' now resolves macros in a sane manner.

       (*) `%substr' can now be used to get other than one-character
           substrings.

       (*) New type of character/string constants, using backquotes
           (``...`'), which support C-style escape sequences.

       (*) `%defstr' and `%idefstr' to stringize macro definitions before
           creation.

       (*) Fix forward references used in `EQU' statements.

C.1.33 Version 2.02

       (*) Additional fixes for MMX operands with explicit `qword', as well
           as (hopefully) SSE operands with `oword'.

       (*) Fix handling of truncated strings with `DO'.

       (*) Fix segfaults due to memory overwrites when floating-point
           constants were used.

       (*) Fix segfaults due to missing include files.

       (*) Fix OpenWatcom Makefiles for DOS and OS/2.

       (*) Add autogenerated instruction list back into the documentation.

       (*) ELF: Fix segfault when generating stabs, and no symbols have
           been defined.

       (*) ELF: Experimental support for DWARF debugging information.

       (*) New compile date and time standard macros.

       (*) `%ifnum' now returns true for negative numbers.

       (*) New `%iftoken' test for a single token.

       (*) New `%ifempty' test for empty expansion.

       (*) Add support for the `XSAVE' instruction group.

       (*) Makefile for Netware/gcc.

       (*) Fix issue with some warnings getting emitted way too many times.

       (*) Autogenerated instruction list added to the documentation.

C.1.34 Version 2.01

       (*) Fix the handling of MMX registers with explicit `qword' tags on
           memory (broken in 2.00 due to 64-bit changes.)

       (*) Fix the PREFETCH instructions.

       (*) Fix the documentation.

       (*) Fix debugging info when using `-f elf' (backwards compatibility
           alias for `-f elf32').

       (*) Man pages for rdoff tools (from the Debian project.)

       (*) ELF: handle large numbers of sections.

       (*) Fix corrupt output when the optimizer runs out of passes.

C.1.35 Version 2.00

       (*) Added c99 data-type compliance.

       (*) Added general x86-64 support.

       (*) Added win64 (x86-64 COFF) output format.

       (*) Added `__BITS__' standard macro.

       (*) Renamed the `elf' output format to `elf32' for clarity.

       (*) Added `elf64' and `macho' (MacOS X) output formats.

       (*) Added Numeric constants in `dq' directive.

       (*) Added `oword', `do' and `reso' pseudo operands.

       (*) Allow underscores in numbers.

       (*) Added 8-, 16- and 128-bit floating-point formats.

       (*) Added binary, octal and hexadecimal floating-point.

       (*) Correct the generation of floating-point constants.

       (*) Added floating-point option control.

       (*) Added Infinity and NaN floating point support.

       (*) Added ELF Symbol Visibility support.

       (*) Added setting OSABI value in ELF header directive.

       (*) Added Generate Makefile Dependencies option.

       (*) Added Unlimited Optimization Passes option.

       (*) Added `%IFN' and `%ELIFN' support.

       (*) Added Logical Negation Operator.

       (*) Enhanced Stack Relative Preprocessor Directives.

       (*) Enhanced ELF Debug Formats.

       (*) Enhanced Send Errors to a File option.

       (*) Added SSSE3, SSE4.1, SSE4.2, SSE5 support.

       (*) Added a large number of additional instructions.

       (*) Significant performance improvements.

       (*) `-w+warning' and `-w-warning' can now be written as -Wwarning
           and -Wno-warning, respectively. See section 2.1.24.

       (*) Add `-w+error' to treat warnings as errors. See section 2.1.24.

       (*) Add `-w+all' and `-w-all' to enable or disable all suppressible
           warnings. See section 2.1.24.

   C.2 NASM 0.98 Series

       The 0.98 series was the production versions of NASM from 1999 to
       2007.

C.2.1 Version 0.98.39

       (*) fix buffer overflow

       (*) fix outas86's `.bss' handling

       (*) "make spotless" no longer deletes config.h.in.

       (*) `%(el)if(n)idn' insensitivity to string quotes difference
           (#809300).

       (*) (nasm.c)`__OUTPUT_FORMAT__' changed to string value instead of
           symbol.

C.2.2 Version 0.98.38

       (*) Add Makefile for 16-bit DOS binaries under OpenWatcom, and
           modify `mkdep.pl' to be able to generate completely pathless
           dependencies, as required by OpenWatcom wmake (it supports path
           searches, but not explicit paths.)

       (*) Fix the `STR' instruction.

       (*) Fix the ELF output format, which was broken under certain
           circumstances due to the addition of stabs support.

       (*) Quick-fix Borland format debug-info for `-f obj'

       (*) Fix for `%rep' with no arguments (#560568)

       (*) Fix concatenation of preprocessor function call (#794686)

       (*) Fix long label causes coredump (#677841)

       (*) Use autoheader as well as autoconf to keep configure from
           generating ridiculously long command lines.

       (*) Make sure that all of the formats which support debugging output
           actually will suppress debugging output when `-g' not specified.

C.2.3 Version 0.98.37

       (*) Paths given in `-I' switch searched for `incbin'-ed as well as
           `%include'-ed files.

       (*) Added stabs debugging for the ELF output format, patch from
           Martin Wawro.

       (*) Fix `output/outbin.c' to allow origin > 80000000h.

       (*) Make `-U' switch work.

       (*) Fix the use of relative offsets with explicit prefixes, e.g.
           `a32 loop foo'.

       (*) Remove `backslash()'.

       (*) Fix the `SMSW' and `SLDT' instructions.

       (*) `-O2' and `-O3' are no longer aliases for `-O10' and `-O15'. If
           you mean the latter, please say so! :)

C.2.4 Version 0.98.36

       (*) Update rdoff - librarian/archiver - common rec - docs!

       (*) Fix signed/unsigned problems.

       (*) Fix `JMP FAR label' and `CALL FAR label'.

       (*) Add new multisection support - map files - fix align bug

       (*) Fix sysexit, movhps/movlps reg,reg bugs in insns.dat

       (*) `Q' or `O' suffixes indicate octal

       (*) Support Prescott new instructions (PNI).

       (*) Cyrix `XSTORE' instruction.

C.2.5 Version 0.98.35

       (*) Fix build failure on 16-bit DOS (Makefile.bc3 workaround for
           compiler bug.)

       (*) Fix dependencies and compiler warnings.

       (*) Add "const" in a number of places.

       (*) Add -X option to specify error reporting format (use -Xvc to
           integrate with Microsoft Visual Studio.)

       (*) Minor changes for code legibility.

       (*) Drop use of tmpnam() in rdoff (security fix.)

C.2.6 Version 0.98.34

       (*) Correct additional address-size vs. operand-size confusions.

       (*) Generate dependencies for all Makefiles automatically.

       (*) Add support for unimplemented (but theoretically available)
           registers such as tr0 and cr5. Segment registers 6 and 7 are
           called segr6 and segr7 for the operations which they can be
           represented.

       (*) Correct some disassembler bugs related to redundant address-size
           prefixes. Some work still remains in this area.

       (*) Correctly generate an error for things like "SEG eax".

       (*) Add the JMPE instruction, enabled by "CPU IA64".

       (*) Correct compilation on newer gcc/glibc platforms.

       (*) Issue an error on things like "jmp far eax".

C.2.7 Version 0.98.33

       (*) New __NASM_PATCHLEVEL__ and __NASM_VERSION_ID__ standard macros
           to round out the version-query macros. version.pl now
           understands X.YYplWW or X.YY.ZZplWW as a version number,
           equivalent to X.YY.ZZ.WW (or X.YY.0.WW, as appropriate).

       (*) New keyword "strict" to disable the optimization of specific
           operands.

       (*) Fix the handing of size overrides with JMP instructions
           (instructions such as "jmp dword foo".)

       (*) Fix the handling of "ABSOLUTE label", where "label" points into
           a relocatable segment.

       (*) Fix OBJ output format with lots of externs.

       (*) More documentation updates.

       (*) Add -Ov option to get verbose information about optimizations.

       (*) Undo a braindead change which broke `%elif' directives.

       (*) Makefile updates.

C.2.8 Version 0.98.32

       (*) Fix NASM crashing when `%macro' directives were left
           unterminated.

       (*) Lots of documentation updates.

       (*) Complete rewrite of the PostScript/PDF documentation generator.

       (*) The MS Visual C++ Makefile was updated and corrected.

       (*) Recognize .rodata as a standard section name in ELF.

       (*) Fix some obsolete Perl4-isms in Perl scripts.

       (*) Fix configure.in to work with autoconf 2.5x.

       (*) Fix a couple of "make cleaner" misses.

       (*) Make the normal "./configure && make" work with Cygwin.

C.2.9 Version 0.98.31

       (*) Correctly build in a separate object directory again.

       (*) Derive all references to the version number from the version
           file.

       (*) New standard macros __NASM_SUBMINOR__ and __NASM_VER__ macros.

       (*) Lots of Makefile updates and bug fixes.

       (*) New `%ifmacro' directive to test for multiline macros.

       (*) Documentation updates.

       (*) Fixes for 16-bit OBJ format output.

       (*) Changed the NASM environment variable to NASMENV.

C.2.10 Version 0.98.30

       (*) Changed doc files a lot: completely removed old READMExx and
           Wishlist files, incorporating all information in CHANGES and
           TODO.

       (*) I waited a long time to rename zoutieee.c to (original)
           outieee.c

       (*) moved all output modules to output/ subdirectory.

       (*) Added 'make strip' target to strip debug info from nasm &
           ndisasm.

       (*) Added INSTALL file with installation instructions.

       (*) Added -v option description to nasm man.

       (*) Added dist makefile target to produce source distributions.

       (*) 16-bit support for ELF output format (GNU extension, but
           useful.)

C.2.11 Version 0.98.28

       (*) Fastcooked this for Debian's Woody release: Frank applied the
           INCBIN bug patch to 0.98.25alt and called it 0.98.28 to not
           confuse poor little apt-get.

C.2.12 Version 0.98.26

       (*) Reorganised files even better from 0.98.25alt

C.2.13 Version 0.98.25alt

       (*) Prettified the source tree. Moved files to more reasonable
           places.

       (*) Added findleak.pl script to misc/ directory.

       (*) Attempted to fix doc.

C.2.14 Version 0.98.25

       (*) Line continuation character `\'.

       (*) Docs inadvertantly reverted - "dos packaging".

C.2.15 Version 0.98.24p1

       (*) FIXME: Someone, document this please.

C.2.16 Version 0.98.24

       (*) Documentation - Ndisasm doc added to Nasm.doc.

C.2.17 Version 0.98.23

       (*) Attempted to remove rdoff version1

       (*) Lino Mastrodomenico's patches to preproc.c (%$$ bug?).

C.2.18 Version 0.98.22

       (*) Update rdoff2 - attempt to remove v1.

C.2.19 Version 0.98.21

       (*) Optimization fixes.

C.2.20 Version 0.98.20

       (*) Optimization fixes.

C.2.21 Version 0.98.19

       (*) H. J. Lu's patch back out.

C.2.22 Version 0.98.18

       (*) Added ".rdata" to "-f win32".

C.2.23 Version 0.98.17

       (*) H. J. Lu's "bogus elf" patch. (Red Hat problem?)

C.2.24 Version 0.98.16

       (*) Fix whitespace before "[section ..." bug.

C.2.25 Version 0.98.15

       (*) Rdoff changes (?).

       (*) Fix fixes to memory leaks.

C.2.26 Version 0.98.14

       (*) Fix memory leaks.

C.2.27 Version 0.98.13

       (*) There was no 0.98.13

C.2.28 Version 0.98.12

       (*) Update optimization (new function of "-O1")

       (*) Changes to test/bintest.asm (?).

C.2.29 Version 0.98.11

       (*) Optimization changes.

       (*) Ndisasm fixed.

C.2.30 Version 0.98.10

       (*) There was no 0.98.10

C.2.31 Version 0.98.09

       (*) Add multiple sections support to "-f bin".

       (*) Changed GLOBAL_TEMP_BASE in outelf.c from 6 to 15.

       (*) Add "-v" as an alias to the "-r" switch.

       (*) Remove "#ifdef" from Tasm compatibility options.

       (*) Remove redundant size-overrides on "mov ds, ex", etc.

       (*) Fixes to SSE2, other insns.dat (?).

       (*) Enable uppercase "I" and "P" switches.

       (*) Case insinsitive "seg" and "wrt".

       (*) Update install.sh (?).

       (*) Allocate tokens in blocks.

       (*) Improve "invalid effective address" messages.

C.2.32 Version 0.98.08

       (*) Add "`%strlen'" and "`%substr'" macro operators

       (*) Fixed broken c16.mac.

       (*) Unterminated string error reported.

       (*) Fixed bugs as per 0.98bf

C.2.33 Version 0.98.09b with John Coffman patches released 28-Oct-2001

       Changes from 0.98.07 release to 98.09b as of 28-Oct-2001

       (*) More closely compatible with 0.98 when -O0 is implied or
           specified. Not strictly identical, since backward branches in
           range of short offsets are recognized, and signed byte values
           with no explicit size specification will be assembled as a
           single byte.

       (*) More forgiving with the PUSH instruction. 0.98 requires a size
           to be specified always. 0.98.09b will imply the size from the
           current BITS setting (16 or 32).

       (*) Changed definition of the optimization flag:

       -O0 strict two-pass assembly, JMP and Jcc are handled more like
       0.98, except that back- ward JMPs are short, if possible.

       -O1 strict two-pass assembly, but forward branches are assembled
       with code guaranteed to reach; may produce larger code than -O0, but
       will produce successful assembly more often if branch offset sizes
       are not specified.

       -O2 multi-pass optimization, minimize branch offsets; also will
       minimize signed immed- iate bytes, overriding size specification.

       -O3 like -O2, but more passes taken, if needed

C.2.34 Version 0.98.07 released 01/28/01

       (*) Added Stepane Denis' SSE2 instructions to a *working* version of
           the code - some earlier versions were based on broken code -
           sorry 'bout that. version "0.98.07"

       01/28/01

       (*) Cosmetic modifications to nasm.c, nasm.h, AUTHORS, MODIFIED

C.2.35 Version 0.98.06f released 01/18/01

       (*) - Add "metalbrain"s jecxz bug fix in insns.dat - alter
           nasmdoc.src to match - version "0.98.06f"

C.2.36 Version 0.98.06e released 01/09/01

       (*) Removed the "outforms.h" file - it appears to be someone's old
           backup of "outform.h". version "0.98.06e"

       01/09/01

       (*) fbk - finally added the fix for the "multiple %includes bug",
           known since 7/27/99 - reported originally (?) and sent to us by
           Austin Lunnen - he reports that John Fine had a fix within the
           day. Here it is...

       (*) Nelson Rush resigns from the group. Big thanks to Nelson for his
           leadership and enthusiasm in getting these changes incorporated
           into Nasm!

       (*) fbk - [list +], [list -] directives - ineptly implemented,
           should be re-written or removed, perhaps.

       (*) Brian Raiter / fbk - "elfso bug" fix - applied to aoutb format
           as well - testing might be desirable...

       08/07/00

       (*) James Seter - -postfix, -prefix command line switches.

       (*) Yuri Zaporogets - rdoff utility changes.

C.2.37 Version 0.98p1

       (*) GAS-like palign (Panos Minos)

       (*) FIXME: Someone, fill this in with details

C.2.38 Version 0.98bf (bug-fixed)

       (*) Fixed - elf and aoutb bug - shared libraries - multiple
           "%include" bug in "-f obj" - jcxz, jecxz bug - unrecognized
           option bug in ndisasm

C.2.39 Version 0.98.03 with John Coffman's changes released 27-Jul-2000

       (*) Added signed byte optimizations for the 0x81/0x83 class of
           instructions: ADC, ADD, AND, CMP, OR, SBB, SUB, XOR: when used
           as 'ADD reg16,imm' or 'ADD reg32,imm.' Also optimization of
           signed byte form of 'PUSH imm' and 'IMUL reg,imm'/'IMUL
           reg,reg,imm.' No size specification is needed.

       (*) Added multi-pass JMP and Jcc offset optimization. Offsets on
           forward references will preferentially use the short form,
           without the need to code a specific size (short or near) for the
           branch. Added instructions for 'Jcc label' to use the form
           'Jnotcc $+3/JMP label', in cases where a short offset is out of
           bounds. If compiling for a 386 or higher CPU, then the 386 form
           of Jcc will be used instead.

       This feature is controlled by a new command-line switch: "O", (upper
       case letter O). "-O0" reverts the assembler to no extra optimization
       passes, "-O1" allows up to 5 extra passes, and "-O2"(default),
       allows up to 10 extra optimization passes.

       (*) Added a new directive: 'cpu XXX', where XXX is any of: 8086,
           186, 286, 386, 486, 586, pentium, 686, PPro, P2, P3 or Katmai.
           All are case insensitive. All instructions will be selected only
           if they apply to the selected cpu or lower. Corrected a couple
           of bugs in cpu-dependence in 'insns.dat'.

       (*) Added to 'standard.mac', the "use16" and "use32" forms of the
           "bits 16/32" directive. This is nothing new, just conforms to a
           lot of other assemblers. (minor)

       (*) Changed label allocation from 320/32 (10000 labels @ 200K+) to
           32/37 (1000 labels); makes running under DOS much easier. Since
           additional label space is allocated dynamically, this should
           have no effect on large programs with lots of labels. The 37 is
           a prime, believed to be better for hashing. (minor)

C.2.40 Version 0.98.03

       "Integrated patchfile 0.98-0.98.01. I call this version 0.98.03 for
       historical reasons: 0.98.02 was trashed." --John Coffman
       <johninsd@san.rr.com>, 27-Jul-2000

       (*) Kendall Bennett's SciTech MGL changes

       (*) Note that you must define "TASM_COMPAT" at compile-time to get
           the Tasm Ideal Mode compatibility.

       (*) All changes can be compiled in and out using the TASM_COMPAT
           macros, and when compiled without TASM_COMPAT defined we get the
           exact same binary as the unmodified 0.98 sources.

       (*) standard.mac, macros.c: Added macros to ignore TASM directives
           before first include

       (*) nasm.h: Added extern declaration for tasm_compatible_mode

       (*) nasm.c: Added global variable tasm_compatible_mode

       (*) Added command line switch for TASM compatible mode (-t)

       (*) Changed version command line to reflect when compiled with TASM
           additions

       (*) Added response file processing to allow all arguments on a
           single line (response file is @resp rather than -@resp for NASM
           format).

       (*) labels.c: Changes islocal() macro to support TASM style @@local
           labels.

       (*) Added islocalchar() macro to support TASM style @@local labels.

       (*) parser.c: Added support for TASM style memory references (ie:
           mov [DWORD eax],10 rather than the NASM style mov DWORD
           [eax],10).

       (*) preproc.c: Added new directives, `%arg', `%local', `%stacksize'
           to directives table

       (*) Added support for TASM style directives without a leading %
           symbol.

       (*) Integrated a block of changes from Andrew Zabolotny
           <bit@eltech.ru>:

       (*) A new keyword `%xdefine' and its case-insensitive counterpart
           `%ixdefine'. They work almost the same way as `%define' and
           `%idefine' but expand the definition immediately, not on the
           invocation. Something like a cross between `%define' and
           `%assign'. The "x" suffix stands for "eXpand", so "xdefine" can
           be deciphered as "expand-and-define". Thus you can do things
           like this:

            %assign ofs     0
      
            %macro  arg     1
                    %xdefine %1 dword [esp+ofs]
                    %assign ofs ofs+4
            %endmacro

       (*) Changed the place where the expansion of %$name macros are
           expanded. Now they are converted into ..@ctxnum.name form when
           detokenizing, so there are no quirks as before when using %$name
           arguments to macros, in macros etc. For example:

            %macro  abc     1
                    %define %1 hello
            %endm
      
            abc     %$here
            %$here

       Now last line will be expanded into "hello" as expected. This also
       allows for lots of goodies, a good example are extended "proc"
       macros included in this archive.

       (*) Added a check for "cstk" in smacro_defined() before calling
           get_ctx() - this allows for things like:

            %ifdef %$abc
            %endif

       to work without warnings even in no context.

       (*) Added a check for "cstk" in %if*ctx and %elif*ctx directives -
           this allows to use `%ifctx' without excessive warnings. If there
           is no active context, `%ifctx' goes through "false" branch.

       (*) Removed "user error: " prefix with `%error' directive: it just
           clobbers the output and has absolutely no functionality.
           Besides, this allows to write macros that does not differ from
           built-in functions in any way.

       (*) Added expansion of string that is output by `%error' directive.
           Now you can do things like:

            %define hello(x) Hello, x!
      
            %define %$name andy
            %error "hello(%$name)"

       Same happened with `%include' directive.

       (*) Now all directives that expect an identifier will try to expand
           and concatenate everything without whitespaces in between before
           usage. For example, with "unfixed" nasm the commands

            %define %$abc hello
            %define __%$abc goodbye
            __%$abc

       would produce "incorrect" output: last line will expand to

            hello goodbyehello

       Not quite what you expected, eh? :-) The answer is that preprocessor
       treats the `%define' construct as if it would be

            %define __ %$abc goodbye

       (note the white space between __ and %$abc). After my "fix" it will
       "correctly" expand into

            goodbye

       as expected. Note that I use quotes around words "correct",
       "incorrect" etc because this is rather a feature not a bug; however
       current behaviour is more logical (and allows more advanced macro
       usage :-).

       Same change was applied to:
       `%push',`%macro',`%imacro',`%define',`%idefine',`%xdefine',`%ixdefine',
       `%assign',`%iassign',`%undef'

       (*) A new directive [WARNING {+|-}warning-id] have been added. It
           works only if the assembly phase is enabled (i.e. it doesn't
           work with nasm -e).

       (*) A new warning type: macro-selfref. By default this warning is
           disabled; when enabled NASM warns when a macro self-references
           itself; for example the following source:

              [WARNING macro-selfref]
      
              %macro          push    1-*
                      %rep    %0
                              push    %1
                              %rotate 1
                      %endrep
              %endmacro
      
                              push    eax,ebx,ecx

       will produce a warning, but if we remove the first line we won't see
       it anymore (which is The Right Thing To Do {tm} IMHO since C
       preprocessor eats such constructs without warnings at all).

       (*) Added a "error" routine to preprocessor which always will set
           ERR_PASS1 bit in severity_code. This removes annoying repeated
           errors on first and second passes from preprocessor.

       (*) Added the %+ operator in single-line macros for concatenating
           two identifiers. Usage example:

              %define _myfunc _otherfunc
              %define cextern(x) _ %+ x
              cextern (myfunc)

       After first expansion, third line will become "_myfunc". After this
       expansion is performed again so it becomes "_otherunc".

       (*) Now if preprocessor is in a non-emitting state, no warning or
           error will be emitted. Example:

              %if 1
                      mov     eax,ebx
              %else
                      put anything you want between these two brackets,
                      even macro-parameter references %1 or local
                      labels %$zz or macro-local labels %%zz - no
                      warning will be emitted.
              %endif

       (*) Context-local variables on expansion as a last resort are looked
           up in outer contexts. For example, the following piece:

              %push   outer
              %define %$a [esp]
      
                      %push   inner
                      %$a
                      %pop
              %pop

       will expand correctly the fourth line to [esp]; if we'll define
       another %$a inside the "inner" context, it will take precedence over
       outer definition. However, this modification has been applied only
       to expand_smacro and not to smacro_define: as a consequence
       expansion looks in outer contexts, but `%ifdef' won't look in outer
       contexts.

       This behaviour is needed because we don't want nested contexts to
       act on already defined local macros. Example:

              %define %$arg1  [esp+4]
              test    eax,eax
              if      nz
                      mov     eax,%$arg1
              endif

       In this example the "if" mmacro enters into the "if" context, so
       %$arg1 is not valid anymore inside "if". Of course it could be
       worked around by using explicitely %$$arg1 but this is ugly IMHO.

       (*) Fixed memory leak in `%undef'. The origline wasn't freed before
           exiting on success.

       (*) Fixed trap in preprocessor when line expanded to empty set of
           tokens. This happens, for example, in the following case:

              #define SOMETHING
              SOMETHING

C.2.41 Version 0.98

       All changes since NASM 0.98p3 have been produced by H. Peter Anvin
       <hpa@zytor.com>.

       (*) The documentation comment delimiter is

       (*) Allow EQU definitions to refer to external labels; reported by
           Pedro Gimeno.

       (*) Re-enable support for RDOFF v1; reported by Pedro Gimeno.

       (*) Updated License file per OK from Simon and Julian.

C.2.42 Version 0.98p9

       (*) Update documentation (although the instruction set reference
           will have to wait; I don't want to hold up the 0.98 release for
           it.)

       (*) Verified that the NASM implementation of the PEXTRW and PMOVMSKB
           instructions is correct. The encoding differs from what the
           Intel manuals document, but the Pentium III behaviour matches
           NASM, not the Intel manuals.

       (*) Fix handling of implicit sizes in PSHUFW and PINSRW, reported by
           Stefan Hoffmeister.

       (*) Resurrect the -s option, which was removed when changing the
           diagnostic output to stdout.

C.2.43 Version 0.98p8

       (*) Fix for "DB" when NASM is running on a bigendian machine.

       (*) Invoke insns.pl once for each output script, making Makefile.in
           legal for "make -j".

       (*) Improve the Unix configure-based makefiles to make package
           creation easier.

       (*) Included an RPM .spec file for building RPM (RedHat Package
           Manager) packages on Linux or Unix systems.

       (*) Fix Makefile dependency problems.

       (*) Change src/rdsrc.pl to include sectioning information in info
           output; required for install-info to work.

       (*) Updated the RDOFF distribution to version 2 from Jules; minor
           massaging to make it compile in my environment.

       (*) Split doc files that can be built by anyone with a Perl
           interpreter off into a separate archive.

       (*) "Dress rehearsal" release!

C.2.44 Version 0.98p7

       (*) Fixed opcodes with a third byte-sized immediate argument to not
           complain if given "byte" on the immediate.

       (*) Allow `%undef' to remove single-line macros with arguments. This
           matches the behaviour of #undef in the C preprocessor.

       (*) Allow -d, -u, -i and -p to be specified as -D, -U, -I and -P for
           compatibility with most C compilers and preprocessors. This
           allows Makefile options to be shared between cc and nasm, for
           example.

       (*) Minor cleanups.

       (*) Went through the list of Katmai instructions and hopefully fixed
           the (rather few) mistakes in it.

       (*) (Hopefully) fixed a number of disassembler bugs related to
           ambiguous instructions (disambiguated by -p) and SSE
           instructions with REP.

       (*) Fix for bug reported by Mark Junger: "call dword 0x12345678"
           should work and may add an OSP (affected CALL, JMP, Jcc).

       (*) Fix for environments when "stderr" isn't a compile-time
           constant.

C.2.45 Version 0.98p6

       (*) Took officially over coordination of the 0.98 release; so drop
           the p3.x notation. Skipped p4 and p5 to avoid confusion with
           John Fine's J4 and J5 releases.

       (*) Update the documentation; however, it still doesn't include
           documentation for the various new instructions. I somehow wonder
           if it makes sense to have an instruction set reference in the
           assembler manual when Intel et al have PDF versions of their
           manuals online.

       (*) Recognize "idt" or "centaur" for the -p option to ndisasm.

       (*) Changed error messages back to stderr where they belong, but add
           an -E option to redirect them elsewhere (the DOS shell cannot
           redirect stderr.)

       (*) -M option to generate Makefile dependencies (based on code from
           Alex Verstak.)

       (*) `%undef' preprocessor directive, and -u option, that undefines a
           single-line macro.

       (*) OS/2 Makefile (Mkfiles/Makefile.os2) for Borland under OS/2;
           from Chuck Crayne.

       (*) Various minor bugfixes (reported by): - Dangling `%s' in
           preproc.c (Martin Junker)

       (*) THERE ARE KNOWN BUGS IN SSE AND THE OTHER KATMAI INSTRUCTIONS. I
           am on a trip and didn't bring the Katmai instruction reference,
           so I can't work on them right now.

       (*) Updated the License file per agreement with Simon and Jules to
           include a GPL distribution clause.

C.2.46 Version 0.98p3.7

       (*) (Hopefully) fixed the canned Makefiles to include the outrdf2
           and zoutieee modules.

       (*) Renamed changes.asm to changed.asm.

C.2.47 Version 0.98p3.6

       (*) Fixed a bunch of instructions that were added in 0.98p3.5 which
           had memory operands, and the address-size prefix was missing
           from the instruction pattern.

C.2.48 Version 0.98p3.5

       (*) Merged in changes from John S. Fine's 0.98-J5 release. John's
           based 0.98-J5 on my 0.98p3.3 release; this merges the changes.

       (*) Expanded the instructions flag field to a long so we can fit
           more flags; mark SSE (KNI) and AMD or Katmai-specific
           instructions as such.

       (*) Fix the "PRIV" flag on a bunch of instructions, and create new
           "PROT" flag for protected-mode-only instructions (orthogonal to
           if the instruction is privileged!) and new "SMM" flag for SMM-
           only instructions.

       (*) Added AMD-only SYSCALL and SYSRET instructions.

       (*) Make SSE actually work, and add new Katmai MMX instructions.

       (*) Added a -p (preferred vendor) option to ndisasm so that it can
           distinguish e.g. Cyrix opcodes also used in SSE. For example:

            ndisasm -p cyrix aliased.bin
            00000000  670F514310        paddsiw mm0,[ebx+0x10]
            00000005  670F514320        paddsiw mm0,[ebx+0x20]
            ndisasm -p intel aliased.bin
            00000000  670F514310        sqrtps xmm0,[ebx+0x10]
            00000005  670F514320        sqrtps xmm0,[ebx+0x20]

       (*) Added a bunch of Cyrix-specific instructions.

C.2.49 Version 0.98p3.4

       (*) Made at least an attempt to modify all the additional Makefiles
           (in the Mkfiles directory). I can't test it, but this was the
           best I could do.

       (*) DOS DJGPP+"Opus Make" Makefile from John S. Fine.

       (*) changes.asm changes from John S. Fine.

C.2.50 Version 0.98p3.3

       (*) Patch from Conan Brink to allow nesting of `%rep' directives.

       (*) If we're going to allow INT01 as an alias for INT1/ICEBP (one of
           Jules 0.98p3 changes), then we should allow INT03 as an alias
           for INT3 as well.

       (*) Updated changes.asm to include the latest changes.

       (*) Tried to clean up the <CR>s that had snuck in from a DOS/Windows
           environment into my Unix environment, and try to make sure than
           DOS/Windows users get them back.

       (*) We would silently generate broken tools if insns.dat wasn't
           sorted properly. Change insns.pl so that the order doesn't
           matter.

       (*) Fix bug in insns.pl (introduced by me) which would cause
           conditional instructions to have an extra "cc" in disassembly,
           e.g. "jnz" disassembled as "jccnz".

C.2.51 Version 0.98p3.2

       (*) Merged in John S. Fine's changes from his 0.98-J4 prerelease;
           see http://www.csoft.net/cz/johnfine/

       (*) Changed previous "spotless" Makefile target (appropriate for
           distribution) to "distclean", and added "cleaner" target which
           is same as "clean" except deletes files generated by Perl
           scripts; "spotless" is union.

       (*) Removed BASIC programs from distribution. Get a Perl interpreter
           instead (see below.)

       (*) Calling this "pre-release 3.2" rather than "p3-hpa2" because of
           John's contributions.

       (*) Actually link in the IEEE output format (zoutieee.c); fix a
           bunch of compiler warnings in that file. Note I don't know what
           IEEE output is supposed to look like, so these changes were made
           "blind".

C.2.52 Version 0.98p3-hpa

       (*) Merged nasm098p3.zip with nasm-0.97.tar.gz to create a fully
           buildable version for Unix systems (Makefile.in updates, etc.)

       (*) Changed insns.pl to create the instruction tables in nasm.h and
           names.c, so that a new instruction can be added by adding it
           *only* to insns.dat.

       (*) Added the following new instructions: SYSENTER, SYSEXIT, FXSAVE,
           FXRSTOR, UD1, UD2 (the latter two are two opcodes that Intel
           guarantee will never be used; one of them is documented as UD2
           in Intel documentation, the other one just as "Undefined Opcode"
           -- calling it UD1 seemed to make sense.)

       (*) MAX_SYMBOL was defined to be 9, but LOADALL286 and LOADALL386
           are 10 characters long. Now MAX_SYMBOL is derived from
           insns.dat.

       (*) A note on the BASIC programs included: forget them. insns.bas is
           already out of date. Get yourself a Perl interpreter for your
           platform of choice at http://www.cpan.org/ports/index.html.

C.2.53 Version 0.98 pre-release 3

       (*) added response file support, improved command line handling, new
           layout help screen

       (*) fixed limit checking bug, 'OUT byte nn, reg' bug, and a couple
           of rdoff related bugs, updated Wishlist; 0.98 Prerelease 3.

C.2.54 Version 0.98 pre-release 2

       (*) fixed bug in outcoff.c to do with truncating section names
           longer than 8 characters, referencing beyond end of string; 0.98
           pre-release 2

C.2.55 Version 0.98 pre-release 1

       (*) Fixed a bug whereby STRUC didn't work at all in RDF.

       (*) Fixed a problem with group specification in PUBDEFs in OBJ.

       (*) Improved ease of adding new output formats. Contribution due to
           Fox Cutter.

       (*) Fixed a bug in relocations in the `bin' format: was showing up
           when a relocatable reference crossed an 8192-byte boundary in
           any output section.

       (*) Fixed a bug in local labels: local-label lookups were
           inconsistent between passes one and two if an EQU occurred
           between the definition of a global label and the subsequent use
           of a local label local to that global.

       (*) Fixed a seg-fault in the preprocessor (again) which happened
           when you use a blank line as the first line of a multi-line
           macro definition and then defined a label on the same line as a
           call to that macro.

       (*) Fixed a stale-pointer bug in the handling of the NASM
           environment variable. Thanks to Thomas McWilliams.

       (*) ELF had a hard limit on the number of sections which caused
           segfaults when transgressed. Fixed.

       (*) Added ability for ndisasm to read from stdin by using `-' as the
           filename.

       (*) ndisasm wasn't outputting the TO keyword. Fixed.

       (*) Fixed error cascade on bogus expression in `%if' - an error in
           evaluation was causing the entire `%if' to be discarded, thus
           creating trouble later when the `%else' or `%endif' was
           encountered.

       (*) Forward reference tracking was instruction-granular not operand-
           granular, which was causing 286-specific code to be generated
           needlessly on code of the form `shr word [forwardref],1'. Thanks
           to Jim Hague for sending a patch.

       (*) All messages now appear on stdout, as sending them to stderr
           serves no useful purpose other than to make redirection
           difficult.

       (*) Fixed the problem with EQUs pointing to an external symbol -
           this now generates an error message.

       (*) Allowed multiple size prefixes to an operand, of which only the
           first is taken into account.

       (*) Incorporated John Fine's changes, including fixes of a large
           number of preprocessor bugs, some small problems in OBJ, and a
           reworking of label handling to define labels before their line
           is assembled, rather than after.

       (*) Reformatted a lot of the source code to be more readable.
           Included 'coding.txt' as a guideline for how to format code for
           contributors.

       (*) Stopped nested `%reps' causing a panic - they now cause a
           slightly more friendly error message instead.

       (*) Fixed floating point constant problems (patch by Pedro Gimeno)

       (*) Fixed the return value of insn_size() not being checked for -1,
           indicating an error.

       (*) Incorporated 3Dnow! instructions.

       (*) Fixed the 'mov eax, eax + ebx' bug.

       (*) Fixed the GLOBAL EQU bug in ELF. Released developers release 3.

       (*) Incorporated John Fine's command line parsing changes

       (*) Incorporated David Lindauer's OMF debug support

       (*) Made changes for LCC 4.0 support (`__NASM_CDecl__', removed
           register size specification warning when sizes agree).

   C.3 NASM 0.9 Series

       Revisions before 0.98.

C.3.1 Version 0.97 released December 1997

       (*) This was entirely a bug-fix release to 0.96, which seems to have
           got cursed. Silly me.

       (*) Fixed stupid mistake in OBJ which caused `MOV EAX,<constant>' to
           fail. Caused by an error in the `MOV EAX,<segment>' support.

       (*) ndisasm hung at EOF when compiled with lcc on Linux because lcc
           on Linux somehow breaks feof(). ndisasm now does not rely on
           feof().

       (*) A heading in the documentation was missing due to a markup error
           in the indexing. Fixed.

       (*) Fixed failure to update all pointers on realloc() within
           extended- operand code in parser.c. Was causing wrong behaviour
           and seg faults on lines such as `dd 0.0,0.0,0.0,0.0,...'

       (*) Fixed a subtle preprocessor bug whereby invoking one multi-line
           macro on the first line of the expansion of another, when the
           second had been invoked with a label defined before it, didn't
           expand the inner macro.

       (*) Added internal.doc back in to the distribution archives - it was
           missing in 0.96 *blush*

       (*) Fixed bug causing 0.96 to be unable to assemble its own test
           files, specifically objtest.asm. *blush again*

       (*) Fixed seg-faults and bogus error messages caused by mismatching
           `%rep' and `%endrep' within multi-line macro definitions.

       (*) Fixed a problem with buffer overrun in OBJ, which was causing
           corruption at ends of long PUBDEF records.

       (*) Separated DOS archives into main-program and documentation to
           reduce download size.

C.3.2 Version 0.96 released November 1997

       (*) Fixed a bug whereby, if `nasm sourcefile' would cause a filename
           collision warning and put output into `nasm.out', then `nasm
           sourcefile -o outputfile' still gave the warning even though the
           `-o' was honoured. Fixed name pollution under Digital UNIX: one
           of its header files defined R_SP, which broke the enum in
           nasm.h.

       (*) Fixed minor instruction table problems: FUCOM and FUCOMP didn't
           have two-operand forms; NDISASM didn't recognise the longer
           register forms of PUSH and POP (eg FF F3 for PUSH BX); TEST
           mem,imm32 was flagged as undocumented; the 32-bit forms of CMOV
           had 16-bit operand size prefixes; `AAD imm' and `AAM imm' are no
           longer flagged as undocumented because the Intel Architecture
           reference documents them.

       (*) Fixed a problem with the local-label mechanism, whereby strange
           types of symbol (EQUs, auto-defined OBJ segment base symbols)
           interfered with the `previous global label' value and screwed up
           local labels.

       (*) Fixed a bug whereby the stub preprocessor didn't communicate
           with the listing file generator, so that the -a and -l options
           in conjunction would produce a useless listing file.

       (*) Merged `os2' object file format back into `obj', after
           discovering that `obj' _also_ shouldn't have a link pass
           separator in a module containing a non-trivial MODEND. Flat
           segments are now declared using the FLAT attribute. `os2' is no
           longer a valid object format name: use `obj'.

       (*) Removed the fixed-size temporary storage in the evaluator. Very
           very long expressions (like `mov ax,1+1+1+1+...' for two hundred
           1s or so) should now no longer crash NASM.

       (*) Fixed a bug involving segfaults on disassembly of MMX
           instructions, by changing the meaning of one of the operand-type
           flags in nasm.h. This may cause other apparently unrelated MMX
           problems; it needs to be tested thoroughly.

       (*) Fixed some buffer overrun problems with large OBJ output files.
           Thanks to DJ Delorie for the bug report and fix.

       (*) Made preprocess-only mode actually listen to the `%line' markers
           as it prints them, so that it can report errors more sanely.

       (*) Re-designed the evaluator to keep more sensible track of
           expressions involving forward references: can now cope with
           previously-nightmare situations such as:

         mov ax,foo | bar
         foo equ 1
         bar equ 2

       (*) Added the ALIGN and ALIGNB standard macros.

       (*) Added PIC support in ELF: use of WRT to obtain the four extra
           relocation types needed.

       (*) Added the ability for output file formats to define their own
           extensions to the GLOBAL, COMMON and EXTERN directives.

       (*) Implemented common-variable alignment, and global-symbol type
           and size declarations, in ELF.

       (*) Implemented NEAR and FAR keywords for common variables, plus
           far-common element size specification, in OBJ.

       (*) Added a feature whereby EXTERNs and COMMONs in OBJ can be given
           a default WRT specification (either a segment or a group).

       (*) Transformed the Unix NASM archive into an auto-configuring
           package.

       (*) Added a sanity-check for people applying SEG to things which are
           already segment bases: this previously went unnoticed by the SEG
           processing and caused OBJ-driver panics later.

       (*) Added the ability, in OBJ format, to deal with `MOV
           EAX,<segment>' type references: OBJ doesn't directly support
           dword-size segment base fixups, but as long as the low two bytes
           of the constant term are zero, a word-size fixup can be
           generated instead and it will work.

       (*) Added the ability to specify sections' alignment requirements in
           Win32 object files and pure binary files.

       (*) Added preprocess-time expression evaluation: the `%assign' (and
           `%iassign') directive and the bare `%if' (and `%elif')
           conditional. Added relational operators to the evaluator, for
           use only in `%if' constructs: the standard relationals = < > <=
           >= <> (and C-like synonyms == and !=) plus low-precedence
           logical operators &&, ^^ and ||.

       (*) Added a preprocessor repeat construct: `%rep' / `%exitrep' /
           `%endrep'.

       (*) Added the __FILE__ and __LINE__ standard macros.

       (*) Added a sanity check for number constants being greater than
           0xFFFFFFFF. The warning can be disabled.

       (*) Added the %0 token whereby a variadic multi-line macro can tell
           how many parameters it's been given in a specific invocation.

       (*) Added `%rotate', allowing multi-line macro parameters to be
           cycled.

       (*) Added the `*' option for the maximum parameter count on multi-
           line macros, allowing them to take arbitrarily many parameters.

       (*) Added the ability for the user-level forms of EXTERN, GLOBAL and
           COMMON to take more than one argument.

       (*) Added the IMPORT and EXPORT directives in OBJ format, to deal
           with Windows DLLs.

       (*) Added some more preprocessor `%if' constructs: `%ifidn' /
           `%ifidni' (exact textual identity), and `%ifid' / `%ifnum' /
           `%ifstr' (token type testing).

       (*) Added the ability to distinguish SHL AX,1 (the 8086 version)
           from SHL AX,BYTE 1 (the 286-and-upwards version whose constant
           happens to be 1).

       (*) Added NetBSD/FreeBSD/OpenBSD's variant of a.out format, complete
           with PIC shared library features.

       (*) Changed NASM's idiosyncratic handling of FCLEX, FDISI, FENI,
           FINIT, FSAVE, FSTCW, FSTENV, and FSTSW to bring it into line
           with the otherwise accepted standard. The previous behaviour,
           though it was a deliberate feature, was a deliberate feature
           based on a misunderstanding. Apologies for the inconvenience.

       (*) Improved the flexibility of ABSOLUTE: you can now give it an
           expression rather than being restricted to a constant, and it
           can take relocatable arguments as well.

       (*) Added the ability for a variable to be declared as EXTERN
           multiple times, and the subsequent definitions are just ignored.

       (*) We now allow instruction prefixes (CS, DS, LOCK, REPZ etc) to be
           alone on a line (without a following instruction).

       (*) Improved sanity checks on whether the arguments to EXTERN,
           GLOBAL and COMMON are valid identifiers.

       (*) Added misc/exebin.mac to allow direct generation of .EXE files
           by hacking up an EXE header using DB and DW; also added
           test/binexe.asm to demonstrate the use of this. Thanks to Yann
           Guidon for contributing the EXE header code.

       (*) ndisasm forgot to check whether the input file had been
           successfully opened. Now it does. Doh!

       (*) Added the Cyrix extensions to the MMX instruction set.

       (*) Added a hinting mechanism to allow [EAX+EBX] and [EBX+EAX] to be
           assembled differently. This is important since [ESI+EBP] and
           [EBP+ESI] have different default base segment registers.

       (*) Added support for the PharLap OMF extension for 4096-byte
           segment alignment.

C.3.3 Version 0.95 released July 1997

       (*) Fixed yet another ELF bug. This one manifested if the user
           relied on the default segment, and attempted to define global
           symbols without first explicitly declaring the target segment.

       (*) Added makefiles (for NASM and the RDF tools) to build Win32
           console apps under Symantec C++. Donated by Mark Junker.

       (*) Added `macros.bas' and `insns.bas', QBasic versions of the Perl
           scripts that convert `standard.mac' to `macros.c' and convert
           `insns.dat' to `insnsa.c' and `insnsd.c'. Also thanks to Mark
           Junker.

       (*) Changed the diassembled forms of the conditional instructions so
           that JB is now emitted as JC, and other similar changes.
           Suggested list by Ulrich Doewich.

       (*) Added `@' to the list of valid characters to begin an identifier
           with.

       (*) Documentary changes, notably the addition of the `Common
           Problems' section in nasm.doc.

       (*) Fixed a bug relating to 32-bit PC-relative fixups in OBJ.

       (*) Fixed a bug in perm_copy() in labels.c which was causing
           exceptions in cleanup_labels() on some systems.

       (*) Positivity sanity check in TIMES argument changed from a warning
           to an error following a further complaint.

       (*) Changed the acceptable limits on byte and word operands to allow
           things like `~10111001b' to work.

       (*) Fixed a major problem in the preprocessor which caused seg-
           faults if macro definitions contained blank lines or comment-
           only lines.

       (*) Fixed inadequate error checking on the commas separating the
           arguments to `db', `dw' etc.

       (*) Fixed a crippling bug in the handling of macros with operand
           counts defined with a `+' modifier.

       (*) Fixed a bug whereby object file formats which stored the input
           file name in the output file (such as OBJ and COFF) weren't
           doing so correctly when the output file name was specified on
           the command line.

       (*) Removed [INC] and [INCLUDE] support for good, since they were
           obsolete anyway.

       (*) Fixed a bug in OBJ which caused all fixups to be output in 16-
           bit (old-format) FIXUPP records, rather than putting the 32-bit
           ones in FIXUPP32 (new-format) records.

       (*) Added, tentatively, OS/2 object file support (as a minor variant
           on OBJ).

       (*) Updates to Fox Cutter's Borland C makefile, Makefile.bc2.

       (*) Removed a spurious second fclose() on the output file.

       (*) Added the `-s' command line option to redirect all messages
           which would go to stderr (errors, help text) to stdout instead.

       (*) Added the `-w' command line option to selectively suppress some
           classes of assembly warning messages.

       (*) Added the `-p' pre-include and `-d' pre-define command-line
           options.

       (*) Added an include file search path: the `-i' command line option.

       (*) Fixed a silly little preprocessor bug whereby starting a line
           with a `%!' environment-variable reference caused an `unknown
           directive' error.

       (*) Added the long-awaited listing file support: the `-l' command
           line option.

       (*) Fixed a problem with OBJ format whereby, in the absence of any
           explicit segment definition, non-global symbols declared in the
           implicit default segment generated spurious EXTDEF records in
           the output.

       (*) Added the NASM environment variable.

       (*) From this version forward, Win32 console-mode binaries will be
           included in the DOS distribution in addition to the 16-bit
           binaries. Added Makefile.vc for this purpose.

       (*) Added `return 0;' to test/objlink.c to prevent compiler
           warnings.

       (*) Added the __NASM_MAJOR__ and __NASM_MINOR__ standard defines.

       (*) Added an alternative memory-reference syntax in which prefixing
           an operand with `&' is equivalent to enclosing it in square
           brackets, at the request of Fox Cutter.

       (*) Errors in pass two now cause the program to return a non-zero
           error code, which they didn't before.

       (*) Fixed the single-line macro cycle detection, which didn't work
           at all on macros with no parameters (caused an infinite loop).
           Also changed the behaviour of single-line macro cycle detection
           to work like cpp, so that macros like `extrn' as given in the
           documentation can be implemented.

       (*) Fixed the implementation of WRT, which was too restrictive in
           that you couldn't do `mov ax,[di+abc wrt dgroup]' because
           (di+abc) wasn't a relocatable reference.

C.3.4 Version 0.94 released April 1997

       (*) Major item: added the macro processor.

       (*) Added undocumented instructions SMI, IBTS, XBTS and LOADALL286.
           Also reorganised CMPXCHG instruction into early-486 and Pentium
           forms. Thanks to Thobias Jones for the information.

       (*) Fixed two more stupid bugs in ELF, which were causing `ld' to
           continue to seg-fault in a lot of non-trivial cases.

       (*) Fixed a seg-fault in the label manager.

       (*) Stopped FBLD and FBSTP from _requiring_ the TWORD keyword, which
           is the only option for BCD loads/stores in any case.

       (*) Ensured FLDCW, FSTCW and FSTSW can cope with the WORD keyword,
           if anyone bothers to provide it. Previously they complained
           unless no keyword at all was present.

       (*) Some forms of FDIV/FDIVR and FSUB/FSUBR were still inverted: a
           vestige of a bug that I thought had been fixed in 0.92. This was
           fixed, hopefully for good this time...

       (*) Another minor phase error (insofar as a phase error can _ever_
           be minor) fixed, this one occurring in code of the form

         rol ax,forward_reference
         forward_reference equ 1

       (*) The number supplied to TIMES is now sanity-checked for
           positivity, and also may be greater than 64K (which previously
           didn't work on 16-bit systems).

       (*) Added Watcom C makefiles, and misc/pmw.bat, donated by Dominik
           Behr.

       (*) Added the INCBIN pseudo-opcode.

       (*) Due to the advent of the preprocessor, the [INCLUDE] and [INC]
           directives have become obsolete. They are still supported in
           this version, with a warning, but won't be in the next.

       (*) Fixed a bug in OBJ format, which caused incorrect object records
           to be output when absolute labels were made global.

       (*) Updates to RDOFF subdirectory, and changes to outrdf.c.

C.3.5 Version 0.93 released January 1997

       This release went out in a great hurry after semi-crippling bugs
       were found in 0.92.

       (*) Really _did_ fix the stack overflows this time. *blush*

       (*) Had problems with EA instruction sizes changing between passes,
           when an offset contained a forward reference and so 4 bytes were
           allocated for the offset in pass one; by pass two the symbol had
           been defined and happened to be a small absolute value, so only
           1 byte got allocated, causing instruction size mismatch between
           passes and hence incorrect address calculations. Fixed.

       (*) Stupid bug in the revised ELF section generation fixed
           (associated string-table section for .symtab was hard-coded as
           7, even when this didn't fit with the real section table). Was
           causing `ld' to seg-fault under Linux.

       (*) Included a new Borland C makefile, Makefile.bc2, donated by Fox
           Cutter <lmb@comtch.iea.com>.

C.3.6 Version 0.92 released January 1997

       (*) The FDIVP/FDIVRP and FSUBP/FSUBRP pairs had been inverted: this
           was fixed. This also affected the LCC driver.

       (*) Fixed a bug regarding 32-bit effective addresses of the form
           `[other_register+ESP]'.

       (*) Documentary changes, notably documentation of the fact that
           Borland Win32 compilers use `obj' rather than `win32' object
           format.

       (*) Fixed the COMENT record in OBJ files, which was formatted
           incorrectly.

       (*) Fixed a bug causing segfaults in large RDF files.

       (*) OBJ format now strips initial periods from segment and group
           definitions, in order to avoid complications with the local
           label syntax.

       (*) Fixed a bug in disassembling far calls and jumps in NDISASM.

       (*) Added support for user-defined sections in COFF and ELF files.

       (*) Compiled the DOS binaries with a sensible amount of stack, to
           prevent stack overflows on any arithmetic expression containing
           parentheses.

       (*) Fixed a bug in handling of files that do not terminate in a
           newline.

C.3.7 Version 0.91 released November 1996

       (*) Loads of bug fixes.

       (*) Support for RDF added.

       (*) Support for DBG debugging format added.

       (*) Support for 32-bit extensions to Microsoft OBJ format added.

       (*) Revised for Borland C: some variable names changed, makefile
           added.

       (*) LCC support revised to actually work.

       (*) JMP/CALL NEAR/FAR notation added.

       (*) `a16', `o16', `a32' and `o32' prefixes added.

       (*) Range checking on short jumps implemented.

       (*) MMX instruction support added.

       (*) Negative floating point constant support added.

       (*) Memory handling improved to bypass 64K barrier under DOS.

       (*) `$' prefix to force treatment of reserved words as identifiers
           added.

       (*) Default-size mechanism for object formats added.

       (*) Compile-time configurability added.

       (*) `#', `@', `~' and c{?} are now valid characters in labels.

       (*) `-e' and `-k' options in NDISASM added.

C.3.8 Version 0.90 released October 1996

       First release version. First support for object file output. Other
       changes from previous version (0.3x) too numerous to document.
回复

使用道具 举报

QQ|Archiver|小黑屋|技术宅的结界 ( 滇ICP备16008837号 )|网站地图

GMT+8, 2024-4-27 08:37 , Processed in 0.088176 second(s), 31 queries , Gzip On.

Powered by Discuz! X3.5

© 2001-2024 Discuz! Team.

快速回复 返回顶部 返回列表