2 ## --------------------------------------------------------------------------
4 ## Copyright 1996-2012 The NASM Authors - All Rights Reserved
5 ## See the file AUTHORS included with the NASM distribution for
6 ## the specific copyright holders.
8 ## Redistribution and use in source and binary forms, with or without
9 ## modification, are permitted provided that the following
10 ## conditions are met:
12 ## * Redistributions of source code must retain the above copyright
13 ## notice, this list of conditions and the following disclaimer.
14 ## * Redistributions in binary form must reproduce the above
15 ## copyright notice, this list of conditions and the following
16 ## disclaimer in the documentation and/or other materials provided
17 ## with the distribution.
19 ## THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND
20 ## CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES,
21 ## INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
22 ## MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
23 ## DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
24 ## CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
25 ## SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
26 ## NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27 ## LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 ## HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
29 ## CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
30 ## OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE,
31 ## EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
33 ## --------------------------------------------------------------------------
38 # Parse insns.dat and produce generated source code files
40 # Opcode prefixes which need their own opcode tables
41 # LONGER PREFIXES FIRST!
42 @disasm_prefixes = qw(0F24 0F25 0F38 0F3A 0F7A 0FA6 0FA7 0F);
44 # This should match MAX_OPERANDS from nasm.h
47 # Add VEX/XOP prefixes
48 @vex_class = ( 'vex', 'xop' );
49 $vex_classes = scalar(@vex_class);
52 for ($c = 0; $c < $vex_classes; $c++) {
53 $vexmap{$vex_class[$c]} = $c;
54 for ($m = 0; $m < 32; $m++) {
55 for ($p = 0; $p < 4; $p++) {
56 push(@vexlist, sprintf("%s%02X%01X", $vex_class[$c], $m, $p));
60 @disasm_prefixes = (@vexlist, @disasm_prefixes);
62 @bytecode_count = (0) x
256;
64 print STDERR
"Reading insns.dat...\n";
68 foreach $arg ( @ARGV ) {
69 if ( $arg =~ /^\-/ ) {
70 if ( $arg =~ /^\-([abdin])$/ ) {
73 die "$0: Unknown option: ${arg}\n";
80 $fname = "insns.dat" unless $fname = $args[0];
81 open (F
, $fname) || die "unable to open $fname";
91 next if ( /^\s*(\;.*|)$/ ); # comments or blank lines
93 unless (/^\s*(\S+)\s+(\S+)\s+(\S+|\[.*\])\s+(\S+)\s*$/) {
94 warn "line $line does not contain four fields\n";
97 @fields = ($1, $2, $3, $4);
98 @field_list = ([@fields, 0]);
100 if ($fields[1] =~ /\*/) {
101 # This instruction has relaxed form(s)
102 if ($fields[2] !~ /^\[/) {
103 warn "line $line has an * operand but uses raw bytecodes\n";
108 @ops = split(/,/, $fields[1]);
109 for ($oi = 0; $oi < scalar @ops; $oi++) {
110 if ($ops[$oi] =~ /\*$/) {
112 warn "line $line has a first operand with a *\n";
119 for ($oi = 1; $oi < (1 << scalar @ops); $oi++) {
120 if (($oi & ~$opmask) == 0) {
123 for ($oj = 0; $oj < scalar(@ops); $oj++) {
125 push(@xops, $ops[$oj]);
129 push(@field_list, [$fields[0], join(',', @xops),
130 $fields[2], $fields[3], $oi]);
135 foreach $fptr (@field_list) {
137 ($formatted, $nd) = format_insn
(@fields);
140 $aname = "aa_$fields[0]";
141 push @
$aname, $formatted;
143 if ( $fields[0] =~ /cc$/ ) {
144 # Conditional instruction
145 $k_opcodes_cc{$fields[0]}++;
147 # Unconditional instruction
148 $k_opcodes{$fields[0]}++;
150 if ($formatted && !$nd) {
151 push @big, $formatted;
152 my @sseq = startseq
($fields[2], $fields[4]);
154 if (!defined($dinstables{$i})) {
155 $dinstables{$i} = [];
157 push(@
{$dinstables{$i}}, $#big);
166 # Generate the bytecode array. At this point, @bytecode_list contains
167 # the full set of bytecodes.
170 # Sort by descending length
171 @bytecode_list = sort { scalar(@
$b) <=> scalar(@
$a) } @bytecode_list;
172 @bytecode_array = ();
175 foreach $bl (@bytecode_list) {
176 my $h = hexstr
(@
$bl);
177 next if (defined($bytecode_pos{$h}));
179 push(@bytecode_array, $bl);
181 $bytecode_pos{$h} = $bytecode_next;
186 undef @bytecode_list;
188 @opcodes = sort keys(%k_opcodes);
189 @opcodes_cc = sort keys(%k_opcodes_cc);
191 if ( !defined($output) || $output eq 'b') {
192 print STDERR
"Writing insnsb.c...\n";
196 print B
"/* This file auto-generated from insns.dat by insns.pl" .
197 " - don't edit it */\n\n";
199 print B
"#include \"nasm.h\"\n";
200 print B
"#include \"insns.h\"\n\n";
202 print B
"const uint8_t nasm_bytecodes[$bytecode_next] = {\n";
205 foreach $bl (@bytecode_array) {
206 printf B
" /* %5d */ ", $p;
217 print B
" * Bytecode frequencies (including reuse):\n";
219 for ($i = 0; $i < 32; $i++) {
221 for ($j = 0; $j < 256; $j += 32) {
222 print B
" |" if ($j);
223 printf B
" %3o:%4d", $i+$j, $bytecode_count[$i+$j];
232 if ( !defined($output) || $output eq 'a' ) {
233 print STDERR
"Writing insnsa.c...\n";
237 print A
"/* This file auto-generated from insns.dat by insns.pl" .
238 " - don't edit it */\n\n";
240 print A
"#include \"nasm.h\"\n";
241 print A
"#include \"insns.h\"\n\n";
243 foreach $i (@opcodes, @opcodes_cc) {
244 print A
"static const struct itemplate instrux_${i}[] = {\n";
246 foreach $j (@
$aname) {
247 print A
" ", codesubst
($j), "\n";
249 print A
" ITEMPLATE_END\n};\n\n";
251 print A
"const struct itemplate * const nasm_instructions[] = {\n";
252 foreach $i (@opcodes, @opcodes_cc) {
253 print A
" instrux_${i},\n";
260 if ( !defined($output) || $output eq 'd' ) {
261 print STDERR
"Writing insnsd.c...\n";
265 print D
"/* This file auto-generated from insns.dat by insns.pl" .
266 " - don't edit it */\n\n";
268 print D
"#include \"nasm.h\"\n";
269 print D
"#include \"insns.h\"\n\n";
271 print D
"static const struct itemplate instrux[] = {\n";
274 printf D
" /* %4d */ %s\n", $n++, codesubst
($j);
278 foreach $h (sort(keys(%dinstables))) {
279 next if ($h eq ''); # Skip pseudo-instructions
280 print D
"\nstatic const struct itemplate * const itable_${h}[] = {\n";
281 foreach $j (@
{$dinstables{$h}}) {
282 print D
" instrux + $j,\n";
288 foreach $h (@disasm_prefixes, '') {
289 for ($c = 0; $c < 256; $c++) {
290 $nn = sprintf("%s%02X", $h, $c);
291 if ($is_prefix{$nn} || defined($dinstables{$nn})) {
292 # At least one entry in this prefix table
293 push(@prefix_list, $h);
300 foreach $h (@prefix_list) {
302 print D
"static " unless ($h eq '');
303 print D
"const struct disasm_index ";
304 print D
($h eq '') ?
'itable' : "itable_$h";
305 print D
"[256] = {\n";
306 for ($c = 0; $c < 256; $c++) {
307 $nn = sprintf("%s%02X", $h, $c);
308 if ($is_prefix{$nn}) {
309 die "$fname: ambiguous decoding of $nn\n"
310 if (defined($dinstables{$nn}));
311 printf D
" /* 0x%02x */ { itable_%s, -1 },\n", $c, $nn;
312 } elsif (defined($dinstables{$nn})) {
313 printf D
" /* 0x%02x */ { itable_%s, %u },\n", $c,
314 $nn, scalar(@
{$dinstables{$nn}});
316 printf D
" /* 0x%02x */ { NULL, 0 },\n", $c;
322 printf D
"\nconst struct disasm_index * const itable_vex[%d][32][4] =\n",
325 for ($c = 0; $c < $vex_classes; $c++) {
327 for ($m = 0; $m < 32; $m++) {
329 for ($p = 0; $p < 4; $p++) {
330 $vp = sprintf("%s%02X%01X", $vex_class[$c], $m, $p);
332 ($is_prefix{$vp} ?
sprintf("itable_%s,", $vp) : 'NULL,');
343 if ( !defined($output) || $output eq 'i' ) {
344 print STDERR
"Writing insnsi.h...\n";
348 print I
"/* This file is auto-generated from insns.dat by insns.pl" .
349 " - don't edit it */\n\n";
350 print I
"/* This file in included by nasm.h */\n\n";
352 print I
"/* Instruction names */\n\n";
353 print I
"#ifndef NASM_INSNSI_H\n";
354 print I
"#define NASM_INSNSI_H 1\n\n";
355 print I
"enum opcode {\n";
357 foreach $i (@opcodes, @opcodes_cc) {
358 print I
"\tI_${i},\n";
360 $len++ if ( $i =~ /cc$/ ); # Condition codes can be 3 characters long
361 $maxlen = $len if ( $len > $maxlen );
363 print I
"\tI_none = -1\n";
365 print I
"#define MAX_INSLEN ", $maxlen, "\n";
366 print I
"#define FIRST_COND_OPCODE I_", $opcodes_cc[0], "\n\n";
367 print I
"#endif /* NASM_INSNSI_H */\n";
372 if ( !defined($output) || $output eq 'n' ) {
373 print STDERR
"Writing insnsn.c...\n";
377 print N
"/* This file is auto-generated from insns.dat by insns.pl" .
378 " - don't edit it */\n\n";
379 print N
"#include \"tables.h\"\n\n";
381 print N
"const char * const nasm_insn_names[] = {";
383 foreach $i (@opcodes, @opcodes_cc) {
384 print N
"," if ( !$first );
387 $ilower =~ s/cc$//; # Remove conditional cc suffix
388 $ilower =~ tr/A-Z/a-z/; # Change to lower case (Perl 4 compatible)
389 print N
"\n\t\"${ilower}\"";
395 printf STDERR
"Done: %d instructions\n", $insns;
397 # Count primary bytecodes, for statistics
398 sub count_bytecodes
(@
) {
400 foreach my $bc (@_) {
405 $bytecode_count[$bc]++;
406 if ($bc >= 01 && $bc <= 04) {
408 } elsif (($bc & ~03) == 010) {
410 } elsif (($bc & ~013) == 0144) {
412 } elsif ($bc == 0172 || $bc == 0173) {
414 } elsif ($bc >= 0260 && $bc <= 0270) {
416 } elsif ($bc == 0330) {
422 sub format_insn
($$$$$) {
423 my ($opcode, $operands, $codes, $flags, $relax) = @_;
426 my $op, @ops, $opp, @opx, @oppx;
428 return (undef, undef) if $operands eq "ignore";
430 # format the operands
431 $operands =~ s/\*//g;
432 $operands =~ s/:/|colon,/g;
434 if ($operands ne 'void') {
435 foreach $op (split(/,/, $operands)) {
437 foreach $opp (split(/\|/, $op)) {
439 if ($opp =~ s/(?<=\D)(8|16|32|64|80|128|256)$//) {
440 push(@oppx, "bits$1");
442 $opp =~ s/^mem$/memory/;
443 $opp =~ s/^memory_offs$/mem_offs/;
444 $opp =~ s/^imm$/immediate/;
445 $opp =~ s/^([a-z]+)rm$/rm_$1/;
446 $opp =~ s/^rm$/rm_gpr/;
447 $opp =~ s/^reg$/reg_gpr/;
448 push(@opx, $opp, @oppx);
450 $op = join('|', @opx);
456 while (scalar(@ops) < $MAX_OPERANDS) {
459 $operands = join(',', @ops);
460 $operands =~ tr/a-z/A-Z/;
463 $flags =~ s/,/|IF_/g;
464 $flags =~ s/(\|IF_ND|IF_ND\|)//, $nd = 1 if $flags =~ /IF_ND/;
465 $flags = "IF_" . $flags;
467 @bytecode = (decodify
($codes, $relax), 0);
468 push(@bytecode_list, [@bytecode]);
469 $codes = hexstr
(@bytecode);
470 count_bytecodes
(@bytecode);
472 ("{I_$opcode, $num, {$operands}, \@\@CODES-$codes\@\@, $flags},", $nd);
476 # Look for @@CODES-xxx@@ sequences and replace them with the appropriate
477 # offset into nasm_bytecodes
483 while ($s =~ /\@\@CODES-([0-9A-F]+)\@\@/) {
484 my $pos = $bytecode_pos{$1};
485 if (!defined($pos)) {
486 die "$fname: no position assigned to byte code $1\n";
488 $s = $` . "nasm_bytecodes+${pos}" . "$'";
494 my ($prefix, @list) = @_;
499 push(@l, sprintf("%s%02X", $prefix, $x));
506 # Turn a code string into a sequence of bytes
509 # Although these are C-syntax strings, by convention they should have
510 # only octal escapes (for directives) and hexadecimal escapes
511 # (for verbatim bytes)
512 my($codestr, $relax) = @_;
514 if ($codestr =~ /^\s*\[([^\]]*)\]\s*$/) {
515 return byte_code_compile($1, $relax);
521 unless ($codestr eq 'ignore') {
523 if ($c =~ /^\\x([0-9a-f]+)(.*)$/i) {
524 push(@codes, hex $1);
527 } elsif ($c =~ /^\\([0-7]{1,3})(.*)$/) {
528 push(@codes, oct $1);
532 die "$fname: unknown code format in \"$codestr\"\n";
540 # Turn a numeric list into a hex string
546 $s .= sprintf("%02X", $c);
551 # Here we determine the range of possible starting bytes for a given
552 # instruction. We need only consider the codes:
553 # \[1234] mean literal bytes, of course
554 # \1[0123] mean byte plus register value
555 # \330 means byte plus condition code
556 # \0 or \340 mean give up and return empty set
557 # \34[4567] mean PUSH/POP of segment registers: special case
558 # \17[234] skip is4 control byte
559 # \26x \270 skip VEX control bytes
561 my ($codestr, $relax) = @_;
568 @codes = decodify($codestr, $relax);
570 while ($c0 = shift(@codes)) {
572 if ($c0 >= 01 && $c0 <= 04) {
576 if ($c0 >= 01 && $c0 <= 04) {
578 $fbs .= sprintf("%02X", shift(@codes));
586 foreach $pfx (@disasm_prefixes) {
587 if (substr($fbs, 0, length($pfx)) eq $pfx) {
589 $fbs = substr($fbs, length($pfx));
595 return ($prefix.substr($fbs,0,2));
598 unshift(@codes, $c0);
599 } elsif ($c0 >= 010 && $c0 <= 013) {
600 return addprefix($prefix, $c1..($c1+7));
601 } elsif (($c0 & ~013) == 0144) {
602 return addprefix($prefix, $c1, $c1|2);
603 } elsif ($c0 == 0330) {
604 return addprefix($prefix, $c1..($c1+15));
605 } elsif ($c0 == 0 || $c0 == 0340) {
607 } elsif (($c0 & ~3) == 0260 || $c0 == 0270) {
610 $wlp = shift(@codes);
613 $prefix .= sprintf('%s%02X%01X', $vex_class[$c], $m, $wlp & 3);
614 } elsif ($c0 >= 0172 && $c0 <= 173) {
615 shift(@codes); # Skip is4 control byte
617 # We really need to be able to distinguish "forbidden"
618 # and "ignorable" codes here
625 # This function takes a series of byte codes in a format which is more
626 # typical of the Intel documentation, and encode it.
628 # The format looks like:
630 # [operands: opcodes]
632 # The operands word lists the order of the operands:
634 # r = register field in the modr/m
638 # s = register field of is4/imz2 field
639 # - = implicit (unencoded) operand
641 # For an operand that should be filled into more than one field,
642 # enter it as e.g. "r+v".
644 sub byte_code_compile($$) {
645 my($str, $relax) = @_;
657 'ib,u' => 024, # Unsigned imm8
659 'ib,s' => 0274, # imm8 sign-extended to opsize or bits
660 'iwd' => 034, # imm16 or imm32, depending on opsize
662 'id,s' => 0254, # imm32 sign-extended to 64 bits
663 'iwdq' => 044, # imm16/32/64, depending on addrsize
667 'rel' => 064, # 16 or 32 bit relative operand
672 'o16' => 0320, # 16-bit operand size
673 'o32' => 0321, # 32-bit operand size
674 'odf' => 0322, # Operand size is default
675 'o64' => 0324, # 64-bit operand size requiring REX.W
676 'o64nw' => 0323, # Implied 64-bit operand size (no REX.W)
679 'adf' => 0312, # Address size is default
683 'f2i' => 0332, # F2 prefix, but 66 for operand size is OK
684 'f3i' => 0333, # F3 prefix, but 66 for operand size is OK
693 'nohi' => 0325, # Use spl/bpl/sil/dil even without REX
694 'nof3' => 0326, # No REP 0xF3 prefix permitted
695 'norep' => 0331, # No REP prefix permitted
696 'wait' => 0341, # Needs a wait prefix
698 'jcc8' => 0370, # Match only if Jcc possible with single byte
699 'jmp8' => 0371, # Match only if JMP possible with single byte
700 'jlen' => 0373, # Length of jump
705 # This instruction takes XMM VSIB
710 # This instruction takes YMM VSIB
716 unless ($str =~ /^(([^\s:]*)\:|)\s*(.*\S)\s*$/) {
717 die "$fname: $line: cannot parse: [$str]\n";
723 for ($i = 0; $i < length($opr); $i++) {
724 my $c = substr($opr,$i,1);
738 foreach $op (split(/\s*(?:\s|(?=[\/\\]))/, $opc)) {
739 my $pc = $plain_codes{$op};
744 } elsif ($prefix_ok && $op =~ /^(66|f2|f3|np)$/) {
745 # 66/F2/F3 prefix used as an opcode extension, or np = no prefix
748 } elsif ($op eq 'f2') {
750 } elsif ($op eq 'f3') {
755 } elsif ($op =~ /^[0-9a-f]{2}$/) {
756 if (defined($litix) && $litix+$codes[$litix]+1 == scalar @codes &&
757 $codes[$litix] < 4) {
759 push(@codes, hex $op);
761 $litix = scalar(@codes);
762 push(@codes, 01, hex $op);
765 } elsif ($op eq '/r') {
766 if (!defined($oppos{'r'}) || !defined($oppos{'m'})) {
767 die "$fname: $line: $op requires r and m operands\n";
769 $opex = (($oppos{'m'} & 4) ? 06 : 0) |
770 (($oppos{'r'} & 4) ? 05 : 0);
771 push(@codes, $opex) if ($opex);
772 push(@codes, 0100 + (($oppos{'m'} & 3) << 3) + ($oppos{'r'} & 3));
774 } elsif ($op =~ m:^/([0-7])$:) {
775 if (!defined($oppos{'m'})) {
776 die "$fname: $line: $op requires m operand\n";
778 push(@codes, 06) if ($oppos{'m'} & 4);
779 push(@codes, 0200 + (($oppos{'m'} & 3) << 3) + $1);
781 } elsif ($op =~ /^(vex|xop)(|\..*)$/) {
783 my ($m,$w,$l,$p) = (undef,2,undef,0);
785 my @subops = split(/\./, $op);
786 shift @subops; # Drop prefix
787 foreach $oq (@subops) {
788 if ($oq eq '128' || $oq eq 'l0' || $oq eq 'lz') {
790 } elsif ($oq eq '256' || $oq eq 'l1') {
792 } elsif ($oq eq 'lig') {
794 } elsif ($oq eq 'w0') {
796 } elsif ($oq eq 'w1') {
798 } elsif ($oq eq 'wig') {
800 } elsif ($oq eq 'ww') {
802 } elsif ($oq eq 'p0') {
804 } elsif ($oq eq '66' || $oq eq 'p1') {
806 } elsif ($oq eq 'f3' || $oq eq 'p2') {
808 } elsif ($oq eq 'f2' || $oq eq 'p3') {
810 } elsif ($oq eq '0f') {
812 } elsif ($oq eq '0f38') {
814 } elsif ($oq eq '0f3a') {
816 } elsif ($oq =~ /^m([0-9]+)$/) {
818 } elsif ($oq eq 'nds' || $oq eq 'ndd' || $oq eq 'dds') {
819 if (!defined($oppos{'v'})) {
820 die "$fname: $line: vex.$oq without 'v' operand\n";
824 die "$fname: $line: undefined VEX subcode: $oq\n";
827 if (!defined($m) || !defined($w) || !defined($l) || !defined($p)) {
828 die "$fname: $line: missing fields in VEX specification\n";
830 if (defined($oppos{'v'}) && !$has_nds) {
831 die "$fname: $line: 'v' operand without vex.nds or vex.ndd\n";
833 push(@codes, defined($oppos{'v'}) ? 0260+($oppos{'v'} & 3) : 0270,
834 ($c << 6)+$m, ($w << 4)+($l << 2)+$p);
836 } elsif (defined $imm_codes{$op}) {
838 if ($last_imm lt 'i') {
839 die "$fname: $line: seg without an immediate operand\n";
843 if ($last_imm gt 'j') {
844 die "$fname: $line: too many immediate operands\n";
847 if (!defined($oppos{$last_imm})) {
848 die "$fname: $line: $op without '$last_imm' operand\n";
850 push(@codes, 05) if ($oppos{$last_imm} & 4);
851 push(@codes, $imm_codes{$op} + ($oppos{$last_imm} & 3));
853 } elsif ($op eq '/is4') {
854 if (!defined($oppos{'s'})) {
855 die "$fname: $line: $op without 's' operand\n";
857 if (defined($oppos{'i'})) {
858 push(@codes, 0172, ($oppos{'s'} << 3)+$oppos{'i'});
860 push(@codes, 05) if ($oppos{'s'} & 4);
861 push(@codes, 0174+($oppos{'s'} & 3));
864 } elsif ($op =~ /^\/is4\=([0-9]+)$/) {
866 if (!defined($oppos{'s'})) {
867 die "$fname: $line: $op without 's' operand\n";
869 if ($imm < 0 || $imm > 15) {
870 die "$fname: $line: invalid imm4 value for $op: $imm\n";
872 push(@codes, 0173, ($oppos{'s'} << 4) + $imm);
874 } elsif ($op =~ /^([0-9a-f]{2})\+c$/) {
875 push(@codes, 0330, hex $1);
877 } elsif ($op =~ /^([0-9a-f]{2})\+r$/) {
878 if (!defined($oppos{'r'})) {
879 die "$fname: $line: $op without 'r' operand\n";
881 push(@codes, 05) if ($oppos{'r'} & 4);
882 push(@codes, 010 + ($oppos{'r'} & 3), hex $1);
884 } elsif ($op =~ /^\\([0-7]+|x[0-9a-f]{2})$/) {
885 # Escape to enter literal bytecodes
886 push(@codes, oct $1);
888 die "$fname: $line: unknown operation: $op\n";