3 # insns.pl produce insnsa.c, insnsd.c, insnsi.h, insnsn.c from insns.dat
5 # The Netwide Assembler is copyright (C) 1996 Simon Tatham and
6 # Julian Hall. All rights reserved. The software is
7 # redistributable under the license given in the file "LICENSE"
8 # distributed in the NASM archive.
10 # Opcode prefixes which need their own opcode tables
11 # LONGER PREFIXES FIRST!
12 @disasm_prefixes = qw(0F24 0F25 0F38 0F3A 0F7A 0FA6 0FA7 0F);
14 # This should match MAX_OPERANDS from nasm.h
17 print STDERR
"Reading insns.dat...\n";
21 foreach $arg ( @ARGV ) {
22 if ( $arg =~ /^\-/ ) {
23 if ( $arg =~ /^\-([abdin])$/ ) {
26 die "$0: Unknown option: ${arg}\n";
33 $fname = "insns.dat" unless $fname = $args[0];
34 open (F
, $fname) || die "unable to open $fname";
44 next if ( /^\s*(\;.*|)$/ ); # comments or blank lines
46 unless (/^\s*(\S+)\s+(\S+)\s+(\S+|\[.*\])\s+(\S+)\s*$/) {
47 warn "line $line does not contain four fields\n";
50 @fields = ($1, $2, $3, $4);
51 ($formatted, $nd) = format_insn
(@fields);
54 $aname = "aa_$fields[0]";
55 push @
$aname, $formatted;
57 if ( $fields[0] =~ /cc$/ ) {
58 # Conditional instruction
59 $k_opcodes_cc{$fields[0]}++;
61 # Unconditional instruction
62 $k_opcodes{$fields[0]}++;
64 if ($formatted && !$nd) {
65 push @big, $formatted;
66 my @sseq = startseq
($fields[2]);
68 if (!defined($dinstables{$i})) {
71 push(@
{$dinstables{$i}}, $#big);
79 # Generate the bytecode array. At this point, @bytecode_list contains
80 # the full set of bytecodes.
83 # Sort by descending length
84 @bytecode_list = sort { scalar(@
$b) <=> scalar(@
$a) } @bytecode_list;
88 foreach $bl (@bytecode_list) {
90 next if (defined($bytecode_pos{$h}));
92 push(@bytecode_array, $bl);
94 $bytecode_pos{$h} = $bytecode_next;
101 @opcodes = sort keys(%k_opcodes);
102 @opcodes_cc = sort keys(%k_opcodes_cc);
104 if ( !defined($output) || $output eq 'b') {
105 print STDERR
"Writing insnsb.c...\n";
109 print B
"/* This file auto-generated from insns.dat by insns.pl" .
110 " - don't edit it */\n\n";
112 print B
"#include \"nasm.h\"\n";
113 print B
"#include \"insns.h\"\n\n";
115 print B
"const uint8_t nasm_bytecodes[$bytecode_next] = {\n";
118 foreach $bl (@bytecode_array) {
119 printf B
" /* %5d */ ", $p;
131 if ( !defined($output) || $output eq 'a' ) {
132 print STDERR
"Writing insnsa.c...\n";
136 print A
"/* This file auto-generated from insns.dat by insns.pl" .
137 " - don't edit it */\n\n";
139 print A
"#include \"nasm.h\"\n";
140 print A
"#include \"insns.h\"\n\n";
142 foreach $i (@opcodes, @opcodes_cc) {
143 print A
"static const struct itemplate instrux_${i}[] = {\n";
145 foreach $j (@
$aname) {
146 print A
" ", codesubst
($j), "\n";
148 print A
" ITEMPLATE_END\n};\n\n";
150 print A
"const struct itemplate * const nasm_instructions[] = {\n";
151 foreach $i (@opcodes, @opcodes_cc) {
152 print A
" instrux_${i},\n";
159 if ( !defined($output) || $output eq 'd' ) {
160 print STDERR
"Writing insnsd.c...\n";
164 print D
"/* This file auto-generated from insns.dat by insns.pl" .
165 " - don't edit it */\n\n";
167 print D
"#include \"nasm.h\"\n";
168 print D
"#include \"insns.h\"\n\n";
170 print D
"static const struct itemplate instrux[] = {\n";
173 printf D
" /* %4d */ %s\n", $n++, codesubst
($j);
177 foreach $h (sort(keys(%dinstables))) {
178 print D
"\nstatic const struct itemplate * const itable_${h}[] = {\n";
179 foreach $j (@
{$dinstables{$h}}) {
180 print D
" instrux + $j,\n";
185 foreach $h (@disasm_prefixes, '') {
188 print D
"static " unless ($h eq '');
189 print D
"const struct disasm_index ";
190 print D
($h eq '') ?
'itable' : "itable_$h";
191 print D
"[256] = {\n";
192 for ($c = 0; $c < 256; $c++) {
193 $nn = sprintf("%s%02X", $h, $c);
194 if ($is_prefix{$nn}) {
195 die "$0: ambiguous decoding of $nn\n"
196 if (defined($dinstables{$nn}));
197 printf D
" { itable_%s, -1 },\n", $nn;
198 } elsif (defined($dinstables{$nn})) {
199 printf D
" { itable_%s, %u },\n",
200 $nn, scalar(@
{$dinstables{$nn}});
202 printf D
" { NULL, 0 },\n";
211 if ( !defined($output) || $output eq 'i' ) {
212 print STDERR
"Writing insnsi.h...\n";
216 print I
"/* This file is auto-generated from insns.dat by insns.pl" .
217 " - don't edit it */\n\n";
218 print I
"/* This file in included by nasm.h */\n\n";
220 print I
"/* Instruction names */\n\n";
221 print I
"#ifndef NASM_INSNSI_H\n";
222 print I
"#define NASM_INSNSI_H 1\n\n";
223 print I
"enum opcode {\n";
225 foreach $i (@opcodes, @opcodes_cc) {
226 print I
"\tI_${i},\n";
228 $len++ if ( $i =~ /cc$/ ); # Condition codes can be 3 characters long
229 $maxlen = $len if ( $len > $maxlen );
231 print I
"\tI_none = -1\n";
233 print I
"#define MAX_INSLEN ", $maxlen, "\n";
234 print I
"#define FIRST_COND_OPCODE I_", $opcodes_cc[0], "\n\n";
235 print I
"#endif /* NASM_INSNSI_H */\n";
240 if ( !defined($output) || $output eq 'n' ) {
241 print STDERR
"Writing insnsn.c...\n";
245 print N
"/* This file is auto-generated from insns.dat by insns.pl" .
246 " - don't edit it */\n\n";
247 print N
"#include \"tables.h\"\n\n";
249 print N
"const char * const nasm_insn_names[] = {";
251 foreach $i (@opcodes) {
252 print N
"," if ( !$first );
255 $ilower =~ tr/A-Z/a-z/; # Change to lower case (Perl 4 compatible)
256 print N
"\n\t\"${ilower}\"";
259 print N
"/* Conditional instructions */\n";
260 print N
"const char * const nasm_cond_insn_names[] = {";
262 foreach $i (@opcodes_cc) {
263 print N
"," if ( !$first );
266 $ilower =~ s/cc$//; # Skip cc suffix
267 $ilower =~ tr/A-Z/a-z/; # Change to lower case (Perl 4 compatible)
268 print N
"\n\t\"${ilower}\"";
272 print N
"/* and the corresponding opcodes */\n";
273 print N
"const enum opcode nasm_cond_insn_opcodes[] = {";
275 foreach $i (@opcodes_cc) {
276 print N
"," if ( !$first );
285 printf STDERR
"Done: %d instructions\n", $insns;
288 my ($opcode, $operands, $codes, $flags) = @_;
292 return (undef, undef) if $operands eq "ignore";
294 # format the operands
295 $operands =~ s/:/|colon,/g;
296 $operands =~ s/mem(\d+)/mem|bits$1/g;
297 $operands =~ s/mem/memory/g;
298 $operands =~ s/memory_offs/mem_offs/g;
299 $operands =~ s/imm(\d+)/imm|bits$1/g;
300 $operands =~ s/imm/immediate/g;
301 $operands =~ s/rm(\d+)/rm_gpr|bits$1/g;
302 $operands =~ s/(mmx|xmm|ymm)rm/rm_$1/g;
303 $operands =~ s/\=([0-9]+)/same_as|$1/g;
304 if ($operands eq 'void') {
307 @ops = split(/\,/, $operands);
310 while (scalar(@ops) < $MAX_OPERANDS) {
313 $operands = join(',', @ops);
314 $operands =~ tr/a-z/A-Z/;
317 $flags =~ s/,/|IF_/g;
318 $flags =~ s/(\|IF_ND|IF_ND\|)//, $nd = 1 if $flags =~ /IF_ND/;
319 $flags = "IF_" . $flags;
321 @bytecode = (decodify
($codes), 0);
322 push(@bytecode_list, [@bytecode]);
323 $codes = hexstr
(@bytecode);
325 ("{I_$opcode, $num, {$operands}, \@\@CODES-$codes\@\@, $flags},", $nd);
329 # Look for @@CODES-xxx@@ sequences and replace them with the appropriate
330 # offset into nasm_bytecodes
336 while ($s =~ /\@\@CODES-([0-9A-F]+)\@\@/) {
337 my $pos = $bytecode_pos{$1};
338 if (!defined($pos)) {
339 die "$0: no position assigned to byte code $1\n";
341 $s = $` . "nasm_bytecodes+${pos}" . "$'";
347 my ($prefix, @list) = @_;
352 push(@l, sprintf("%s%02X", $prefix, $x));
359 # Turn a code string into a sequence of bytes
362 # Although these are C-syntax strings, by convention they should have
363 # only octal escapes (for directives) and hexadecimal escapes
364 # (for verbatim bytes)
367 if ($codestr =~ /^\s*\[([^\]]*)\]\s*$/) {
368 return byte_code_compile($1);
375 if ($c =~ /^\\x([0-9a-f]+)(.*)$/i) {
376 push(@codes, hex $1);
379 } elsif ($c =~ /^\\([0-7]{1,3})(.*)$/) {
380 push(@codes, oct $1);
384 die "$0: unknown code format in \"$codestr\"\n";
391 # Turn a numeric list into a hex string
397 $s .= sprintf("%02X", $c);
402 # Here we determine the range of possible starting bytes for a given
403 # instruction. We need only consider the codes:
404 # \1 \2 \3 mean literal bytes, of course
405 # \4 \5 \6 \7 mean PUSH/POP of segment registers: special case
406 # \1[0123] mean byte plus register value
407 # \330 means byte plus condition code
408 # \0 or \340 mean give up and return empty set
417 @codes = decodify($codestr);
419 while ($c0 = shift(@codes)) {
421 if ($c0 == 01 || $c0 == 02 || $c0 == 03) {
425 if ($c0 == 01 || $c0 == 02 || $c0 == 03) {
427 $fbs .= sprintf("%02X", shift(@codes));
435 foreach $pfx (@disasm_prefixes) {
436 if (substr($fbs, 0, length($pfx)) eq $pfx) {
438 $fbs = substr($fbs, length($pfx));
444 return ($prefix.substr($fbs,0,2));
447 unshift(@codes, $c0);
448 } elsif ($c0 == 04) {
449 return addprefix($prefix, 0x07, 0x17, 0x1F);
450 } elsif ($c0 == 05) {
451 return addprefix($prefix, 0xA1, 0xA9);
452 } elsif ($c0 == 06) {
453 return addprefix($prefix, 0x06, 0x0E, 0x16, 0x1E);
454 } elsif ($c0 == 07) {
455 return addprefix($prefix, 0xA0, 0xA8);
456 } elsif ($c0 >= 010 && $c0 <= 013) {
457 return addprefix($prefix, $c1..($c1+7));
458 } elsif (($c0 & ~013) == 0144) {
459 return addprefix($prefix, $c1, $c1|2);
460 } elsif ($c0 == 0330) {
461 return addprefix($prefix, $c1..($c1+15));
462 } elsif ($c0 == 0 || $c0 == 0340) {
464 } elsif (($c0 & ~3) == 0260 || $c0 == 270) {
467 } elsif ($c0 == 0172) {
470 # We really need to be able to distinguish "forbidden"
471 # and "ignorable" codes here
478 # This function takes a series of byte codes in a format which is more
479 # typical of the Intel documentation, and encode it.
481 # The format looks like:
483 # [operands: opcodes]
485 # The operands word lists the order of the operands:
487 # r = register field in the modr/m
490 # d = DREX "dst" field
492 # s = register field of is4/imz2 field
493 # - = implicit (unencoded) operand
495 # For an operand that should be filled into more than one field,
496 # enter it as e.g. "r+v".
498 sub byte_code_compile($) {
508 if ($str =~ /^(\S*)\:\s*(.*\S)\s*$/) {
517 for ($i = 0; $i < length($opr); $i++) {
518 my $c = substr($opr,$i,1);
527 foreach $op (split(/\s*(?:\s|(?=[\/\\]))/, $opc)) {
530 } elsif ($op eq 'o32') {
532 } elsif ($op eq 'o64') { # 64-bit operand size requiring REX.W
534 } elsif ($op eq 'o64i') { # Implied 64-bit operand size (no REX.W)
536 } elsif ($op eq 'a16') {
538 } elsif ($op eq 'a32') {
540 } elsif ($op eq 'a64') {
542 } elsif ($op eq '!osp') {
544 } elsif ($op eq '!asp') {
546 } elsif ($op eq 'rex.l') {
548 } elsif ($op eq 'repe') {
550 } elsif ($prefix_ok && $op =~ /^(66|f2|f3|np)$/) {
551 # 66/F2/F3 prefix used as an opcode extension, or np = no prefix
554 } elsif ($op eq 'f2') {
556 } elsif ($op eq 'f3') {
561 } elsif ($op =~ /^[0-9a-f]{2}$/) {
562 if (defined($litix) && $litix+$codes[$litix]+1 == scalar @codes) {
564 push(@codes, hex $op);
566 $litix = scalar(@codes);
567 push(@codes, 01, hex $op);
570 } elsif ($op eq '/r') {
571 if (!defined($oppos{'r'}) || !defined($oppos{'m'})) {
572 die "$0: $line: $op requires r and m operands\n";
574 push(@codes, 0100 + ($oppos{'m'} << 3) + $oppos{'r'});
576 } elsif ($op =~ m:^/([0-7])$:) {
577 if (!defined($oppos{'m'})) {
578 die "$0: $line: $op requires m operand\n";
580 push(@codes, 0200 + ($oppos{'m'} << 3) + $1);
582 } elsif ($op =~ /^vex(|\..*)$/) {
583 my ($m,$w,$l,$p) = (undef,2,undef,0);
584 foreach $oq (split(/\./, $op)) {
587 } elsif ($oq eq '128' || $oq eq 'l0') {
589 } elsif ($oq eq '256' || $oq eq 'l1') {
591 } elsif ($oq eq 'w0') {
593 } elsif ($oq eq 'w1') {
595 } elsif ($oq eq '66') {
597 } elsif ($oq eq 'f3') {
599 } elsif ($oq eq 'f2') {
601 } elsif ($oq eq '0f') {
603 } elsif ($oq eq '0f38') {
605 } elsif ($oq eq '0f3a') {
607 } elsif ($oq =~ /^m([0-9]+)$/) {
609 } elsif ($oq eq 'nds' || $oq eq 'ndd') {
610 if (!defined($oppos{'v'})) {
611 die "$0: $line: vex.$oq without 'v' operand\n";
614 die "$0: $line: undefined VEX subcode: $oq\n";
617 if (!defined($m) || !defined($w) || !defined($l) || !defined($p)) {
618 die "$0: $line: missing fields in VEX specification\n";
620 push(@codes, defined($oppos{'v'}) ? 0260+$oppos{'v'} : 0270,
621 $m, ($w << 3)+($l << 2)+$p);
623 } elsif ($op =~ /^drex(|..*)$/) {
625 foreach $oq (split(/\./, $op)) {
628 } elsif ($oq eq 'oc0') {
631 die "$0: $line: undefined DREX subcode: $oq\n";
634 if (!defined($oppos{'d'})) {
635 die "$0: $line: DREX without a 'd' operand\n";
637 push(@codes, 0160+$oppos{'d'}+($oc0 ? 4 : 0));
638 } elsif ($op =~ /^(ib\,s|ib|ib\,w|iw|iwd|id|iwdq|rel|rel8|rel16|rel32|iq|seg|ibw|ibd|ibd,s)$/) {
639 if (!defined($oppos{'i'})) {
640 die "$0: $op without 'i' operand\n";
642 if ($op eq 'ib,s') { # Signed imm8
643 push(@codes, 014+$oppos{'i'});
644 } elsif ($op eq 'ib') { # imm8
645 push(@codes, 020+$oppos{'i'});
646 } elsif ($op eq 'ib,u') { # Unsigned imm8
647 push(@codes, 024+$oppos{'i'});
648 } elsif ($op eq 'iw') { # imm16
649 push(@codes, 030+$oppos{'i'});
650 } elsif ($op eq 'iwd') { # imm16 or imm32, depending on opsize
651 push(@codes, 034+$oppos{'i'});
652 } elsif ($op eq 'id') { # imm32
653 push(@codes, 040+$oppos{'i'});
654 } elsif ($op eq 'iwdq') { # imm16/32/64, depending on opsize
655 push(@codes, 044+$oppos{'i'});
656 } elsif ($op eq 'rel8') {
657 push(@codes, 050+$oppos{'i'});
658 } elsif ($op eq 'iq') {
659 push(@codes, 054+$oppos{'i'});
660 } elsif ($op eq 'rel16') {
661 push(@codes, 060+$oppos{'i'});
662 } elsif ($op eq 'rel') { # 16 or 32 bit relative operand
663 push(@codes, 064+$oppos{'i'});
664 } elsif ($op eq 'rel32') {
665 push(@codes, 070+$oppos{'i'});
666 } elsif ($op eq 'seg') {
667 push(@codes, 074+$oppos{'i'});
668 } elsif ($op eq 'ibw') { # imm16 that can be bytified
669 if (!defined($s_pos)) {
670 die "$0: $line: $op without a +s byte\n";
672 $codes[$s_pos] += 0144;
673 push(@codes, 0140+$oppos{'i'});
674 } elsif ($op eq 'ibd') { # imm32 that can be bytified
675 if (!defined($s_pos)) {
676 die "$0: $line: $op without a +s byte\n";
678 $codes[$s_pos] += 0154;
679 push(@codes, 0150+$oppos{'i'});
680 } elsif ($op eq 'ibd,s') {
681 # imm32 that can be bytified, sign extended to 64 bits
682 if (!defined($s_pos)) {
683 die "$0: $line: $op without a +s byte\n";
685 $codes[$s_pos] += 0154;
686 push(@codes, 0250+$oppos{'i'});
689 } elsif ($op eq '/is4') {
690 if (!defined($oppos{'s'})) {
691 die "$0: $line: $op without 's' operand\n";
693 if (defined($oppos{'i'})) {
694 push(@codes, 0172, ($oppos{'s'} << 3)+$oppos{'i'});
696 push(@codes, 0174, $oppos{'s'});
699 } elsif ($op =~ /^\/is4\=([0-9]+)$/) {
701 if (!defined($oppos{'s'})) {
702 die "$0: $line: $op without 's' operand\n";
704 if ($imm < 0 || $imm > 15) {
705 die "$0: $line: invalid imm4 value for $op: $imm\n";
707 push(@codes, 0173, ($oppos{'s'} << 4) + $imm);
709 } elsif ($op =~ /^([0-9a-f]{2})\+s$/) {
710 if (!defined($oppos{'i'})) {
711 die "$0: $op without 'i' operand\n";
713 $s_pos = scalar @codes;
714 push(@codes, $oppos{'i'}, hex $1);
716 } elsif ($op =~ /^([0-9a-f]{2})\+c$/) {
717 push(@codes, 0330, hex $1);
719 } elsif ($op =~ /^\\([0-7]+|x[0-9a-f]{2})$/) {
720 # Escape to enter literal bytecodes
721 push(@codes, oct $1);
723 die "$0: unknown operation: $op\n";