insns.pl

   1 #!/usr/bin/perl
   2 ## --------------------------------------------------------------------------
   3 ##
   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.
   7 ##
   8 ##   Redistribution and use in source and binary forms, with or without
   9 ##   modification, are permitted provided that the following
  10 ##   conditions are met:
  11 ##
  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.
  18 ##
  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.
  32 ##
  33 ## --------------------------------------------------------------------------
  34
  35 #
  36 # insns.pl
  37 #
  38 # Parse insns.dat and produce generated source code files
  39
  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);
  43
  44 # This should match MAX_OPERANDS from nasm.h
  45 $MAX_OPERANDS = 5;
  46
  47 # Add VEX/XOP prefixes
  48 @vex_class = ( 'vex', 'xop' );
  49 $vex_classes = scalar(@vex_class);
  50 @vexlist = ();
  51 %vexmap = ();
  52 for ($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));
  57         }
  58     }
  59 }
  60 @disasm_prefixes = (@vexlist, @disasm_prefixes);
  61
  62 @bytecode_count = (0) x 256;
  63
  64 print STDERR "Reading insns.dat...\n";
  65
  66 @args   = ();
  67 undef $output;
  68 foreach $arg ( @ARGV ) {
  69     if ( $arg =~ /^\-/ ) {
  70         if  ( $arg =~ /^\-([abdin])$/ ) {
  71             $output = $1;
  72         } else {
  73             die "$0: Unknown option: ${arg}\n";
  74         }
  75     } else {
  76         push (@args, $arg);
  77     }
  78 }
  79
  80 $fname = "insns.dat" unless $fname = $args[0];
  81 open (F, $fname) || die "unable to open $fname";
  82
  83 %dinstables = ();
  84 @bytecode_list = ();
  85
  86 $line = 0;
  87 $insns = 0;
  88 while (<F>) {
  89     $line++;
  90     chomp;
  91     next if ( /^\s*(\;.*|)$/ );   # comments or blank lines
  92
  93     unless (/^\s*(\S+)\s+(\S+)\s+(\S+|\[.*\])\s+(\S+)\s*$/) {
  94         warn "line $line does not contain four fields\n";
  95         next;
  96     }
  97     @fields = ($1, $2, $3, $4);
  98     @field_list = ([@fields, 0]);
  99
 100     if ($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";
 104             next;
 105         }
 106
 107         $opmask = 0;
 108         @ops = split(/,/, $fields[1]);
 109         for ($oi = 0; $oi < scalar @ops; $oi++) {
 110             if ($ops[$oi] =~ /\*$/) {
 111                 if ($oi == 0) {
 112                     warn "line $line has a first operand with a *\n";
 113                     next;
 114                 }
 115                 $opmask |= 1 << $oi;
 116             }
 117         }
 118
 119         for ($oi = 1; $oi < (1 << scalar @ops); $oi++) {
 120             if (($oi & ~$opmask) == 0) {
 121                 my @xops = ();
 122                 my $omask = ~$oi;
 123                 for ($oj = 0; $oj < scalar(@ops); $oj++) {
 124                     if ($omask & 1) {
 125                         push(@xops, $ops[$oj]);
 126                     }
 127                     $omask >>= 1;
 128                 }
 129                 push(@field_list, [$fields[0], join(',', @xops),
 130                      $fields[2], $fields[3], $oi]);
 131             }
 132         }
 133     }
 134
 135     foreach $fptr (@field_list) {
 136         @fields = @$fptr;
 137         ($formatted, $nd) = format_insn(@fields);
 138         if ($formatted) {
 139             $insns++;
 140             $aname = "aa_$fields[0]";
 141             push @$aname, $formatted;
 142         }
 143         if ( $fields[0] =~ /cc$/ ) {
 144             # Conditional instruction
 145             $k_opcodes_cc{$fields[0]}++;
 146         } else {
 147             # Unconditional instruction
 148             $k_opcodes{$fields[0]}++;
 149         }
 150         if ($formatted && !$nd) {
 151             push @big, $formatted;
 152             my @sseq = startseq($fields[2], $fields[4]);
 153             foreach $i (@sseq) {
 154                 if (!defined($dinstables{$i})) {
 155                     $dinstables{$i} = [];
 156                 }
 157                 push(@{$dinstables{$i}}, $#big);
 158             }
 159         }
 160     }
 161 }
 162
 163 close F;
 164
 165 #
 166 # Generate the bytecode array.  At this point, @bytecode_list contains
 167 # the full set of bytecodes.
 168 #
 169
 170 # Sort by descending length
 171 @bytecode_list = sort { scalar(@$b) <=> scalar(@$a) } @bytecode_list;
 172 @bytecode_array = ();
 173 %bytecode_pos = ();
 174 $bytecode_next = 0;
 175 foreach $bl (@bytecode_list) {
 176     my $h = hexstr(@$bl);
 177     next if (defined($bytecode_pos{$h}));
 178
 179     push(@bytecode_array, $bl);
 180     while ($h ne '') {
 181         $bytecode_pos{$h} = $bytecode_next;
 182         $h = substr($h, 2);
 183         $bytecode_next++;
 184     }
 185 }
 186 undef @bytecode_list;
 187
 188 @opcodes    = sort keys(%k_opcodes);
 189 @opcodes_cc = sort keys(%k_opcodes_cc);
 190
 191 if ( !defined($output) || $output eq 'b') {
 192     print STDERR "Writing insnsb.c...\n";
 193
 194     open B, ">insnsb.c";
 195
 196     print B "/* This file auto-generated from insns.dat by insns.pl" .
 197         " - don't edit it */\n\n";
 198
 199     print B "#include \"nasm.h\"\n";
 200     print B "#include \"insns.h\"\n\n";
 201
 202     print B "const uint8_t nasm_bytecodes[$bytecode_next] = {\n";
 203
 204     $p = 0;
 205     foreach $bl (@bytecode_array) {
 206         printf B "    /* %5d */ ", $p;
 207         foreach $d (@$bl) {
 208             printf B "%#o,", $d;
 209             $p++;
 210         }
 211         printf B "\n";
 212     }
 213     print B "};\n";
 214
 215     print B "\n";
 216     print B "/*\n";
 217     print B " * Bytecode frequencies (including reuse):\n";
 218     print B " *\n";
 219     for ($i = 0; $i < 32; $i++) {
 220         print B " *";
 221         for ($j = 0; $j < 256; $j += 32) {
 222             print B " |" if ($j);
 223             printf B " %3o:%4d", $i+$j, $bytecode_count[$i+$j];
 224         }
 225         print B "\n";
 226     }
 227     print B " */\n";
 228
 229     close B;
 230 }
 231
 232 if ( !defined($output) || $output eq 'a' ) {
 233     print STDERR "Writing insnsa.c...\n";
 234
 235     open A, ">insnsa.c";
 236
 237     print A "/* This file auto-generated from insns.dat by insns.pl" .
 238         " - don't edit it */\n\n";
 239
 240     print A "#include \"nasm.h\"\n";
 241     print A "#include \"insns.h\"\n\n";
 242
 243     foreach $i (@opcodes, @opcodes_cc) {
 244         print A "static const struct itemplate instrux_${i}[] = {\n";
 245         $aname = "aa_$i";
 246         foreach $j (@$aname) {
 247             print A "    ", codesubst($j), "\n";
 248         }
 249         print A "    ITEMPLATE_END\n};\n\n";
 250     }
 251     print A "const struct itemplate * const nasm_instructions[] = {\n";
 252     foreach $i (@opcodes, @opcodes_cc) {
 253         print A "    instrux_${i},\n";
 254     }
 255     print A "};\n";
 256
 257     close A;
 258 }
 259
 260 if ( !defined($output) || $output eq 'd' ) {
 261     print STDERR "Writing insnsd.c...\n";
 262
 263     open D, ">insnsd.c";
 264
 265     print D "/* This file auto-generated from insns.dat by insns.pl" .
 266         " - don't edit it */\n\n";
 267
 268     print D "#include \"nasm.h\"\n";
 269     print D "#include \"insns.h\"\n\n";
 270
 271     print D "static const struct itemplate instrux[] = {\n";
 272     $n = 0;
 273     foreach $j (@big) {
 274         printf D "    /* %4d */ %s\n", $n++, codesubst($j);
 275     }
 276     print D "};\n";
 277
 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";
 283         }
 284         print D "};\n";
 285     }
 286
 287     @prefix_list = ();
 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);
 294                 $is_prefix{$h} = 1;
 295                 last;
 296             }
 297         }
 298     }
 299
 300     foreach $h (@prefix_list) {
 301         print D "\n";
 302         print D "static " unless ($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}});
 315             } else {
 316                 printf D "    /* 0x%02x */ { NULL, 0 },\n", $c;
 317             }
 318         }
 319         print D "};\n";
 320     }
 321
 322     printf D "\nconst struct disasm_index * const itable_vex[%d][32][4] =\n",
 323         $vex_classes;
 324     print D "{\n";
 325     for ($c = 0; $c < $vex_classes; $c++) {
 326         print D "    {\n";
 327         for ($m = 0; $m < 32; $m++) {
 328             print D "        { ";
 329             for ($p = 0; $p < 4; $p++) {
 330                 $vp = sprintf("%s%02X%01X", $vex_class[$c], $m, $p);
 331                 printf D "%-15s",
 332                        ($is_prefix{$vp} ? sprintf("itable_%s,", $vp) : 'NULL,');
 333             }
 334             print D "},\n";
 335         }
 336         print D "    },\n";
 337     }
 338     print D "};\n";
 339
 340     close D;
 341 }
 342
 343 if ( !defined($output) || $output eq 'i' ) {
 344     print STDERR "Writing insnsi.h...\n";
 345
 346     open I, ">insnsi.h";
 347
 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";
 351
 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";
 356     $maxlen = 0;
 357     foreach $i (@opcodes, @opcodes_cc) {
 358         print I "\tI_${i},\n";
 359         $len = length($i);
 360         $len++ if ( $i =~ /cc$/ ); # Condition codes can be 3 characters long
 361         $maxlen = $len if ( $len > $maxlen );
 362     }
 363     print I "\tI_none = -1\n";
 364     print I "};\n\n";
 365     print I "#define MAX_INSLEN ", $maxlen, "\n";
 366     print I "#define FIRST_COND_OPCODE I_", $opcodes_cc[0], "\n\n";
 367     print I "#endif /* NASM_INSNSI_H */\n";
 368
 369     close I;
 370 }
 371
 372 if ( !defined($output) || $output eq 'n' ) {
 373     print STDERR "Writing insnsn.c...\n";
 374
 375     open N, ">insnsn.c";
 376
 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";
 380
 381     print N "const char * const nasm_insn_names[] = {";
 382     $first = 1;
 383     foreach $i (@opcodes, @opcodes_cc) {
 384         print N "," if ( !$first );
 385         $first = 0;
 386         $ilower = $i;
 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}\"";
 390     }
 391     print N "\n};\n";
 392     close N;
 393 }
 394
 395 printf STDERR "Done: %d instructions\n", $insns;
 396
 397 # Count primary bytecodes, for statistics
 398 sub count_bytecodes(@) {
 399     my $skip = 0;
 400     foreach my $bc (@_) {
 401         if ($skip) {
 402             $skip--;
 403             next;
 404         }
 405         $bytecode_count[$bc]++;
 406         if ($bc >= 01 && $bc <= 04) {
 407             $skip = $bc;
 408         } elsif (($bc & ~03) == 010) {
 409             $skip = 1;
 410         } elsif (($bc & ~013) == 0144) {
 411             $skip = 1;
 412         } elsif ($bc == 0172 || $bc == 0173) {
 413             $skip = 1;
 414         } elsif ($bc >= 0260 && $bc <= 0270) {
 415             $skip = 2;
 416         } elsif ($bc == 0330) {
 417             $skip = 1;
 418         }
 419     }
 420 }
 421
 422 sub format_insn($$$$$) {
 423     my ($opcode, $operands, $codes, $flags, $relax) = @_;
 424     my $num, $nd = 0;
 425     my @bytecode;
 426     my $op, @ops, $opp, @opx, @oppx;
 427
 428     return (undef, undef) if $operands eq "ignore";
 429
 430     # format the operands
 431     $operands =~ s/\*//g;
 432     $operands =~ s/:/|colon,/g;
 433     @ops = ();
 434     if ($operands ne 'void') {
 435         foreach $op (split(/,/, $operands)) {
 436             if ($op =~ /^\=([0-9]+)$/) {
 437                 $op = "same_as|$1";
 438             } else {
 439                 @opx = ();
 440                 foreach $opp (split(/\|/, $op)) {
 441                     @oppx = ();
 442                     if ($opp =~ /^(.*[^\d])(8|16|32|64|80|128|256)$/) {
 443                         my $ox = $1;
 444                         my $on = $2;
 445                         if ($ox !~ /^(sbyte|sdword|udword)$/) {
 446                             $opp = $ox;
 447                             push(@oppx, "bits$on");
 448                         }
 449                     }
 450                     $opp =~ s/^mem$/memory/;
 451                     $opp =~ s/^memory_offs$/mem_offs/;
 452                     $opp =~ s/^imm$/immediate/;
 453                     $opp =~ s/^([a-z]+)rm$/rm_$1/;
 454                     $opp =~ s/^rm$/rm_gpr/;
 455                     $opp =~ s/^reg$/reg_gpr/;
 456                     push(@opx, $opp, @oppx);
 457                 }
 458                 $op = join('|', @opx);
 459             }
 460             push(@ops, $op);
 461         }
 462     }
 463
 464     $num = scalar(@ops);
 465     while (scalar(@ops) < $MAX_OPERANDS) {
 466         push(@ops, '0');
 467     }
 468     $operands = join(',', @ops);
 469     $operands =~ tr/a-z/A-Z/;
 470
 471     # format the flags
 472     $flags =~ s/,/|IF_/g;
 473     $flags =~ s/(\|IF_ND|IF_ND\|)//, $nd = 1 if $flags =~ /IF_ND/;
 474     $flags = "IF_" . $flags;
 475
 476     @bytecode = (decodify($codes, $relax), 0);
 477     push(@bytecode_list, [@bytecode]);
 478     $codes = hexstr(@bytecode);
 479     count_bytecodes(@bytecode);
 480
 481     ("{I_$opcode, $num, {$operands}, \@\@CODES-$codes\@\@, $flags},", $nd);
 482 }
 483
 484 #
 485 # Look for @@CODES-xxx@@ sequences and replace them with the appropriate
 486 # offset into nasm_bytecodes
 487 #
 488 sub codesubst($) {
 489     my($s) = @_;
 490     my $n;
 491
 492     while ($s =~ /\@\@CODES-([0-9A-F]+)\@\@/) {
 493         my $pos = $bytecode_pos{$1};
 494         if (!defined($pos)) {
 495             die "$fname: no position assigned to byte code $1\n";
 496         }
 497         $s = $` . "nasm_bytecodes+${pos}" . "$'";
 498     }
 499     return $s;
 500 }
 501
 502 sub addprefix ($@) {
 503     my ($prefix, @list) = @_;
 504     my $x;
 505     my @l = ();
 506
 507     foreach $x (@list) {
 508         push(@l, sprintf("%s%02X", $prefix, $x));
 509     }
 510
 511     return @l;
 512 }
 513
 514 #
 515 # Turn a code string into a sequence of bytes
 516 #
 517 sub decodify($$) {
 518   # Although these are C-syntax strings, by convention they should have
 519   # only octal escapes (for directives) and hexadecimal escapes
 520   # (for verbatim bytes)
 521     my($codestr, $relax) = @_;
 522
 523     if ($codestr =~ /^\s*\[([^\]]*)\]\s*$/) {
 524         return byte_code_compile($1, $relax);
 525     }
 526
 527     my $c = $codestr;
 528     my @codes = ();
 529
 530     unless ($codestr eq 'ignore') {
 531         while ($c ne '') {
 532             if ($c =~ /^\\x([0-9a-f]+)(.*)$/i) {
 533                 push(@codes, hex $1);
 534                 $c = $2;
 535                 next;
 536             } elsif ($c =~ /^\\([0-7]{1,3})(.*)$/) {
 537                 push(@codes, oct $1);
 538                 $c = $2;
 539                 next;
 540             } else {
 541                 die "$fname: unknown code format in \"$codestr\"\n";
 542             }
 543         }
 544     }
 545
 546     return @codes;
 547 }
 548
 549 # Turn a numeric list into a hex string
 550 sub hexstr(@) {
 551     my $s = '';
 552     my $c;
 553
 554     foreach $c (@_) {
 555         $s .= sprintf("%02X", $c);
 556     }
 557     return $s;
 558 }
 559
 560 # Here we determine the range of possible starting bytes for a given
 561 # instruction. We need only consider the codes:
 562 # \[1234]      mean literal bytes, of course
 563 # \1[0123]     mean byte plus register value
 564 # \330         means byte plus condition code
 565 # \0 or \340   mean give up and return empty set
 566 # \34[4567]    mean PUSH/POP of segment registers: special case
 567 # \17[234]     skip is4 control byte
 568 # \26x \270    skip VEX control bytes
 569 sub startseq($$) {
 570     my ($codestr, $relax) = @_;
 571     my $word, @range;
 572     my @codes = ();
 573     my $c = $codestr;
 574     my $c0, $c1, $i;
 575     my $prefix = '';
 576
 577     @codes = decodify($codestr, $relax);
 578
 579     while ($c0 = shift(@codes)) {
 580         $c1 = $codes[0];
 581         if ($c0 >= 01 && $c0 <= 04) {
 582             # Fixed byte string
 583             my $fbs = $prefix;
 584             while (1) {
 585                 if ($c0 >= 01 && $c0 <= 04) {
 586                     while ($c0--) {
 587                         $fbs .= sprintf("%02X", shift(@codes));
 588                     }
 589                 } else {
 590                     last;
 591                 }
 592                 $c0 = shift(@codes);
 593             }
 594
 595             foreach $pfx (@disasm_prefixes) {
 596                 if (substr($fbs, 0, length($pfx)) eq $pfx) {
 597                     $prefix = $pfx;
 598                     $fbs = substr($fbs, length($pfx));
 599                     last;
 600                 }
 601             }
 602
 603             if ($fbs ne '') {
 604                 return ($prefix.substr($fbs,0,2));
 605             }
 606
 607             unshift(@codes, $c0);
 608         } elsif ($c0 >= 010 && $c0 <= 013) {
 609             return addprefix($prefix, $c1..($c1+7));
 610         } elsif (($c0 & ~013) == 0144) {
 611             return addprefix($prefix, $c1, $c1|2);
 612         } elsif ($c0 == 0330) {
 613             return addprefix($prefix, $c1..($c1+15));
 614         } elsif ($c0 == 0 || $c0 == 0340) {
 615             return $prefix;
 616         } elsif ($c0 == 0344) {
 617             return addprefix($prefix, 0x06, 0x0E, 0x16, 0x1E);
 618         } elsif ($c0 == 0345) {
 619             return addprefix($prefix, 0x07, 0x17, 0x1F);
 620         } elsif ($c0 == 0346) {
 621             return addprefix($prefix, 0xA0, 0xA8);
 622         } elsif ($c0 == 0347) {
 623             return addprefix($prefix, 0xA1, 0xA9);
 624         } elsif (($c0 & ~3) == 0260 || $c0 == 0270) {
 625             my $c,$m,$wlp;
 626             $m   = shift(@codes);
 627             $wlp = shift(@codes);
 628             $c = ($m >> 6);
 629             $m = $m & 31;
 630             $prefix .= sprintf('%s%02X%01X', $vex_class[$c], $m, $wlp & 3);
 631         } elsif ($c0 >= 0172 && $c0 <= 173) {
 632             shift(@codes);      # Skip is4 control byte
 633         } else {
 634             # We really need to be able to distinguish "forbidden"
 635             # and "ignorable" codes here
 636         }
 637     }
 638     return $prefix;
 639 }
 640
 641 #
 642 # This function takes a series of byte codes in a format which is more
 643 # typical of the Intel documentation, and encode it.
 644 #
 645 # The format looks like:
 646 #
 647 # [operands: opcodes]
 648 #
 649 # The operands word lists the order of the operands:
 650 #
 651 # r = register field in the modr/m
 652 # m = modr/m
 653 # v = VEX "v" field
 654 # i = immediate
 655 # s = register field of is4/imz2 field
 656 # - = implicit (unencoded) operand
 657 #
 658 # For an operand that should be filled into more than one field,
 659 # enter it as e.g. "r+v".
 660 #
 661 sub byte_code_compile($$) {
 662     my($str, $relax) = @_;
 663     my $opr;
 664     my $opc;
 665     my @codes = ();
 666     my $litix = undef;
 667     my %oppos = ();
 668     my $i;
 669     my $op, $oq;
 670     my $opex;
 671
 672     my %imm_codes = (
 673         'ib,s' => 014,  # Signed imm8
 674         'ib' => 020,    # imm8
 675         'ib,u' => 024,  # Unsigned imm8
 676         'iw' => 030,    # imm16
 677         'ibx' => 0274,  # imm8 sign-extended to opsize
 678         'iwd' => 034,   # imm16 or imm32, depending on opsize
 679         'id' => 040,    # imm32
 680         'idx' => 0254,  # imm32 extended to 64 bits
 681         'iwdq' => 044,  # imm16/32/64, depending on opsize
 682         'rel8' => 050,
 683         'iq' => 054,
 684         'rel16' => 060,
 685         'rel' => 064,   # 16 or 32 bit relative operand
 686         'rel32' => 070,
 687         'seg' => 074,
 688         'ibw' => 0140,  # imm16 that can be bytified
 689         'ibd' => 0150,  # imm32 that can be bytified
 690         'ibd,s' => 0250 # imm32 that can be bytified, sign extended to 64 bits
 691     );
 692     my %imm_codes_bytifiers = (
 693         'ibw' => 0144,
 694         'ibd' => 0154,
 695         'ibd,s' => 0154
 696     );
 697     my %plain_codes = (
 698         'o16' => 0320,          # 16-bit operand size
 699         'o32' => 0321,          # 32-bit operand size
 700         'odf' => 0322,          # Operand size is default
 701         'o64' => 0324,          # 64-bit operand size requiring REX.W
 702         'o64nw' => 0323,        # Implied 64-bit operand size (no REX.W)
 703         'a16' => 0310,
 704         'a32' => 0311,
 705         'adf' => 0312,          # Address size is default
 706         'a64' => 0313,
 707         '!osp' => 0364,
 708         '!asp' => 0365,
 709         'f2i' => 0332,          # F2 prefix, but 66 for operand size is OK
 710         'f3i' => 0333,          # F3 prefix, but 66 for operand size is OK
 711         'pushseg' => 0344,
 712         'popseg' => 0345,
 713         'pushseg2' => 0346,
 714         'popseg2' => 0347,
 715         'mustrep' => 0336,
 716         'mustrepne' => 0337,
 717         'rex.l' => 0334,
 718         'norexb' => 0314,
 719         'norexx' => 0315,
 720         'norexr' => 0316,
 721         'norexw' => 0317,
 722         'repe' => 0335,
 723         'nohi' => 0325,         # Use spl/bpl/sil/dil even without REX
 724         'wait' => 0341,         # Needs a wait prefix
 725         'resb' => 0340,
 726         'jcc8' => 0370,         # Match only if Jcc possible with single byte
 727         'jmp8' => 0371,         # Match only if JMP possible with single byte
 728         'jlen' => 0373,         # Length of jump
 729         'nohle' => 0264,
 730         'hlexr' => 0265,
 731         'hlenl' => 0266,
 732         'hle' => 0267,
 733         # This instruction takes XMM VSIB
 734         'vsibx' => 0374,
 735         'vm32x' => 0374,
 736         'vm64x' => 0374,
 737         # This instruction takes YMM VSIB
 738         'vsiby' => 0375,
 739         'vm32y' => 0375,
 740         'vm64y' => 0375
 741     );
 742
 743     unless ($str =~ /^(([^\s:]*)\:|)\s*(.*\S)\s*$/) {
 744         die "$fname: $line: cannot parse: [$str]\n";
 745     }
 746     $opr = "\L$2";
 747     $opc = "\L$3";
 748
 749     my $op = 0;
 750     for ($i = 0; $i < length($opr); $i++) {
 751         my $c = substr($opr,$i,1);
 752         if ($c eq '+') {
 753             $op--;
 754         } else {
 755             if ($relax & 1) {
 756                 $op--;
 757             }
 758             $relax >>= 1;
 759             $oppos{$c} = $op++;
 760         }
 761     }
 762
 763     my $last_imm = 'h';
 764     my $prefix_ok = 1;
 765     foreach $op (split(/\s*(?:\s|(?=[\/\\]))/, $opc)) {
 766         my $pc = $plain_codes{$op};
 767
 768         if (defined $pc) {
 769             # Plain code
 770             push(@codes, $pc);
 771         } elsif ($prefix_ok && $op =~ /^(66|f2|f3|np)$/) {
 772             # 66/F2/F3 prefix used as an opcode extension, or np = no prefix
 773             if ($op eq '66') {
 774                 push(@codes, 0361);
 775             } elsif ($op eq 'f2') {
 776                 push(@codes, 0362);
 777             } elsif ($op eq 'f3') {
 778                 push(@codes, 0363);
 779             } else {
 780                 push(@codes, 0360);
 781             }
 782         } elsif ($op =~ /^[0-9a-f]{2}$/) {
 783             if (defined($litix) && $litix+$codes[$litix]+1 == scalar @codes &&
 784                 $codes[$litix] < 4) {
 785                 $codes[$litix]++;
 786                 push(@codes, hex $op);
 787             } else {
 788                 $litix = scalar(@codes);
 789                 push(@codes, 01, hex $op);
 790             }
 791             $prefix_ok = 0;
 792         } elsif ($op eq '/r') {
 793             if (!defined($oppos{'r'}) || !defined($oppos{'m'})) {
 794                 die "$fname: $line: $op requires r and m operands\n";
 795             }
 796             $opex = (($oppos{'m'} & 4) ? 06 : 0) |
 797                 (($oppos{'r'} & 4) ? 05 : 0);
 798             push(@codes, $opex) if ($opex);
 799             push(@codes, 0100 + (($oppos{'m'} & 3) << 3) + ($oppos{'r'} & 3));
 800             $prefix_ok = 0;
 801         } elsif ($op =~ m:^/([0-7])$:) {
 802             if (!defined($oppos{'m'})) {
 803                 die "$fname: $line: $op requires m operand\n";
 804             }
 805             push(@codes, 06) if ($oppos{'m'} & 4);
 806             push(@codes, 0200 + (($oppos{'m'} & 3) << 3) + $1);
 807             $prefix_ok = 0;
 808         } elsif ($op =~ /^(vex|xop)(|\..*)$/) {
 809             my $c = $vexmap{$1};
 810             my ($m,$w,$l,$p) = (undef,2,undef,0);
 811             my $has_nds = 0;
 812             my @subops = split(/\./, $op);
 813             shift @subops;      # Drop prefix
 814                 foreach $oq (@subops) {
 815                     if ($oq eq '128' || $oq eq 'l0' || $oq eq 'lz') {
 816                         $l = 0;
 817                     } elsif ($oq eq '256' || $oq eq 'l1') {
 818                         $l = 1;
 819                     } elsif ($oq eq 'lig') {
 820                         $l = 2;
 821                     } elsif ($oq eq 'w0') {
 822                         $w = 0;
 823                     } elsif ($oq eq 'w1') {
 824                         $w = 1;
 825                     } elsif ($oq eq 'wig') {
 826                         $w = 2;
 827                     } elsif ($oq eq 'ww') {
 828                         $w = 3;
 829                     } elsif ($oq eq 'p0') {
 830                         $p = 0;
 831                     } elsif ($oq eq '66' || $oq eq 'p1') {
 832                         $p = 1;
 833                     } elsif ($oq eq 'f3' || $oq eq 'p2') {
 834                         $p = 2;
 835                     } elsif ($oq eq 'f2' || $oq eq 'p3') {
 836                         $p = 3;
 837                     } elsif ($oq eq '0f') {
 838                         $m = 1;
 839                     } elsif ($oq eq '0f38') {
 840                         $m = 2;
 841                     } elsif ($oq eq '0f3a') {
 842                         $m = 3;
 843                     } elsif ($oq =~ /^m([0-9]+)$/) {
 844                         $m = $1+0;
 845                     } elsif ($oq eq 'nds' || $oq eq 'ndd' || $oq eq 'dds') {
 846                         if (!defined($oppos{'v'})) {
 847                             die "$fname: $line: vex.$oq without 'v' operand\n";
 848                         }
 849                         $has_nds = 1;
 850                     } else {
 851                         die "$fname: $line: undefined VEX subcode: $oq\n";
 852                     }
 853                 }
 854             if (!defined($m) || !defined($w) || !defined($l) || !defined($p)) {
 855                 die "$fname: $line: missing fields in VEX specification\n";
 856             }
 857             if (defined($oppos{'v'}) && !$has_nds) {
 858                 die "$fname: $line: 'v' operand without vex.nds or vex.ndd\n";
 859             }
 860             push(@codes, defined($oppos{'v'}) ? 0260+($oppos{'v'} & 3) : 0270,
 861                  ($c << 6)+$m, ($w << 4)+($l << 2)+$p);
 862             $prefix_ok = 0;
 863         } elsif (defined $imm_codes{$op}) {
 864             if ($op eq 'seg') {
 865                 if ($last_imm lt 'i') {
 866                     die "$fname: $line: seg without an immediate operand\n";
 867                 }
 868             } else {
 869                 $last_imm++;
 870                 if ($last_imm gt 'j') {
 871                     die "$fname: $line: too many immediate operands\n";
 872                 }
 873             }
 874             if (!defined($oppos{$last_imm})) {
 875                 die "$fname: $line: $op without '$last_imm' operand\n";
 876             }
 877             push(@codes, 05) if ($oppos{$last_imm} & 4);
 878             push(@codes, $imm_codes{$op} + ($oppos{$last_imm} & 3));
 879             if (defined $imm_codes_bytifiers{$op}) {
 880                 if (!defined($s_pos)) {
 881                     die "$fname: $line: $op without a +s byte\n";
 882                 }
 883                 $codes[$s_pos] += $imm_codes_bytifiers{$op};
 884             }
 885             $prefix_ok = 0;
 886         } elsif ($op eq '/is4') {
 887             if (!defined($oppos{'s'})) {
 888                 die "$fname: $line: $op without 's' operand\n";
 889             }
 890             if (defined($oppos{'i'})) {
 891                 push(@codes, 0172, ($oppos{'s'} << 3)+$oppos{'i'});
 892             } else {
 893                 push(@codes, 05) if ($oppos{'s'} & 4);
 894                 push(@codes, 0174+($oppos{'s'} & 3));
 895             }
 896             $prefix_ok = 0;
 897         } elsif ($op =~ /^\/is4\=([0-9]+)$/) {
 898             my $imm = $1;
 899             if (!defined($oppos{'s'})) {
 900                 die "$fname: $line: $op without 's' operand\n";
 901             }
 902             if ($imm < 0 || $imm > 15) {
 903                 die "$fname: $line: invalid imm4 value for $op: $imm\n";
 904             }
 905             push(@codes, 0173, ($oppos{'s'} << 4) + $imm);
 906             $prefix_ok = 0;
 907         } elsif ($op =~ /^([0-9a-f]{2})\+s$/) {
 908             if (!defined($oppos{'i'})) {
 909                 die "$fname: $line: $op without 'i' operand\n";
 910             }
 911             $s_pos = scalar @codes;
 912             push(@codes, 05) if ($oppos{'i'} & 4);
 913             push(@codes, $oppos{'i'} & 3, hex $1);
 914             $prefix_ok = 0;
 915         } elsif ($op =~ /^([0-9a-f]{2})\+c$/) {
 916             push(@codes, 0330, hex $1);
 917             $prefix_ok = 0;
 918         } elsif ($op =~ /^([0-9a-f]{2})\+r$/) {
 919             if (!defined($oppos{'r'})) {
 920                 die "$fname: $line: $op without 'r' operand\n";
 921             }
 922             push(@codes, 05) if ($oppos{'r'} & 4);
 923             push(@codes, 010 + ($oppos{'r'} & 3), hex $1);
 924             $prefix_ok = 0;
 925         } elsif ($op =~ /^\\([0-7]+|x[0-9a-f]{2})$/) {
 926             # Escape to enter literal bytecodes
 927             push(@codes, oct $1);
 928         } else {
 929             die "$fname: $line: unknown operation: $op\n";
 930         }
 931     }
 932
 933     return @codes;
 934 }