tokhash.pl

   1 #!/usr/bin/perl
   2 #
   3 # Generate a perfect hash for token parsing
   4 #
   5 # Usage: tokenhash.pl insns.dat regs.dat tokens.dat
   6 #
   7
   8 require 'phash.ph';
   9
  10 my($insns_dat, $regs_dat, $tokens_dat) = @ARGV;
  11
  12 %tokens = ();
  13 @tokendata = ();
  14
  15 #
  16 # List of condition codes
  17 #
  18 @conditions = ('a', 'ae', 'b', 'be', 'c', 'e', 'g', 'ge', 'l', 'le',
  19                'na', 'nae', 'nb', 'nbe', 'nc', 'ne', 'ng', 'nge', 'nl',
  20                'nle', 'no', 'np', 'ns', 'nz', 'o', 'p', 'pe', 'po', 's', 'z');
  21
  22 #
  23 # Read insns.dat
  24 #
  25 open(ID, "< ${insns_dat}") or die "$0: cannot open $insns_dat: $!\n";
  26 while (defined($line = <ID>)) {
  27     if ($line =~ /^([A-Z0-9_]+)(|cc)\s/) {
  28         $insn = $1.$2;
  29         ($token = $1) =~ tr/A-Z/a-z/;
  30
  31         if ($2 eq '') {
  32             # Single instruction token
  33             if (!defined($tokens{$token})) {
  34                 $tokens{$token} = scalar @tokendata;
  35                 push(@tokendata, "\"${token}\", TOKEN_INSN, C_none, I_${insn}");
  36             }
  37         } else {
  38             # Conditional instruction
  39             foreach $cc (@conditions) {
  40                 if (!defined($tokens{$token.$cc})) {
  41                     $tokens{$token.$cc} = scalar @tokendata;
  42                     push(@tokendata, "\"${token}${cc}\", TOKEN_INSN, C_\U$cc\E, I_${insn}");
  43                 }
  44             }
  45         }
  46     }
  47 }
  48 close(ID);
  49
  50 #
  51 # Read regs.dat
  52 #
  53 open(RD, "< ${regs_dat}") or die "$0: cannot open $regs_dat: $!\n";
  54 while (defined($line = <RD>)) {
  55     if ($line =~ /^([a-z0-9_-]+)\s/) {
  56         $reg = $1;
  57
  58         if ($reg =~ /^(.*[^0-9])([0-9]+)\-([0-9]+)(|[^0-9].*)$/) {
  59             $nregs = $3-$2+1;
  60             $reg = $1.$2.$4;
  61             $reg_nr = $2;
  62             $reg_prefix = $1;
  63             $reg_suffix = $4;
  64         } else {
  65             $nregs = 1;
  66             undef $reg_prefix, $reg_suffix;
  67         }
  68
  69         while ($nregs--) {
  70             if (defined($tokens{$reg})) {
  71                 die "Duplicate definition: $reg\n";
  72             }
  73             $tokens{$reg} = scalar @tokendata;
  74             push(@tokendata, "\"${reg}\", TOKEN_REG, 0, R_\U${reg}\E");
  75
  76             if (defined($reg_prefix)) {
  77                 $reg_nr++;
  78                 $reg = sprintf("%s%u%s", $reg_prefix, $reg_nr, $reg_suffix);
  79             } else {
  80                 # Not a dashed sequence
  81                 die if ($nregs);
  82             }
  83         }
  84     }
  85 }
  86 close(RD);
  87
  88 #
  89 # Read tokens.dat
  90 #
  91 open(TD, "< ${tokens_dat}") or die "$0: cannot open $tokens_dat: $!\n";
  92 while (defined($line = <TD>)) {
  93     if ($line =~ /^\%\s+(.*)$/) {
  94         $pattern = $1;
  95     } elsif ($line =~ /^([a-z0-9_-]+)/) {
  96         $token = $1;
  97
  98         if (defined($tokens{$token})) {
  99             die "Duplicate definition: $token\n";
 100         }
 101         $tokens{$token} = scalar @tokendata;
 102
 103         $data = $pattern;
 104         $data =~ s/\*/\U$token/g;
 105
 106         push(@tokendata, "\"$token\", $data");
 107     }
 108 }
 109 close(TD);
 110
 111 #
 112 # Actually generate the hash
 113 #
 114 @hashinfo = gen_perfect_hash(\%tokens);
 115 if (!defined(@hashinfo)) {
 116     die "$0: no hash found\n";
 117 }
 118
 119 # Paranoia...
 120 verify_hash_table(\%tokens, \@hashinfo);
 121
 122 ($n, $sv, $g) = @hashinfo;
 123 $sv2 = $sv+2;
 124
 125 die if ($n & ($n-1));
 126
 127 print "/*\n";
 128 print " * This file is generated from insns.dat, regs.dat and token.dat\n";
 129 print " * by tokhash.pl; do not edit.\n";
 130 print " */\n";
 131 print "\n";
 132
 133 print "#include <string.h>\n";
 134 print "#include \"nasm.h\"\n";
 135 print "#include \"insns.h\"\n";
 136 print "\n";
 137
 138 print "#define rot(x,y) (((uint32_t)(x) << (y))+((uint32_t)(x) >> (32-(y))))\n";
 139 print "\n";
 140
 141 # These somewhat odd sizes and ordering thereof are due to the
 142 # relative ranges of the types; this makes it fit in 16 bytes on
 143 # 64-bit machines and 12 bytes on 32-bit machines.
 144 print "struct tokendata {\n";
 145 print "    const char *string;\n";
 146 print "    int16_t tokentype;\n";
 147 print "    int16_t aux;\n";
 148 print "    int32_t num;\n";
 149 print "};\n";
 150 print "\n";
 151
 152 print "int nasm_token_hash(const char *token, struct tokenval *tv)\n";
 153 print "{\n";
 154
 155 # Put a large value in unused slots.  This makes it extremely unlikely
 156 # that any combination that involves unused slot will pass the range test.
 157 # This speeds up rejection of unrecognized tokens, i.e. identifiers.
 158 print "#define UNUSED 16383\n";
 159
 160 print "    static const int16_t hash1[$n] = {\n";
 161 for ($i = 0; $i < $n; $i++) {
 162     my $h = ${$g}[$i*2+0];
 163     print "        ", defined($h) ? $h : 'UNUSED', ",\n";
 164 }
 165 print "    };\n";
 166
 167 print "    static const int16_t hash2[$n] = {\n";
 168 for ($i = 0; $i < $n; $i++) {
 169     my $h = ${$g}[$i*2+1];
 170     print "        ", defined($h) ? $h : 'UNUSED', ",\n";
 171 }
 172 print "    };\n";
 173
 174 printf "    static const struct tokendata tokendata[%d] = {\n", scalar(@tokendata);
 175 foreach $d (@tokendata) {
 176     print "        { ", $d, " },\n";
 177 }
 178 print  "    };\n";
 179
 180 print  "    uint32_t k1 = 0, k2 = 0;\n";
 181 print  "    uint8_t c;\n";
 182 # For correct overflow behavior, "ix" should be unsigned of the same
 183 # width as the hash arrays.
 184 print  "    uint16_t ix;\n";
 185 print  "    const struct tokendata *data;\n";
 186 print  "    const char *p = token;\n";
 187 print  "\n";
 188
 189 print  "    while ((c = *p++) != 0) {\n";
 190 printf "        uint32_t kn1 = rot(k1,%2d)^(rot(k2,%2d) + c);\n", ${$sv}[0], ${$sv}[1];
 191 printf "        uint32_t kn2 = rot(k2,%2d)^(rot(k1,%2d) + c);\n", ${$sv}[2], ${$sv}[3];
 192 print  "        k1 = kn1; k2 = kn2;\n";
 193 print  "    }\n";
 194 print  "\n";
 195 printf "    ix = hash1[k1 & 0x%x] + hash2[k2 & 0x%x];\n", $n-1, $n-1;
 196 printf "    if (ix >= %d)\n", scalar(@tokendata);
 197 print  "        return tv->t_type = TOKEN_ID;\n";
 198 print  "\n";
 199 print  "    data = &tokendata[ix];\n";
 200
 201 # print  "    fprintf(stderr, \"Looked for: %s found: %s\\n\", token, data->string);\n\n";
 202
 203 print  "    if (strcmp(data->string, token))\n";
 204 print  "        return tv->t_type = TOKEN_ID;\n";
 205 print  "\n";
 206 print  "    tv->t_integer = data->num;\n";
 207 print  "    tv->t_inttwo  = data->aux;\n";
 208 print  "    return tv->t_type = data->tokentype;\n";
 209 print  "}\n";