mkopcodeh.awk

   1 #!/usr/bin/awk -f
   2 #
   3 # Generate the file opcodes.h.
   4 #
   5 # This AWK script scans a concatenation of the parse.h output file from the
   6 # parser and the vdbe.c source file in order to generate the opcodes numbers
   7 # for all opcodes.
   8 #
   9 # The lines of the vdbe.c that we are interested in are of the form:
  10 #
  11 #       case OP_aaaa:      /* same as TK_bbbbb */
  12 #
  13 # The TK_ comment is optional.  If it is present, then the value assigned to
  14 # the OP_ is the same as the TK_ value.  If missing, the OP_ value is assigned
  15 # a small integer that is different from every other OP_ value.
  16 #
  17 # We go to the trouble of making some OP_ values the same as TK_ values
  18 # as an optimization.  During parsing, things like expression operators
  19 # are coded with TK_ values such as TK_ADD, TK_DIVIDE, and so forth.  Later
  20 # during code generation, we need to generate corresponding opcodes like
  21 # OP_Add and OP_Divide.  By making TK_ADD==OP_Add and TK_DIVIDE==OP_Divide,
  22 # code to translate from one to the other is avoided.  This makes the
  23 # code generator run (infinitesimally) faster and more importantly it makes
  24 # the library footprint smaller.
  25 #
  26 # This script also scans for lines of the form:
  27 #
  28 #       case OP_aaaa:       /* jump, in1, in2, in3, out2-prerelease, out3 */
  29 #
  30 # When such comments are found on an opcode, it means that certain
  31 # properties apply to that opcode.  Set corresponding flags using the
  32 # OPFLG_INITIALIZER macro.
  33 #
  34
  35
  36 # Remember the TK_ values from the parse.h file
  37 /^#define TK_/ {
  38   tk[$2] = 0+$3    # tk[x] holds the numeric value for TK symbol X
  39 }
  40
  41 # Find "/* Opcode: " lines in the vdbe.c file.  Each one introduces
  42 # a new opcode.  Remember which parameters are used.
  43 /^.. Opcode: / {
  44   currentOp = "OP_" $3
  45   m = 0
  46   for(i=4; i<=NF; i++){
  47     x = $i
  48     if( x=="P1" ) m += 1
  49     if( x=="P2" ) m += 2
  50     if( x=="P3" ) m += 4
  51     if( x=="P4" ) m += 8
  52     if( x=="P5" ) m += 16
  53   }
  54   paramused[currentOp] = m
  55 }
  56
  57 # Find "** Synopsis: " lines that follow Opcode:
  58 /^.. Synopsis: / {
  59   if( currentOp ){
  60     x = $3
  61     for(i=4; i<=NF; i++){
  62       x = x " " $i
  63     }
  64     synopsis[currentOp] = x
  65   }
  66 }
  67
  68 # Scan for "case OP_aaaa:" lines in the vdbe.c file
  69 /^case OP_/ {
  70   name = $2
  71   sub(/:/,"",name)
  72   sub("\r","",name)
  73   op[name] = -1       # op[x] holds the numeric value for OP symbol x
  74   jump[name] = 0
  75   out2_prerelease[name] = 0
  76   in1[name] = 0
  77   in2[name] = 0
  78   in3[name] = 0
  79   out2[name] = 0
  80   out3[name] = 0
  81   for(i=3; i<NF; i++){
  82     if($i=="same" && $(i+1)=="as"){
  83       sym = $(i+2)
  84       sub(/,/,"",sym)
  85       val = tk[sym]
  86       op[name] = val
  87       used[val] = 1
  88       sameas[val] = sym
  89       def[val] = name
  90     }
  91     x = $i
  92     sub(",","",x)
  93     if(x=="jump"){
  94       jump[name] = 1
  95     }else if(x=="out2-prerelease"){
  96       out2_prerelease[name] = 1
  97     }else if(x=="in1"){
  98       in1[name] = 1
  99     }else if(x=="in2"){
 100       in2[name] = 1
 101     }else if(x=="in3"){
 102       in3[name] = 1
 103     }else if(x=="out2"){
 104       out2[name] = 1
 105     }else if(x=="out3"){
 106       out3[name] = 1
 107     }
 108   }
 109   order[n_op++] = name;
 110 }
 111
 112 # Assign numbers to all opcodes and output the result.
 113 END {
 114   cnt = 0
 115   max = 0
 116   print "/* Automatically generated.  Do not edit */"
 117   print "/* See the mkopcodeh.awk script for details */"
 118   op["OP_Noop"] = -1;
 119   order[n_op++] = "OP_Noop";
 120   op["OP_Explain"] = -1;
 121   order[n_op++] = "OP_Explain";
 122
 123   # Assign small values to opcodes that are processed by resolveP2Values()
 124   # to make code generation for the switch() statement smaller and faster.
 125   for(i=0; i<n_op; i++){
 126     name = order[i];
 127     if( op[name]>=0 ) continue;
 128     if( name=="OP_Function"      \
 129      || name=="OP_AggStep"       \
 130      || name=="OP_Transaction"   \
 131      || name=="OP_AutoCommit"    \
 132      || name=="OP_Savepoint"     \
 133      || name=="OP_Checkpoint"    \
 134      || name=="OP_Vacuum"        \
 135      || name=="OP_JournalMode"   \
 136      || name=="OP_VUpdate"       \
 137      || name=="OP_VFilter"       \
 138      || name=="OP_Next"          \
 139      || name=="OP_NextIfOpen"    \
 140      || name=="OP_SorterNext"    \
 141      || name=="OP_Prev"          \
 142      || name=="OP_PrevIfOpen"    \
 143     ){
 144       cnt++
 145       while( used[cnt] ) cnt++
 146       op[name] = cnt
 147       used[cnt] = 1
 148       def[cnt] = name
 149     }
 150   }
 151
 152   # Generate the numeric values for opcodes
 153   for(i=0; i<n_op; i++){
 154     name = order[i];
 155     if( op[name]<0 ){
 156       cnt++
 157       while( used[cnt] ) cnt++
 158       op[name] = cnt
 159       used[cnt] = 1
 160       def[cnt] = name
 161     }
 162   }
 163   max = cnt
 164   for(i=1; i<=max; i++){
 165     if( !used[i] ){
 166       def[i] = "OP_NotUsed_" i
 167     }
 168     printf "#define %-16s %3d", def[i], i
 169     com = ""
 170     if( sameas[i] ){
 171       com = "same as " sameas[i]
 172     }
 173     x = synopsis[def[i]]
 174     if( x ){
 175       if( com=="" ){
 176         com = "synopsis: " x
 177       } else {
 178         com = com ", synopsis: " x
 179       }
 180     }
 181     if( com!="" ){
 182       printf " /* %-42s */", com
 183     }
 184     printf "\n"
 185   }
 186
 187   # Generate the bitvectors:
 188   #
 189   #  bit 0:     jump
 190   #  bit 1:     pushes a result onto stack
 191   #  bit 2:     output to p1.  release p1 before opcode runs
 192   #
 193   for(i=0; i<=max; i++){
 194     name = def[i]
 195     a0 = a1 = a2 = a3 = a4 = a5 = a6 = a7 = 0
 196     if( jump[name] ) a0 = 1;
 197     if( out2_prerelease[name] ) a1 = 2;
 198     if( in1[name] ) a2 = 4;
 199     if( in2[name] ) a3 = 8;
 200     if( in3[name] ) a4 = 16;
 201     if( out2[name] ) a5 = 32;
 202     if( out3[name] ) a6 = 64;
 203     bv[i] = a0+a1+a2+a3+a4+a5+a6+a7;
 204   }
 205   print "\n"
 206   print "/* Properties such as \"out2\" or \"jump\" that are specified in"
 207   print "** comments following the \"case\" for each opcode in the vdbe.c"
 208   print "** are encoded into bitvectors as follows:"
 209   print "*/"
 210   print "#define OPFLG_JUMP            0x0001  /* jump:  P2 holds jmp target */"
 211   print "#define OPFLG_OUT2_PRERELEASE 0x0002  /* out2-prerelease: */"
 212   print "#define OPFLG_IN1             0x0004  /* in1:   P1 is an input */"
 213   print "#define OPFLG_IN2             0x0008  /* in2:   P2 is an input */"
 214   print "#define OPFLG_IN3             0x0010  /* in3:   P3 is an input */"
 215   print "#define OPFLG_OUT2            0x0020  /* out2:  P2 is an output */"
 216   print "#define OPFLG_OUT3            0x0040  /* out3:  P3 is an output */"
 217   print "#define OPFLG_INITIALIZER {\\"
 218   for(i=0; i<=max; i++){
 219     if( i%8==0 ) printf("/* %3d */",i)
 220     printf " 0x%02x,", bv[i]
 221     if( i%8==7 ) printf("\\\n");
 222   }
 223   print "}"
 224   if( 0 ){
 225     print "\n/* Bitmask to indicate which fields (P1..P5) of each opcode are"
 226     print "** actually used.\n*/"
 227     print "#define OP_PARAM_USED_INITIALIZER {\\"
 228     for(i=0; i<=max; i++){
 229       if( i%8==0 ) printf("/* %3d */",i)
 230       printf " 0x%02x,", paramused[def[i]]
 231       if( i%8==7 ) printf("\\\n");
 232     }
 233     print "}"
 234   }
 235 }