1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
|
#!/usr/bin/tclsh
#
# Generate the file opcodes.h.
#
# This TCL script scans a concatenation of the parse.h output file from the
# parser and the vdbe.c source file in order to generate the opcodes numbers
# for all opcodes.
#
# The lines of the vdbe.c that we are interested in are of the form:
#
# case OP_aaaa: /* same as TK_bbbbb */
#
# The TK_ comment is optional. If it is present, then the value assigned to
# the OP_ is the same as the TK_ value. If missing, the OP_ value is assigned
# a small integer that is different from every other OP_ value.
#
# We go to the trouble of making some OP_ values the same as TK_ values
# as an optimization. During parsing, things like expression operators
# are coded with TK_ values such as TK_ADD, TK_DIVIDE, and so forth. Later
# during code generation, we need to generate corresponding opcodes like
# OP_Add and OP_Divide. By making TK_ADD==OP_Add and TK_DIVIDE==OP_Divide,
# code to translate from one to the other is avoided. This makes the
# code generator smaller and faster.
#
# This script also scans for lines of the form:
#
# case OP_aaaa: /* jump, in1, in2, in3, out2, out3 */
#
# When such comments are found on an opcode, it means that certain
# properties apply to that opcode. Set corresponding flags using the
# OPFLG_INITIALIZER macro.
#
set in stdin
set currentOp {}
set prevName {}
set nOp 0
set nGroup 0
while {![eof $in]} {
set line [gets $in]
# Remember the TK_ values from the parse.h file.
# NB: The "TK_" prefix stands for "ToKen", not the graphical Tk toolkit
# commonly associated with TCL.
#
if {[regexp {^#define TK_} $line]} {
set tk([lindex $line 1]) [lindex $line 2]
continue
}
# Find "/* Opcode: " lines in the vdbe.c file. Each one introduces
# a new opcode. Remember which parameters are used.
#
if {[regexp {^.. Opcode: } $line]} {
set currentOp OP_[lindex $line 2]
set m 0
foreach term $line {
switch $term {
P1 {incr m 1}
P2 {incr m 2}
P3 {incr m 4}
P4 {incr m 8}
P5 {incr m 16}
}
}
set paramused($currentOp) $m
}
# Find "** Synopsis: " lines that follow Opcode:
#
if {[regexp {^.. Synopsis: (.*)} $line all x] && $currentOp!=""} {
set synopsis($currentOp) [string trim $x]
}
# Scan for "case OP_aaaa:" lines in the vdbe.c file
#
if {[regexp {^case OP_} $line]} {
set line [split $line]
set name [string trim [lindex $line 1] :]
if {$name=="OP_Abortable"} continue; # put OP_Abortable last
set op($name) -1
set group($name) 0
set jump($name) 0
set in1($name) 0
set in2($name) 0
set in3($name) 0
set out2($name) 0
set out3($name) 0
set ncycle($name) 0
for {set i 3} {$i<[llength $line]-1} {incr i} {
switch [string trim [lindex $line $i] ,] {
same {
incr i
if {[lindex $line $i]=="as"} {
incr i
set sym [string trim [lindex $line $i] ,]
set val $tk($sym)
set op($name) $val
set used($val) 1
set sameas($val) $sym
set def($val) $name
}
}
group {set group($name) 1}
jump {set jump($name) 1}
in1 {set in1($name) 1}
in2 {set in2($name) 1}
in3 {set in3($name) 1}
out2 {set out2($name) 1}
out3 {set out3($name) 1}
ncycle {set ncycle($name) 1}
}
}
if {$group($name)} {
set newGroup 0
if {[info exists groups($nGroup)]} {
if {$prevName=="" || !$group($prevName)} {
set newGroup 1
}
}
lappend groups($nGroup) $name
if {$newGroup} {incr nGroup}
} else {
if {$prevName!="" && $group($prevName)} {
incr nGroup
}
}
set order($nOp) $name
set prevName $name
incr nOp
}
}
# Assign numbers to all opcodes and output the result.
#
puts "/* Automatically generated. Do not edit */"
puts "/* See the tool/mkopcodeh.tcl script for details */"
foreach name {OP_Noop OP_Explain OP_Abortable} {
set jump($name) 0
set in1($name) 0
set in2($name) 0
set in3($name) 0
set out2($name) 0
set out3($name) 0
set ncycle($name) 0
set op($name) -1
set order($nOp) $name
incr nOp
}
# The following are the opcodes that receive special processing in the
# resolveP2Values() routine. Update this list whenever new cases are
# added to the pOp->opcode switch within resolveP2Values().
#
set rp2v_ops {
OP_Transaction
OP_AutoCommit
OP_Savepoint
OP_Checkpoint
OP_Vacuum
OP_JournalMode
OP_VUpdate
OP_VFilter
OP_Init
}
# Assign the smallest values to opcodes that are processed by resolveP2Values()
# to make code generation for the switch() statement smaller and faster.
#
set cnt -1
for {set i 0} {$i<$nOp} {incr i} {
set name $order($i)
if {[lsearch $rp2v_ops $name]>=0} {
incr cnt
while {[info exists used($cnt)]} {incr cnt}
set op($name) $cnt
set used($cnt) 1
set def($cnt) $name
}
}
set mxCase1 $cnt
# Assign the next group of values to JUMP opcodes
#
for {set i 0} {$i<$nOp} {incr i} {
set name $order($i)
if {$op($name)>=0} continue
if {!$jump($name)} continue
incr cnt
while {[info exists used($cnt)]} {incr cnt}
set op($name) $cnt
set used($cnt) 1
set def($cnt) $name
}
# Find the numeric value for the largest JUMP opcode
#
set mxJump -1
for {set i 0} {$i<$nOp} {incr i} {
set name $order($i)
if {$jump($name) && $op($name)>$mxJump} {set mxJump $op($name)}
}
# Generate the numeric values for all remaining opcodes, while
# preserving any groupings of opcodes (i.e. those that must be
# together).
#
for {set g 0} {$g<$nGroup} {incr g} {
set gLen [llength $groups($g)]
set ok 0; set start -1
set seek $cnt
while {!$ok} {
incr seek
while {[info exists used($seek)]} {incr seek}
set ok 1; set start $seek
for {set j 0} {$j<$gLen} {incr j} {
incr seek
if {[info exists used($seek)]} {
set ok 0; break
}
}
}
if {$ok} {
set next $start
for {set j 0} {$j<$gLen} {incr j} {
set name [lindex $groups($g) $j]
if {$op($name)>=0} continue
set op($name) $next
set used($next) 1
set def($next) $name
incr next
}
} else {
error "cannot find opcodes for group: $groups($g)"
}
}
for {set i 0} {$i<$nOp} {incr i} {
set name $order($i)
if {$op($name)<0} {
incr cnt
while {[info exists used($cnt)]} {incr cnt}
set op($name) $cnt
set used($cnt) 1
set def($cnt) $name
}
}
set max [lindex [lsort -decr -integer [array names used]] 0]
for {set i 0} {$i<=$max} {incr i} {
if {![info exists used($i)]} {
set def($i) "OP_NotUsed_$i"
}
if {$i>$max} {set max $i}
set name $def($i)
puts -nonewline [format {#define %-16s %3d} $name $i]
set com {}
if {[info exists jump($name)] && $jump($name)} {
lappend com "jump"
}
if {[info exists sameas($i)]} {
lappend com "same as $sameas($i)"
}
if {[info exists synopsis($name)]} {
lappend com "synopsis: $synopsis($name)"
}
if {[llength $com]} {
puts -nonewline [format " /* %-42s */" [join $com {, }]]
}
puts ""
}
if {$max>255} {
error "More than 255 opcodes - VdbeOp.opcode is of type u8!"
}
# Generate the bitvectors:
#
set bv(0) 0
for {set i 0} {$i<=$max} {incr i} {
set x 0
set name $def($i)
if {[string match OP_NotUsed* $name]==0} {
if {$jump($name)} {incr x 1}
if {$in1($name)} {incr x 2}
if {$in2($name)} {incr x 4}
if {$in3($name)} {incr x 8}
if {$out2($name)} {incr x 16}
if {$out3($name)} {incr x 32}
if {$ncycle($name)} {incr x 64}
}
set bv($i) $x
}
puts ""
puts "/* Properties such as \"out2\" or \"jump\" that are specified in"
puts "** comments following the \"case\" for each opcode in the vdbe.c"
puts "** are encoded into bitvectors as follows:"
puts "*/"
puts "#define OPFLG_JUMP 0x01 /* jump: P2 holds jmp target */"
puts "#define OPFLG_IN1 0x02 /* in1: P1 is an input */"
puts "#define OPFLG_IN2 0x04 /* in2: P2 is an input */"
puts "#define OPFLG_IN3 0x08 /* in3: P3 is an input */"
puts "#define OPFLG_OUT2 0x10 /* out2: P2 is an output */"
puts "#define OPFLG_OUT3 0x20 /* out3: P3 is an output */"
puts "#define OPFLG_NCYCLE 0x40 /* ncycle:Cycles count against P1 */"
puts "#define OPFLG_INITIALIZER \173\\"
for {set i 0} {$i<=$max} {incr i} {
if {$i%8==0} {
puts -nonewline [format "/* %3d */" $i]
}
puts -nonewline [format " 0x%02x," $bv($i)]
if {$i%8==7} {
puts "\\"
}
}
puts "\175"
puts ""
puts "/* The resolve3P2Values() routine is able to run faster if it knows"
puts "** the value of the largest JUMP opcode. The smaller the maximum"
puts "** JUMP opcode the better, so the mkopcodeh.tcl script that"
puts "** generated this include file strives to group all JUMP opcodes"
puts "** together near the beginning of the list."
puts "*/"
puts "#define SQLITE_MX_JUMP_OPCODE $mxJump /* Maximum JUMP opcode */"
|