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
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
|
import sys
from rpython.jit.metainterp.history import ConstInt
from rpython.jit.metainterp.optimize import InvalidLoop
from rpython.jit.metainterp.optimizeopt.intutils import (IntBound, IntLowerBound,
IntUpperBound)
from rpython.jit.metainterp.optimizeopt.optimizer import (Optimization, CONST_1,
CONST_0, MODE_ARRAY, MODE_STR, MODE_UNICODE)
from rpython.jit.metainterp.optimizeopt.util import make_dispatcher_method
from rpython.jit.metainterp.resoperation import rop
from rpython.jit.backend.llsupport import symbolic
def get_integer_min(is_unsigned, byte_size):
if is_unsigned:
return 0
else:
return -(1 << ((byte_size << 3) - 1))
def get_integer_max(is_unsigned, byte_size):
if is_unsigned:
return (1 << (byte_size << 3)) - 1
else:
return (1 << ((byte_size << 3) - 1)) - 1
def next_pow2_m1(n):
"""Calculate next power of 2 greater than n minus one."""
n |= n >> 1
n |= n >> 2
n |= n >> 4
n |= n >> 8
n |= n >> 16
n |= n >> 32
return n
class OptIntBounds(Optimization):
"""Keeps track of the bounds placed on integers by guards and remove
redundant guards"""
def propagate_forward(self, op):
dispatch_opt(self, op)
def opt_default(self, op):
assert not op.is_ovf()
self.emit_operation(op)
def propagate_bounds_backward(self, box):
# FIXME: This takes care of the instruction where box is the reuslt
# but the bounds produced by all instructions where box is
# an argument might also be tighten
v = self.getvalue(box)
b = v.intbound
if b.has_lower and b.has_upper and b.lower == b.upper:
v.make_constant(ConstInt(b.lower))
try:
op = self.optimizer.producer[box]
except KeyError:
return
dispatch_bounds_ops(self, op)
def optimize_GUARD_TRUE(self, op):
self.emit_operation(op)
self.propagate_bounds_backward(op.getarg(0))
optimize_GUARD_FALSE = optimize_GUARD_TRUE
optimize_GUARD_VALUE = optimize_GUARD_TRUE
def optimize_INT_OR_or_XOR(self, op):
v1 = self.getvalue(op.getarg(0))
v2 = self.getvalue(op.getarg(1))
if v1 is v2:
if op.getopnum() == rop.INT_OR:
self.make_equal_to(op.result, v1)
else:
self.make_constant_int(op.result, 0)
return
self.emit_operation(op)
if v1.intbound.known_ge(IntBound(0, 0)) and \
v2.intbound.known_ge(IntBound(0, 0)):
r = self.getvalue(op.result)
mostsignificant = v1.intbound.upper | v2.intbound.upper
r.intbound.intersect(IntBound(0, next_pow2_m1(mostsignificant)))
optimize_INT_OR = optimize_INT_OR_or_XOR
optimize_INT_XOR = optimize_INT_OR_or_XOR
def optimize_INT_AND(self, op):
v1 = self.getvalue(op.getarg(0))
v2 = self.getvalue(op.getarg(1))
self.emit_operation(op)
r = self.getvalue(op.result)
if v2.is_constant():
val = v2.box.getint()
if val >= 0:
r.intbound.intersect(IntBound(0, val))
elif v1.is_constant():
val = v1.box.getint()
if val >= 0:
r.intbound.intersect(IntBound(0, val))
elif v1.intbound.known_ge(IntBound(0, 0)) and \
v2.intbound.known_ge(IntBound(0, 0)):
lesser = min(v1.intbound.upper, v2.intbound.upper)
r.intbound.intersect(IntBound(0, next_pow2_m1(lesser)))
def optimize_INT_SUB(self, op):
v1 = self.getvalue(op.getarg(0))
v2 = self.getvalue(op.getarg(1))
self.emit_operation(op)
r = self.getvalue(op.result)
b = v1.intbound.sub_bound(v2.intbound)
if b.bounded():
r.intbound.intersect(b)
def optimize_INT_ADD(self, op):
v1 = self.getvalue(op.getarg(0))
v2 = self.getvalue(op.getarg(1))
self.emit_operation(op)
r = self.getvalue(op.result)
b = v1.intbound.add_bound(v2.intbound)
if b.bounded():
r.intbound.intersect(b)
def optimize_INT_MUL(self, op):
v1 = self.getvalue(op.getarg(0))
v2 = self.getvalue(op.getarg(1))
self.emit_operation(op)
r = self.getvalue(op.result)
b = v1.intbound.mul_bound(v2.intbound)
if b.bounded():
r.intbound.intersect(b)
def optimize_INT_FLOORDIV(self, op):
v1 = self.getvalue(op.getarg(0))
v2 = self.getvalue(op.getarg(1))
self.emit_operation(op)
r = self.getvalue(op.result)
r.intbound.intersect(v1.intbound.div_bound(v2.intbound))
def optimize_INT_MOD(self, op):
v1 = self.getvalue(op.getarg(0))
v2 = self.getvalue(op.getarg(1))
known_nonneg = (v1.intbound.known_ge(IntBound(0, 0)) and
v2.intbound.known_ge(IntBound(0, 0)))
if known_nonneg and v2.is_constant():
val = v2.box.getint()
if (val & (val-1)) == 0:
# nonneg % power-of-two ==> nonneg & (power-of-two - 1)
arg1 = op.getarg(0)
arg2 = ConstInt(val-1)
op = op.copy_and_change(rop.INT_AND, args=[arg1, arg2])
self.emit_operation(op)
if v2.is_constant():
val = v2.box.getint()
r = self.getvalue(op.result)
if val < 0:
if val == -sys.maxint-1:
return # give up
val = -val
if known_nonneg:
r.intbound.make_ge(IntBound(0, 0))
else:
r.intbound.make_gt(IntBound(-val, -val))
r.intbound.make_lt(IntBound(val, val))
def optimize_INT_LSHIFT(self, op):
v1 = self.getvalue(op.getarg(0))
v2 = self.getvalue(op.getarg(1))
self.emit_operation(op)
r = self.getvalue(op.result)
b = v1.intbound.lshift_bound(v2.intbound)
r.intbound.intersect(b)
# intbound.lshift_bound checks for an overflow and if the
# lshift can be proven not to overflow sets b.has_upper and
# b.has_lower
if b.has_lower and b.has_upper:
# Synthesize the reverse op for optimize_default to reuse
self.pure(rop.INT_RSHIFT, [op.result, op.getarg(1)], op.getarg(0))
def optimize_INT_RSHIFT(self, op):
v1 = self.getvalue(op.getarg(0))
v2 = self.getvalue(op.getarg(1))
b = v1.intbound.rshift_bound(v2.intbound)
if b.has_lower and b.has_upper and b.lower == b.upper:
# constant result (likely 0, for rshifts that kill all bits)
self.make_constant_int(op.result, b.lower)
else:
self.emit_operation(op)
r = self.getvalue(op.result)
r.intbound.intersect(b)
def optimize_GUARD_NO_OVERFLOW(self, op):
lastop = self.last_emitted_operation
if lastop is not None:
opnum = lastop.getopnum()
args = lastop.getarglist()
result = lastop.result
# If the INT_xxx_OVF was replaced with INT_xxx or removed
# completely, then we can kill the GUARD_NO_OVERFLOW.
if (opnum != rop.INT_ADD_OVF and
opnum != rop.INT_SUB_OVF and
opnum != rop.INT_MUL_OVF):
return
# Else, synthesize the non overflowing op for optimize_default to
# reuse, as well as the reverse op
elif opnum == rop.INT_ADD_OVF:
self.pure(rop.INT_ADD, args[:], result)
self.pure(rop.INT_SUB, [result, args[1]], args[0])
self.pure(rop.INT_SUB, [result, args[0]], args[1])
elif opnum == rop.INT_SUB_OVF:
self.pure(rop.INT_SUB, args[:], result)
self.pure(rop.INT_ADD, [result, args[1]], args[0])
self.pure(rop.INT_SUB, [args[0], result], args[1])
elif opnum == rop.INT_MUL_OVF:
self.pure(rop.INT_MUL, args[:], result)
self.emit_operation(op)
def optimize_GUARD_OVERFLOW(self, op):
# If INT_xxx_OVF was replaced by INT_xxx, *but* we still see
# GUARD_OVERFLOW, then the loop is invalid.
lastop = self.last_emitted_operation
if lastop is None:
raise InvalidLoop('An INT_xxx_OVF was proven not to overflow but' +
'guarded with GUARD_OVERFLOW')
opnum = lastop.getopnum()
if opnum not in (rop.INT_ADD_OVF, rop.INT_SUB_OVF, rop.INT_MUL_OVF):
raise InvalidLoop('An INT_xxx_OVF was proven not to overflow but' +
'guarded with GUARD_OVERFLOW')
self.emit_operation(op)
def optimize_INT_ADD_OVF(self, op):
v1 = self.getvalue(op.getarg(0))
v2 = self.getvalue(op.getarg(1))
resbound = v1.intbound.add_bound(v2.intbound)
if resbound.bounded():
# Transform into INT_ADD. The following guard will be killed
# by optimize_GUARD_NO_OVERFLOW; if we see instead an
# optimize_GUARD_OVERFLOW, then InvalidLoop.
op = op.copy_and_change(rop.INT_ADD)
self.emit_operation(op) # emit the op
r = self.getvalue(op.result)
r.intbound.intersect(resbound)
def optimize_INT_SUB_OVF(self, op):
v1 = self.getvalue(op.getarg(0))
v2 = self.getvalue(op.getarg(1))
if v1 is v2:
self.make_constant_int(op.result, 0)
return
resbound = v1.intbound.sub_bound(v2.intbound)
if resbound.bounded():
op = op.copy_and_change(rop.INT_SUB)
self.emit_operation(op) # emit the op
r = self.getvalue(op.result)
r.intbound.intersect(resbound)
def optimize_INT_MUL_OVF(self, op):
v1 = self.getvalue(op.getarg(0))
v2 = self.getvalue(op.getarg(1))
resbound = v1.intbound.mul_bound(v2.intbound)
if resbound.bounded():
op = op.copy_and_change(rop.INT_MUL)
self.emit_operation(op)
r = self.getvalue(op.result)
r.intbound.intersect(resbound)
def optimize_INT_LT(self, op):
v1 = self.getvalue(op.getarg(0))
v2 = self.getvalue(op.getarg(1))
if v1.intbound.known_lt(v2.intbound):
self.make_constant_int(op.result, 1)
elif v1.intbound.known_ge(v2.intbound) or v1 is v2:
self.make_constant_int(op.result, 0)
else:
self.emit_operation(op)
def optimize_INT_GT(self, op):
v1 = self.getvalue(op.getarg(0))
v2 = self.getvalue(op.getarg(1))
if v1.intbound.known_gt(v2.intbound):
self.make_constant_int(op.result, 1)
elif v1.intbound.known_le(v2.intbound) or v1 is v2:
self.make_constant_int(op.result, 0)
else:
self.emit_operation(op)
def optimize_INT_LE(self, op):
v1 = self.getvalue(op.getarg(0))
v2 = self.getvalue(op.getarg(1))
if v1.intbound.known_le(v2.intbound) or v1 is v2:
self.make_constant_int(op.result, 1)
elif v1.intbound.known_gt(v2.intbound):
self.make_constant_int(op.result, 0)
else:
self.emit_operation(op)
def optimize_INT_GE(self, op):
v1 = self.getvalue(op.getarg(0))
v2 = self.getvalue(op.getarg(1))
if v1.intbound.known_ge(v2.intbound) or v1 is v2:
self.make_constant_int(op.result, 1)
elif v1.intbound.known_lt(v2.intbound):
self.make_constant_int(op.result, 0)
else:
self.emit_operation(op)
def optimize_INT_EQ(self, op):
v1 = self.getvalue(op.getarg(0))
v2 = self.getvalue(op.getarg(1))
if v1.intbound.known_gt(v2.intbound):
self.make_constant_int(op.result, 0)
elif v1.intbound.known_lt(v2.intbound):
self.make_constant_int(op.result, 0)
elif v1 is v2:
self.make_constant_int(op.result, 1)
else:
self.emit_operation(op)
def optimize_INT_NE(self, op):
v1 = self.getvalue(op.getarg(0))
v2 = self.getvalue(op.getarg(1))
if v1.intbound.known_gt(v2.intbound):
self.make_constant_int(op.result, 1)
elif v1.intbound.known_lt(v2.intbound):
self.make_constant_int(op.result, 1)
elif v1 is v2:
self.make_constant_int(op.result, 0)
else:
self.emit_operation(op)
def optimize_INT_FORCE_GE_ZERO(self, op):
value = self.getvalue(op.getarg(0))
if value.intbound.known_ge(IntBound(0, 0)):
self.make_equal_to(op.result, value)
else:
self.emit_operation(op)
def optimize_ARRAYLEN_GC(self, op):
self.emit_operation(op)
array = self.getvalue(op.getarg(0))
result = self.getvalue(op.result)
array.make_len_gt(MODE_ARRAY, op.getdescr(), -1)
array.lenbound.bound.intersect(result.intbound)
result.intbound = array.lenbound.bound
def optimize_STRLEN(self, op):
self.emit_operation(op)
array = self.getvalue(op.getarg(0))
result = self.getvalue(op.result)
array.make_len_gt(MODE_STR, op.getdescr(), -1)
array.lenbound.bound.intersect(result.intbound)
result.intbound = array.lenbound.bound
def optimize_UNICODELEN(self, op):
self.emit_operation(op)
array = self.getvalue(op.getarg(0))
result = self.getvalue(op.result)
array.make_len_gt(MODE_UNICODE, op.getdescr(), -1)
array.lenbound.bound.intersect(result.intbound)
result.intbound = array.lenbound.bound
def optimize_STRGETITEM(self, op):
self.emit_operation(op)
v1 = self.getvalue(op.result)
v1.intbound.make_ge(IntLowerBound(0))
v1.intbound.make_lt(IntUpperBound(256))
def optimize_GETFIELD_RAW(self, op):
self.emit_operation(op)
descr = op.getdescr()
if descr.is_integer_bounded():
v1 = self.getvalue(op.result)
v1.intbound.make_ge(IntLowerBound(descr.get_integer_min()))
v1.intbound.make_le(IntUpperBound(descr.get_integer_max()))
optimize_GETFIELD_GC = optimize_GETFIELD_RAW
optimize_GETINTERIORFIELD_GC = optimize_GETFIELD_RAW
def optimize_GETARRAYITEM_RAW(self, op):
self.emit_operation(op)
descr = op.getdescr()
if descr and descr.is_item_integer_bounded():
v1 = self.getvalue(op.result)
v1.intbound.make_ge(IntLowerBound(descr.get_item_integer_min()))
v1.intbound.make_le(IntUpperBound(descr.get_item_integer_max()))
optimize_GETARRAYITEM_GC = optimize_GETARRAYITEM_RAW
def optimize_UNICODEGETITEM(self, op):
self.emit_operation(op)
v1 = self.getvalue(op.result)
v1.intbound.make_ge(IntLowerBound(0))
def make_int_lt(self, box1, box2):
v1 = self.getvalue(box1)
v2 = self.getvalue(box2)
if v1.intbound.make_lt(v2.intbound):
self.propagate_bounds_backward(box1)
if v2.intbound.make_gt(v1.intbound):
self.propagate_bounds_backward(box2)
def make_int_le(self, box1, box2):
v1 = self.getvalue(box1)
v2 = self.getvalue(box2)
if v1.intbound.make_le(v2.intbound):
self.propagate_bounds_backward(box1)
if v2.intbound.make_ge(v1.intbound):
self.propagate_bounds_backward(box2)
def make_int_gt(self, box1, box2):
self.make_int_lt(box2, box1)
def make_int_ge(self, box1, box2):
self.make_int_le(box2, box1)
def propagate_bounds_INT_LT(self, op):
r = self.getvalue(op.result)
if r.is_constant():
if r.box.same_constant(CONST_1):
self.make_int_lt(op.getarg(0), op.getarg(1))
else:
self.make_int_ge(op.getarg(0), op.getarg(1))
def propagate_bounds_INT_GT(self, op):
r = self.getvalue(op.result)
if r.is_constant():
if r.box.same_constant(CONST_1):
self.make_int_gt(op.getarg(0), op.getarg(1))
else:
self.make_int_le(op.getarg(0), op.getarg(1))
def propagate_bounds_INT_LE(self, op):
r = self.getvalue(op.result)
if r.is_constant():
if r.box.same_constant(CONST_1):
self.make_int_le(op.getarg(0), op.getarg(1))
else:
self.make_int_gt(op.getarg(0), op.getarg(1))
def propagate_bounds_INT_GE(self, op):
r = self.getvalue(op.result)
if r.is_constant():
if r.box.same_constant(CONST_1):
self.make_int_ge(op.getarg(0), op.getarg(1))
else:
self.make_int_lt(op.getarg(0), op.getarg(1))
def propagate_bounds_INT_EQ(self, op):
r = self.getvalue(op.result)
if r.is_constant():
if r.box.same_constant(CONST_1):
v1 = self.getvalue(op.getarg(0))
v2 = self.getvalue(op.getarg(1))
if v1.intbound.intersect(v2.intbound):
self.propagate_bounds_backward(op.getarg(0))
if v2.intbound.intersect(v1.intbound):
self.propagate_bounds_backward(op.getarg(1))
def propagate_bounds_INT_NE(self, op):
r = self.getvalue(op.result)
if r.is_constant():
if r.box.same_constant(CONST_0):
v1 = self.getvalue(op.getarg(0))
v2 = self.getvalue(op.getarg(1))
if v1.intbound.intersect(v2.intbound):
self.propagate_bounds_backward(op.getarg(0))
if v2.intbound.intersect(v1.intbound):
self.propagate_bounds_backward(op.getarg(1))
def propagate_bounds_INT_IS_TRUE(self, op):
r = self.getvalue(op.result)
if r.is_constant():
if r.box.same_constant(CONST_1):
v1 = self.getvalue(op.getarg(0))
if v1.intbound.known_ge(IntBound(0, 0)):
v1.intbound.make_gt(IntBound(0, 0))
self.propagate_bounds_backward(op.getarg(0))
def propagate_bounds_INT_IS_ZERO(self, op):
r = self.getvalue(op.result)
if r.is_constant():
if r.box.same_constant(CONST_1):
v1 = self.getvalue(op.getarg(0))
# Clever hack, we can't use self.make_constant_int yet because
# the args aren't in the values dictionary yet so it runs into
# an assert, this is a clever way of expressing the same thing.
v1.intbound.make_ge(IntBound(0, 0))
v1.intbound.make_lt(IntBound(1, 1))
self.propagate_bounds_backward(op.getarg(0))
def propagate_bounds_INT_ADD(self, op):
v1 = self.getvalue(op.getarg(0))
v2 = self.getvalue(op.getarg(1))
r = self.getvalue(op.result)
b = r.intbound.sub_bound(v2.intbound)
if v1.intbound.intersect(b):
self.propagate_bounds_backward(op.getarg(0))
b = r.intbound.sub_bound(v1.intbound)
if v2.intbound.intersect(b):
self.propagate_bounds_backward(op.getarg(1))
def propagate_bounds_INT_SUB(self, op):
v1 = self.getvalue(op.getarg(0))
v2 = self.getvalue(op.getarg(1))
r = self.getvalue(op.result)
b = r.intbound.add_bound(v2.intbound)
if v1.intbound.intersect(b):
self.propagate_bounds_backward(op.getarg(0))
b = r.intbound.sub_bound(v1.intbound).mul(-1)
if v2.intbound.intersect(b):
self.propagate_bounds_backward(op.getarg(1))
def propagate_bounds_INT_MUL(self, op):
v1 = self.getvalue(op.getarg(0))
v2 = self.getvalue(op.getarg(1))
r = self.getvalue(op.result)
b = r.intbound.div_bound(v2.intbound)
if v1.intbound.intersect(b):
self.propagate_bounds_backward(op.getarg(0))
b = r.intbound.div_bound(v1.intbound)
if v2.intbound.intersect(b):
self.propagate_bounds_backward(op.getarg(1))
def propagate_bounds_INT_LSHIFT(self, op):
v1 = self.getvalue(op.getarg(0))
v2 = self.getvalue(op.getarg(1))
r = self.getvalue(op.result)
b = r.intbound.rshift_bound(v2.intbound)
if v1.intbound.intersect(b):
self.propagate_bounds_backward(op.getarg(0))
propagate_bounds_INT_ADD_OVF = propagate_bounds_INT_ADD
propagate_bounds_INT_SUB_OVF = propagate_bounds_INT_SUB
propagate_bounds_INT_MUL_OVF = propagate_bounds_INT_MUL
dispatch_opt = make_dispatcher_method(OptIntBounds, 'optimize_',
default=OptIntBounds.opt_default)
dispatch_bounds_ops = make_dispatcher_method(OptIntBounds, 'propagate_bounds_')
|
