-
Notifications
You must be signed in to change notification settings - Fork 22
/
optimizer.py
479 lines (403 loc) · 13.7 KB
/
optimizer.py
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
from structures import InternalFunction
from lifter import Lifter
from blockstate import ConstantState
from blockstate import CopyState
from blockstate import MemState
from opcodes import *
from baseexecutor import execute_binop, execute_monop
from instructions import MoveInstruction, Instruction
import sys, math
def apply_peephole_optimizations(func):
for block in func.graph:
__order_operands(block)
__size_one_rewrites(block)
__size_two_rewrites(block)
rewrite_free_ptr(block)
# __sha3_rewrites(block)
def __order_operands(block):
for instruction in block:
if instruction.opcode not in {"AND", "ADD", "MUL", "OR", "EQ", "XOR"}:
continue
reads = instruction.reads
if isinstance(reads[0], str) != isinstance(reads[1], str):
if isinstance(reads[0], str):
instruction.reads.reverse()
def __size_one_rewrites(block):
instructions = block.get_instructions()
for i, instruction in enumerate(instructions):
__remove_self_assign(i, instruction, block)
__fold_constant(i, instruction, block)
__rewrite_word_add(instruction)
__rewrite_shift(i, instruction, block)
__rewrite_move(i, instruction, block)
# __rewrite_assert(i, instruction, block)
def __remove_self_assign(i, instruction, block):
if instruction.opcode == "MOVE" and \
(instruction.reads[0] == instruction.writes[0]):
block.set_nop_instruction(i)
if instruction.opcode == "AND" and instruction.reads[1] == WORD_MASK and \
(instruction.reads[0] == instruction.writes[1]):
block.set_nop_instruction(i)
def __fold_constant(i, instruction, block):
opcode = instruction.opcode
constants = instruction.get_constants()
if not constants:
return
inputs = [opcode] + constants
writes = instruction.writes
address = instruction.address
if opcode in bin_ops:
try:
value = execute_binop(inputs)
except ZeroDivisionError:
return
new_instruction = MoveInstruction("MOVE", [value], writes, address)
block.set_instruction(i, new_instruction)
elif opcode in mono_ops:
value = execute_monop(inputs)
new_instruction = MoveInstruction("MOVE", [value], writes, address)
block.set_instruction(i, new_instruction)
def __rewrite_word_add(instruction):
if instruction.opcode != "ADD":
return
if instruction.reads[0] != WORD_MASK:
return
instruction.opcode = "SUB"
instruction.reads[0] = instruction.reads[1]
instruction.reads[1] = 1
def __rewrite_shift(i, instruction, block):
opcode = instruction.opcode
address = instruction.address
if opcode not in {"DIV", "MUL"}:
return
if opcode == "DIV":
num = instruction.reads[1]
if isinstance(num, str):
return
if not (((num & (num - 1)) == 0) and num > 256):
return
exp = int(math.log(num, 2))
new_instruction = \
Instruction("SR", [instruction.reads[0], exp], instruction.writes, address)
block.set_instruction(i, new_instruction)
elif opcode == "MUL":
num = instruction.reads[0]
if isinstance(num, str):
return
if not (((num & (num - 1)) == 0) and num > 256):
return
exp = int(math.log(num, 2))
new_instruction = \
Instruction("SL", [instruction.reads[1], exp], instruction.writes, address)
block.set_instruction(i, new_instruction)
def __rewrite_move(i, instruction, block):
opcode = instruction.opcode
reads = instruction.reads
writes = instruction.writes
address = instruction.address
if opcode == "DIV" and reads[1] == 1:
new_instruction = MoveInstruction("MOVE", [reads[0]], writes, address)
block.set_instruction(i, new_instruction)
elif opcode == "ADD" and reads[0] == 0:
new_instruction = MoveInstruction("MOVE", [reads[1]], writes, address)
block.set_instruction(i, new_instruction)
elif opcode == "MUL" and reads[0] == 1:
new_instruction = MoveInstruction("MOVE", [reads[1]], writes, address)
block.set_instruction(i, new_instruction)
elif opcode == "AND" and reads[0] == WORD_MASK:
new_instruction = MoveInstruction("MOVE", [reads[1]], writes, address)
block.set_instruction(i, new_instruction)
elif opcode == "SUB" and reads[0] == reads[1]:
new_instruction = MoveInstruction("MOVE", [0], writes, address)
block.set_instruction(i, new_instruction)
# def __rewrite_assert(i, instruction, block):
# reads = instruction.reads
# if instruction.opcode == "JUMPI" and reads[0] == 0:
# new_instruction = Instruction("ASSERT", [reads[1]], [], instruction.address)
# block.set_instruction(i, new_instruction)
def __size_two_rewrites(block):
instructions = block.get_instructions()
for i in range(1, len(instructions)):
ins_0, ins_1 = instructions[i - 1:i + 1]
__remove_double_mask(i, ins_0, ins_1, block)
__remove_address_mask(i, ins_0, ins_1, block)
__remove_doube_iszero(i, ins_0, ins_1, block)
# want to remove double iszero first
for i in range(1, len(instructions)):
ins_0, ins_1 = instructions[i - 1:i + 1]
__rewrite_negate_ops(i, ins_0, ins_1, block)
def __remove_double_mask(i, instruction_0, instruction_1, block):
if instruction_0.opcode != "AND" or \
instruction_1.opcode != "AND":
return
w_0 = instruction_0.writes[0]
w_1 = instruction_1.writes[0]
reads_0 = instruction_0.reads
reads_1 = instruction_1.reads
if reads_0[0] != reads_1[0] or \
w_0 != reads_1[1] or \
w_0 != w_1:
return
block.set_nop_instruction(i - 1)
instruction_1.reads[1] = reads_0[1]
def __remove_address_mask(i, instruction_0, instruction_1, block):
if instruction_0.opcode not in {"CALLER", "ADDRESS"} or \
instruction_1.opcode != "AND":
return
w_0 = instruction_0.writes[0]
w_1 = instruction_1.writes[0]
reads_1 = instruction_1.reads
if reads_1[0] != ADDRESS_MASK or reads_1[1] != w_0:
return
address = instruction_1.address
instruction_1 = MoveInstruction("MOVE", [w_0], [w_1], address)
block.set_instruction(i, instruction_1)
def __remove_doube_iszero(i, instruction_0, instruction_1, block):
if instruction_0.opcode != "ISZERO" or \
instruction_1.opcode != "ISZERO" or \
instruction_1.address != instruction_0.address + 1:
return
block.set_nop_instruction(i - 1)
block.set_nop_instruction(i)
def __rewrite_negate_ops(i, instruction_0, instruction_1, block):
opcode = instruction_0.opcode
if instruction_0.opcode not in negate_ops or \
instruction_1.opcode != "ISZERO" or \
instruction_1.address != instruction_0.address + 1:
return
block.set_nop_instruction(i - 1)
opcode = negate_ops[opcode]
new_instruction = \
Instruction(opcode, instruction_0.reads, instruction_0.writes, instruction_0.address)
block.set_instruction(i, new_instruction)
def __sha3_rewrites(block):
local_memory = MemState()
instructions = block.get_instructions()
for index, instruction in enumerate(instructions):
if instruction.opcode == "SHA3":
begin, end = instruction.reads
if begin == 0 and not isinstance(end, str):
addresses = range(begin, end, 32)
items = local_memory.lookup_mapping(addresses)
if len(items) != 0:
values, indices = zip(*items)
for i in indices:
block.set_nop_instruction(i)
operation = Instruction("SHA3R", list(values), instruction.writes, instruction.address)
block.set_instruction(index, operation)
local_memory.add_mapping(index, instruction)
def rewrite_free_ptr(block):
for i, instruction in enumerate(block.get_instructions()):
opcode = instruction.opcode
if opcode == "MLOAD" and \
instruction.reads[0] == 64:
new_instruction = MoveInstruction("MOVE", ["$m"], instruction.writes, instruction.address)
block.set_instruction(i, new_instruction)
elif opcode == "MSTORE" and \
instruction.reads[0] == 64:
new_instruction = MoveInstruction("MOVE", [instruction.reads[1]], ["$m"], instruction.address)
block.set_instruction(i, new_instruction)
def push_down_definition(block):
instructions = block.get_instructions()
for i, instruction in enumerate(instructions):
writes = instruction.writes
if len(writes) != 1:
continue
ui = get_single_use(i, instruction, instructions)
if ui == -1:
continue
if can_reach(instruction, i, ui, instructions):
block.set_nop_instruction(i)
instruction.writes[0] = instructions[ui].writes[0]
block.set_instruction(ui, instruction)
def get_single_use(begin, d, instructions):
wr = d.writes[0]
ui, count = -1, 0
killed = False
for i, u in enumerate(instructions):
if i < begin + 1:
continue
if u.reads_from(wr):
ui = i
count += 1
if u.writes_to(wr):
killed = True
break
u = instructions[ui]
if count == 1 and killed and u.opcode == "MOVE":
return ui
return -1
def can_reach(d, begin, end, instructions):
mload = d.opcode == "MLOAD"
if d.opcode in mem_read_ops | {"SLOAD"} and not mload:
return False
targets = set(d.writes + d.reads)
# print(instructions[begin])
for u in instructions[begin + 1:end]:
if mload and u.opcode in mem_write_ops:
return False
for r in u.writes:
if r in targets:
return False
return True
class Optimizer(Lifter):
def __init__(self, binary):
Lifter.__init__(self, binary)
# return
self.__debug = False
for func in self.get_all_functions():
self.__optimize_function(func)
def __optimize_function(self, func):
self.change = True
while self.change:
self.change = False
self.__propagate_constant_values(func)
self.__propagate_copy_instructions(func)
apply_peephole_optimizations(func)
for block in func.graph:
push_down_definition(block)
self.__eliminate_dead_instructions(func)
def __propagate_constant_values(self, func):
self.__outs = dict()
self.__compute_constant_states(func)
self.__apply_constant_propagation(func)
def __compute_constant_states(self, func):
graph = func.graph
for block in graph:
block_id = block.get_id()
self.__outs[block_id] = ConstantState()
change = True
while change:
change = False
for block in graph:
block_id = block.get_id()
new_out = ConstantState()
for pre in graph.get_predecessor_ids(block_id):
new_out.join(self.__outs[pre])
for inst in block:
new_out.add_mapping(inst)
if not (self.__outs[block_id] == new_out):
self.__outs[block_id] = new_out
change = True
def __apply_constant_propagation(self, func):
graph = func.graph
for block in graph:
block_id = block.get_id()
before = ConstantState()
for pre in graph.get_predecessor_ids(block_id):
before.join(self.__outs[pre])
for instruction in block:
before.apply_mapping(instruction)
before.add_mapping(instruction)
def get_liveness_states(self, func):
self.__outs = dict()
self._uses, self._defs = dict(), dict()
self.__compute_liveness_states(func)
if isinstance(func, InternalFunction): # must keep all return registers alive
assert func.exit_id != -1
self.__outs[func.exit_id] |= set(func.writes)
return self.__outs
def __eliminate_dead_instructions(self, func):
self.__outs = dict()
self._uses, self._defs = dict(), dict()
self.__compute_liveness_states(func)
if isinstance(func, InternalFunction): # must keep all return registers alive
assert func.exit_id != -1
self.__outs[func.exit_id] |= set(func.writes)
self.__apply_instruction_elimination(func)
def __compute_liveness_states(self, func):
graph = func.graph
for block in graph:
block_id = block.get_id()
self.__compute_use_def(block)
self.__outs[block_id] = set()
change = True
while change:
change = False
for block in graph:
block_id = block.get_id()
# anything
new_out = set()
for suc in graph.get_successor_ids(block_id):
u, d = self._uses[suc], self._defs[suc]
# u is used in suc without definition
# d is defined in suc, kills out[suc]
new_out |= (u | (self.__outs[suc] - d))
if self.__outs[block_id] != new_out:
self.__outs[block_id] = new_out
change = True
for out in self.__outs.values():
out.add("$m")
def __compute_use_def(self, block):
block_id = block.get_id()
d, u = set(), set()
for instruction in block:
for read in instruction.get_read_registers():
if read not in d:
u.add(read)
for write in instruction.writes:
d.add(write)
self._defs[block_id] = d
self._uses[block_id] = u
def __apply_instruction_elimination(self, func):
# change = False
graph = func.graph
for block in graph:
block_id = block.get_id()
out = self.__outs[block_id]
new_instructions = list()
for instruction in reversed(block.get_instructions()):
reads = instruction.get_read_registers()
writes = instruction.get_write_registers()
opcode = instruction.opcode
if opcode in throw_away_ops and len(writes & out) == 0:
self.change = True
continue # do not include
else:
out = (out - writes) | reads
new_instructions.append(instruction)
new_instructions.reverse()
block.set_instructions(new_instructions)
def __propagate_copy_instructions(self, func):
self.__outs = dict()
self.__compute_copy_states(func)
self.__apply_copy_propagation(func)
def __compute_copy_states(self, func):
graph = func.graph
for block in graph:
block_id = block.get_id()
self.__outs[block_id] = CopyState()
change = True
while change:
change = False
for block in graph:
block_id = block.get_id()
new_out = CopyState()
for pre in graph.get_predecessor_ids(block_id):
new_out.join(self.__outs[pre])
for instruction in block:
new_out.add_mapping(instruction)
if not (self.__outs[block_id] == new_out):
self.__outs[block_id] = new_out
change = True
def __apply_copy_propagation(self, func):
graph = func.graph
for block in graph:
block_id = block.get_id()
before = CopyState()
for pre in graph.get_predecessor_ids(block_id):
before.join(self.__outs[pre])
# print(before.mapping)
for instruction in block:
before.apply_mapping(instruction)
before.add_mapping(instruction)
# self.__outs[block_id] = CopyState()
if __name__ == "__main__":
input_file = open(sys.argv[1])
line = input_file.readline().strip()
if " " in line:
line = line.split(" ")[1]
input_file.close()
a = Optimizer(line)
if "-d" in sys.argv:
a.debug_functions()