forked from objcode/v8
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathcodegen-mips.cc
1205 lines (1051 loc) · 44.2 KB
/
codegen-mips.cc
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
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
// Copyright 2012 the V8 project authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
#include "src/v8.h"
#if V8_TARGET_ARCH_MIPS
#include "src/codegen.h"
#include "src/macro-assembler.h"
#include "src/mips/simulator-mips.h"
namespace v8 {
namespace internal {
#define __ masm.
#if defined(USE_SIMULATOR)
byte* fast_exp_mips_machine_code = NULL;
double fast_exp_simulator(double x) {
return Simulator::current(Isolate::Current())->CallFP(
fast_exp_mips_machine_code, x, 0);
}
#endif
UnaryMathFunction CreateExpFunction() {
if (!FLAG_fast_math) return &std::exp;
size_t actual_size;
byte* buffer =
static_cast<byte*>(base::OS::Allocate(1 * KB, &actual_size, true));
if (buffer == NULL) return &std::exp;
ExternalReference::InitializeMathExpData();
MacroAssembler masm(NULL, buffer, static_cast<int>(actual_size));
{
DoubleRegister input = f12;
DoubleRegister result = f0;
DoubleRegister double_scratch1 = f4;
DoubleRegister double_scratch2 = f6;
Register temp1 = t0;
Register temp2 = t1;
Register temp3 = t2;
__ MovFromFloatParameter(input);
__ Push(temp3, temp2, temp1);
MathExpGenerator::EmitMathExp(
&masm, input, result, double_scratch1, double_scratch2,
temp1, temp2, temp3);
__ Pop(temp3, temp2, temp1);
__ MovToFloatResult(result);
__ Ret();
}
CodeDesc desc;
masm.GetCode(&desc);
ASSERT(!RelocInfo::RequiresRelocation(desc));
CpuFeatures::FlushICache(buffer, actual_size);
base::OS::ProtectCode(buffer, actual_size);
#if !defined(USE_SIMULATOR)
return FUNCTION_CAST<UnaryMathFunction>(buffer);
#else
fast_exp_mips_machine_code = buffer;
return &fast_exp_simulator;
#endif
}
#if defined(V8_HOST_ARCH_MIPS)
MemCopyUint8Function CreateMemCopyUint8Function(MemCopyUint8Function stub) {
#if defined(USE_SIMULATOR)
return stub;
#else
size_t actual_size;
byte* buffer =
static_cast<byte*>(base::OS::Allocate(3 * KB, &actual_size, true));
if (buffer == NULL) return stub;
// This code assumes that cache lines are 32 bytes and if the cache line is
// larger it will not work correctly.
MacroAssembler masm(NULL, buffer, static_cast<int>(actual_size));
{
Label lastb, unaligned, aligned, chkw,
loop16w, chk1w, wordCopy_loop, skip_pref, lastbloop,
leave, ua_chk16w, ua_loop16w, ua_skip_pref, ua_chkw,
ua_chk1w, ua_wordCopy_loop, ua_smallCopy, ua_smallCopy_loop;
// The size of each prefetch.
uint32_t pref_chunk = 32;
// The maximum size of a prefetch, it must not be less then pref_chunk.
// If the real size of a prefetch is greater then max_pref_size and
// the kPrefHintPrepareForStore hint is used, the code will not work
// correctly.
uint32_t max_pref_size = 128;
ASSERT(pref_chunk < max_pref_size);
// pref_limit is set based on the fact that we never use an offset
// greater then 5 on a store pref and that a single pref can
// never be larger then max_pref_size.
uint32_t pref_limit = (5 * pref_chunk) + max_pref_size;
int32_t pref_hint_load = kPrefHintLoadStreamed;
int32_t pref_hint_store = kPrefHintPrepareForStore;
uint32_t loadstore_chunk = 4;
// The initial prefetches may fetch bytes that are before the buffer being
// copied. Start copies with an offset of 4 so avoid this situation when
// using kPrefHintPrepareForStore.
ASSERT(pref_hint_store != kPrefHintPrepareForStore ||
pref_chunk * 4 >= max_pref_size);
// If the size is less than 8, go to lastb. Regardless of size,
// copy dst pointer to v0 for the retuen value.
__ slti(t2, a2, 2 * loadstore_chunk);
__ bne(t2, zero_reg, &lastb);
__ mov(v0, a0); // In delay slot.
// If src and dst have different alignments, go to unaligned, if they
// have the same alignment (but are not actually aligned) do a partial
// load/store to make them aligned. If they are both already aligned
// we can start copying at aligned.
__ xor_(t8, a1, a0);
__ andi(t8, t8, loadstore_chunk - 1); // t8 is a0/a1 word-displacement.
__ bne(t8, zero_reg, &unaligned);
__ subu(a3, zero_reg, a0); // In delay slot.
__ andi(a3, a3, loadstore_chunk - 1); // Copy a3 bytes to align a0/a1.
__ beq(a3, zero_reg, &aligned); // Already aligned.
__ subu(a2, a2, a3); // In delay slot. a2 is the remining bytes count.
if (kArchEndian == kLittle) {
__ lwr(t8, MemOperand(a1));
__ addu(a1, a1, a3);
__ swr(t8, MemOperand(a0));
__ addu(a0, a0, a3);
} else {
__ lwl(t8, MemOperand(a1));
__ addu(a1, a1, a3);
__ swl(t8, MemOperand(a0));
__ addu(a0, a0, a3);
}
// Now dst/src are both aligned to (word) aligned addresses. Set a2 to
// count how many bytes we have to copy after all the 64 byte chunks are
// copied and a3 to the dst pointer after all the 64 byte chunks have been
// copied. We will loop, incrementing a0 and a1 until a0 equals a3.
__ bind(&aligned);
__ andi(t8, a2, 0x3f);
__ beq(a2, t8, &chkw); // Less than 64?
__ subu(a3, a2, t8); // In delay slot.
__ addu(a3, a0, a3); // Now a3 is the final dst after loop.
// When in the loop we prefetch with kPrefHintPrepareForStore hint,
// in this case the a0+x should be past the "t0-32" address. This means:
// for x=128 the last "safe" a0 address is "t0-160". Alternatively, for
// x=64 the last "safe" a0 address is "t0-96". In the current version we
// will use "pref hint, 128(a0)", so "t0-160" is the limit.
if (pref_hint_store == kPrefHintPrepareForStore) {
__ addu(t0, a0, a2); // t0 is the "past the end" address.
__ Subu(t9, t0, pref_limit); // t9 is the "last safe pref" address.
}
__ Pref(pref_hint_load, MemOperand(a1, 0 * pref_chunk));
__ Pref(pref_hint_load, MemOperand(a1, 1 * pref_chunk));
__ Pref(pref_hint_load, MemOperand(a1, 2 * pref_chunk));
__ Pref(pref_hint_load, MemOperand(a1, 3 * pref_chunk));
if (pref_hint_store != kPrefHintPrepareForStore) {
__ Pref(pref_hint_store, MemOperand(a0, 1 * pref_chunk));
__ Pref(pref_hint_store, MemOperand(a0, 2 * pref_chunk));
__ Pref(pref_hint_store, MemOperand(a0, 3 * pref_chunk));
}
__ bind(&loop16w);
__ lw(t0, MemOperand(a1));
if (pref_hint_store == kPrefHintPrepareForStore) {
__ sltu(v1, t9, a0); // If a0 > t9, don't use next prefetch.
__ Branch(USE_DELAY_SLOT, &skip_pref, gt, v1, Operand(zero_reg));
}
__ lw(t1, MemOperand(a1, 1, loadstore_chunk)); // Maybe in delay slot.
__ Pref(pref_hint_store, MemOperand(a0, 4 * pref_chunk));
__ Pref(pref_hint_store, MemOperand(a0, 5 * pref_chunk));
__ bind(&skip_pref);
__ lw(t2, MemOperand(a1, 2, loadstore_chunk));
__ lw(t3, MemOperand(a1, 3, loadstore_chunk));
__ lw(t4, MemOperand(a1, 4, loadstore_chunk));
__ lw(t5, MemOperand(a1, 5, loadstore_chunk));
__ lw(t6, MemOperand(a1, 6, loadstore_chunk));
__ lw(t7, MemOperand(a1, 7, loadstore_chunk));
__ Pref(pref_hint_load, MemOperand(a1, 4 * pref_chunk));
__ sw(t0, MemOperand(a0));
__ sw(t1, MemOperand(a0, 1, loadstore_chunk));
__ sw(t2, MemOperand(a0, 2, loadstore_chunk));
__ sw(t3, MemOperand(a0, 3, loadstore_chunk));
__ sw(t4, MemOperand(a0, 4, loadstore_chunk));
__ sw(t5, MemOperand(a0, 5, loadstore_chunk));
__ sw(t6, MemOperand(a0, 6, loadstore_chunk));
__ sw(t7, MemOperand(a0, 7, loadstore_chunk));
__ lw(t0, MemOperand(a1, 8, loadstore_chunk));
__ lw(t1, MemOperand(a1, 9, loadstore_chunk));
__ lw(t2, MemOperand(a1, 10, loadstore_chunk));
__ lw(t3, MemOperand(a1, 11, loadstore_chunk));
__ lw(t4, MemOperand(a1, 12, loadstore_chunk));
__ lw(t5, MemOperand(a1, 13, loadstore_chunk));
__ lw(t6, MemOperand(a1, 14, loadstore_chunk));
__ lw(t7, MemOperand(a1, 15, loadstore_chunk));
__ Pref(pref_hint_load, MemOperand(a1, 5 * pref_chunk));
__ sw(t0, MemOperand(a0, 8, loadstore_chunk));
__ sw(t1, MemOperand(a0, 9, loadstore_chunk));
__ sw(t2, MemOperand(a0, 10, loadstore_chunk));
__ sw(t3, MemOperand(a0, 11, loadstore_chunk));
__ sw(t4, MemOperand(a0, 12, loadstore_chunk));
__ sw(t5, MemOperand(a0, 13, loadstore_chunk));
__ sw(t6, MemOperand(a0, 14, loadstore_chunk));
__ sw(t7, MemOperand(a0, 15, loadstore_chunk));
__ addiu(a0, a0, 16 * loadstore_chunk);
__ bne(a0, a3, &loop16w);
__ addiu(a1, a1, 16 * loadstore_chunk); // In delay slot.
__ mov(a2, t8);
// Here we have src and dest word-aligned but less than 64-bytes to go.
// Check for a 32 bytes chunk and copy if there is one. Otherwise jump
// down to chk1w to handle the tail end of the copy.
__ bind(&chkw);
__ Pref(pref_hint_load, MemOperand(a1, 0 * pref_chunk));
__ andi(t8, a2, 0x1f);
__ beq(a2, t8, &chk1w); // Less than 32?
__ nop(); // In delay slot.
__ lw(t0, MemOperand(a1));
__ lw(t1, MemOperand(a1, 1, loadstore_chunk));
__ lw(t2, MemOperand(a1, 2, loadstore_chunk));
__ lw(t3, MemOperand(a1, 3, loadstore_chunk));
__ lw(t4, MemOperand(a1, 4, loadstore_chunk));
__ lw(t5, MemOperand(a1, 5, loadstore_chunk));
__ lw(t6, MemOperand(a1, 6, loadstore_chunk));
__ lw(t7, MemOperand(a1, 7, loadstore_chunk));
__ addiu(a1, a1, 8 * loadstore_chunk);
__ sw(t0, MemOperand(a0));
__ sw(t1, MemOperand(a0, 1, loadstore_chunk));
__ sw(t2, MemOperand(a0, 2, loadstore_chunk));
__ sw(t3, MemOperand(a0, 3, loadstore_chunk));
__ sw(t4, MemOperand(a0, 4, loadstore_chunk));
__ sw(t5, MemOperand(a0, 5, loadstore_chunk));
__ sw(t6, MemOperand(a0, 6, loadstore_chunk));
__ sw(t7, MemOperand(a0, 7, loadstore_chunk));
__ addiu(a0, a0, 8 * loadstore_chunk);
// Here we have less than 32 bytes to copy. Set up for a loop to copy
// one word at a time. Set a2 to count how many bytes we have to copy
// after all the word chunks are copied and a3 to the dst pointer after
// all the word chunks have been copied. We will loop, incrementing a0
// and a1 untill a0 equals a3.
__ bind(&chk1w);
__ andi(a2, t8, loadstore_chunk - 1);
__ beq(a2, t8, &lastb);
__ subu(a3, t8, a2); // In delay slot.
__ addu(a3, a0, a3);
__ bind(&wordCopy_loop);
__ lw(t3, MemOperand(a1));
__ addiu(a0, a0, loadstore_chunk);
__ addiu(a1, a1, loadstore_chunk);
__ bne(a0, a3, &wordCopy_loop);
__ sw(t3, MemOperand(a0, -1, loadstore_chunk)); // In delay slot.
__ bind(&lastb);
__ Branch(&leave, le, a2, Operand(zero_reg));
__ addu(a3, a0, a2);
__ bind(&lastbloop);
__ lb(v1, MemOperand(a1));
__ addiu(a0, a0, 1);
__ addiu(a1, a1, 1);
__ bne(a0, a3, &lastbloop);
__ sb(v1, MemOperand(a0, -1)); // In delay slot.
__ bind(&leave);
__ jr(ra);
__ nop();
// Unaligned case. Only the dst gets aligned so we need to do partial
// loads of the source followed by normal stores to the dst (once we
// have aligned the destination).
__ bind(&unaligned);
__ andi(a3, a3, loadstore_chunk - 1); // Copy a3 bytes to align a0/a1.
__ beq(a3, zero_reg, &ua_chk16w);
__ subu(a2, a2, a3); // In delay slot.
if (kArchEndian == kLittle) {
__ lwr(v1, MemOperand(a1));
__ lwl(v1,
MemOperand(a1, 1, loadstore_chunk, MemOperand::offset_minus_one));
__ addu(a1, a1, a3);
__ swr(v1, MemOperand(a0));
__ addu(a0, a0, a3);
} else {
__ lwl(v1, MemOperand(a1));
__ lwr(v1,
MemOperand(a1, 1, loadstore_chunk, MemOperand::offset_minus_one));
__ addu(a1, a1, a3);
__ swl(v1, MemOperand(a0));
__ addu(a0, a0, a3);
}
// Now the dst (but not the source) is aligned. Set a2 to count how many
// bytes we have to copy after all the 64 byte chunks are copied and a3 to
// the dst pointer after all the 64 byte chunks have been copied. We will
// loop, incrementing a0 and a1 until a0 equals a3.
__ bind(&ua_chk16w);
__ andi(t8, a2, 0x3f);
__ beq(a2, t8, &ua_chkw);
__ subu(a3, a2, t8); // In delay slot.
__ addu(a3, a0, a3);
if (pref_hint_store == kPrefHintPrepareForStore) {
__ addu(t0, a0, a2);
__ Subu(t9, t0, pref_limit);
}
__ Pref(pref_hint_load, MemOperand(a1, 0 * pref_chunk));
__ Pref(pref_hint_load, MemOperand(a1, 1 * pref_chunk));
__ Pref(pref_hint_load, MemOperand(a1, 2 * pref_chunk));
if (pref_hint_store != kPrefHintPrepareForStore) {
__ Pref(pref_hint_store, MemOperand(a0, 1 * pref_chunk));
__ Pref(pref_hint_store, MemOperand(a0, 2 * pref_chunk));
__ Pref(pref_hint_store, MemOperand(a0, 3 * pref_chunk));
}
__ bind(&ua_loop16w);
__ Pref(pref_hint_load, MemOperand(a1, 3 * pref_chunk));
if (kArchEndian == kLittle) {
__ lwr(t0, MemOperand(a1));
__ lwr(t1, MemOperand(a1, 1, loadstore_chunk));
__ lwr(t2, MemOperand(a1, 2, loadstore_chunk));
if (pref_hint_store == kPrefHintPrepareForStore) {
__ sltu(v1, t9, a0);
__ Branch(USE_DELAY_SLOT, &ua_skip_pref, gt, v1, Operand(zero_reg));
}
__ lwr(t3, MemOperand(a1, 3, loadstore_chunk)); // Maybe in delay slot.
__ Pref(pref_hint_store, MemOperand(a0, 4 * pref_chunk));
__ Pref(pref_hint_store, MemOperand(a0, 5 * pref_chunk));
__ bind(&ua_skip_pref);
__ lwr(t4, MemOperand(a1, 4, loadstore_chunk));
__ lwr(t5, MemOperand(a1, 5, loadstore_chunk));
__ lwr(t6, MemOperand(a1, 6, loadstore_chunk));
__ lwr(t7, MemOperand(a1, 7, loadstore_chunk));
__ lwl(t0,
MemOperand(a1, 1, loadstore_chunk, MemOperand::offset_minus_one));
__ lwl(t1,
MemOperand(a1, 2, loadstore_chunk, MemOperand::offset_minus_one));
__ lwl(t2,
MemOperand(a1, 3, loadstore_chunk, MemOperand::offset_minus_one));
__ lwl(t3,
MemOperand(a1, 4, loadstore_chunk, MemOperand::offset_minus_one));
__ lwl(t4,
MemOperand(a1, 5, loadstore_chunk, MemOperand::offset_minus_one));
__ lwl(t5,
MemOperand(a1, 6, loadstore_chunk, MemOperand::offset_minus_one));
__ lwl(t6,
MemOperand(a1, 7, loadstore_chunk, MemOperand::offset_minus_one));
__ lwl(t7,
MemOperand(a1, 8, loadstore_chunk, MemOperand::offset_minus_one));
} else {
__ lwl(t0, MemOperand(a1));
__ lwl(t1, MemOperand(a1, 1, loadstore_chunk));
__ lwl(t2, MemOperand(a1, 2, loadstore_chunk));
if (pref_hint_store == kPrefHintPrepareForStore) {
__ sltu(v1, t9, a0);
__ Branch(USE_DELAY_SLOT, &ua_skip_pref, gt, v1, Operand(zero_reg));
}
__ lwl(t3, MemOperand(a1, 3, loadstore_chunk)); // Maybe in delay slot.
__ Pref(pref_hint_store, MemOperand(a0, 4 * pref_chunk));
__ Pref(pref_hint_store, MemOperand(a0, 5 * pref_chunk));
__ bind(&ua_skip_pref);
__ lwl(t4, MemOperand(a1, 4, loadstore_chunk));
__ lwl(t5, MemOperand(a1, 5, loadstore_chunk));
__ lwl(t6, MemOperand(a1, 6, loadstore_chunk));
__ lwl(t7, MemOperand(a1, 7, loadstore_chunk));
__ lwr(t0,
MemOperand(a1, 1, loadstore_chunk, MemOperand::offset_minus_one));
__ lwr(t1,
MemOperand(a1, 2, loadstore_chunk, MemOperand::offset_minus_one));
__ lwr(t2,
MemOperand(a1, 3, loadstore_chunk, MemOperand::offset_minus_one));
__ lwr(t3,
MemOperand(a1, 4, loadstore_chunk, MemOperand::offset_minus_one));
__ lwr(t4,
MemOperand(a1, 5, loadstore_chunk, MemOperand::offset_minus_one));
__ lwr(t5,
MemOperand(a1, 6, loadstore_chunk, MemOperand::offset_minus_one));
__ lwr(t6,
MemOperand(a1, 7, loadstore_chunk, MemOperand::offset_minus_one));
__ lwr(t7,
MemOperand(a1, 8, loadstore_chunk, MemOperand::offset_minus_one));
}
__ Pref(pref_hint_load, MemOperand(a1, 4 * pref_chunk));
__ sw(t0, MemOperand(a0));
__ sw(t1, MemOperand(a0, 1, loadstore_chunk));
__ sw(t2, MemOperand(a0, 2, loadstore_chunk));
__ sw(t3, MemOperand(a0, 3, loadstore_chunk));
__ sw(t4, MemOperand(a0, 4, loadstore_chunk));
__ sw(t5, MemOperand(a0, 5, loadstore_chunk));
__ sw(t6, MemOperand(a0, 6, loadstore_chunk));
__ sw(t7, MemOperand(a0, 7, loadstore_chunk));
if (kArchEndian == kLittle) {
__ lwr(t0, MemOperand(a1, 8, loadstore_chunk));
__ lwr(t1, MemOperand(a1, 9, loadstore_chunk));
__ lwr(t2, MemOperand(a1, 10, loadstore_chunk));
__ lwr(t3, MemOperand(a1, 11, loadstore_chunk));
__ lwr(t4, MemOperand(a1, 12, loadstore_chunk));
__ lwr(t5, MemOperand(a1, 13, loadstore_chunk));
__ lwr(t6, MemOperand(a1, 14, loadstore_chunk));
__ lwr(t7, MemOperand(a1, 15, loadstore_chunk));
__ lwl(t0,
MemOperand(a1, 9, loadstore_chunk, MemOperand::offset_minus_one));
__ lwl(t1,
MemOperand(a1, 10, loadstore_chunk, MemOperand::offset_minus_one));
__ lwl(t2,
MemOperand(a1, 11, loadstore_chunk, MemOperand::offset_minus_one));
__ lwl(t3,
MemOperand(a1, 12, loadstore_chunk, MemOperand::offset_minus_one));
__ lwl(t4,
MemOperand(a1, 13, loadstore_chunk, MemOperand::offset_minus_one));
__ lwl(t5,
MemOperand(a1, 14, loadstore_chunk, MemOperand::offset_minus_one));
__ lwl(t6,
MemOperand(a1, 15, loadstore_chunk, MemOperand::offset_minus_one));
__ lwl(t7,
MemOperand(a1, 16, loadstore_chunk, MemOperand::offset_minus_one));
} else {
__ lwl(t0, MemOperand(a1, 8, loadstore_chunk));
__ lwl(t1, MemOperand(a1, 9, loadstore_chunk));
__ lwl(t2, MemOperand(a1, 10, loadstore_chunk));
__ lwl(t3, MemOperand(a1, 11, loadstore_chunk));
__ lwl(t4, MemOperand(a1, 12, loadstore_chunk));
__ lwl(t5, MemOperand(a1, 13, loadstore_chunk));
__ lwl(t6, MemOperand(a1, 14, loadstore_chunk));
__ lwl(t7, MemOperand(a1, 15, loadstore_chunk));
__ lwr(t0,
MemOperand(a1, 9, loadstore_chunk, MemOperand::offset_minus_one));
__ lwr(t1,
MemOperand(a1, 10, loadstore_chunk, MemOperand::offset_minus_one));
__ lwr(t2,
MemOperand(a1, 11, loadstore_chunk, MemOperand::offset_minus_one));
__ lwr(t3,
MemOperand(a1, 12, loadstore_chunk, MemOperand::offset_minus_one));
__ lwr(t4,
MemOperand(a1, 13, loadstore_chunk, MemOperand::offset_minus_one));
__ lwr(t5,
MemOperand(a1, 14, loadstore_chunk, MemOperand::offset_minus_one));
__ lwr(t6,
MemOperand(a1, 15, loadstore_chunk, MemOperand::offset_minus_one));
__ lwr(t7,
MemOperand(a1, 16, loadstore_chunk, MemOperand::offset_minus_one));
}
__ Pref(pref_hint_load, MemOperand(a1, 5 * pref_chunk));
__ sw(t0, MemOperand(a0, 8, loadstore_chunk));
__ sw(t1, MemOperand(a0, 9, loadstore_chunk));
__ sw(t2, MemOperand(a0, 10, loadstore_chunk));
__ sw(t3, MemOperand(a0, 11, loadstore_chunk));
__ sw(t4, MemOperand(a0, 12, loadstore_chunk));
__ sw(t5, MemOperand(a0, 13, loadstore_chunk));
__ sw(t6, MemOperand(a0, 14, loadstore_chunk));
__ sw(t7, MemOperand(a0, 15, loadstore_chunk));
__ addiu(a0, a0, 16 * loadstore_chunk);
__ bne(a0, a3, &ua_loop16w);
__ addiu(a1, a1, 16 * loadstore_chunk); // In delay slot.
__ mov(a2, t8);
// Here less than 64-bytes. Check for
// a 32 byte chunk and copy if there is one. Otherwise jump down to
// ua_chk1w to handle the tail end of the copy.
__ bind(&ua_chkw);
__ Pref(pref_hint_load, MemOperand(a1));
__ andi(t8, a2, 0x1f);
__ beq(a2, t8, &ua_chk1w);
__ nop(); // In delay slot.
if (kArchEndian == kLittle) {
__ lwr(t0, MemOperand(a1));
__ lwr(t1, MemOperand(a1, 1, loadstore_chunk));
__ lwr(t2, MemOperand(a1, 2, loadstore_chunk));
__ lwr(t3, MemOperand(a1, 3, loadstore_chunk));
__ lwr(t4, MemOperand(a1, 4, loadstore_chunk));
__ lwr(t5, MemOperand(a1, 5, loadstore_chunk));
__ lwr(t6, MemOperand(a1, 6, loadstore_chunk));
__ lwr(t7, MemOperand(a1, 7, loadstore_chunk));
__ lwl(t0,
MemOperand(a1, 1, loadstore_chunk, MemOperand::offset_minus_one));
__ lwl(t1,
MemOperand(a1, 2, loadstore_chunk, MemOperand::offset_minus_one));
__ lwl(t2,
MemOperand(a1, 3, loadstore_chunk, MemOperand::offset_minus_one));
__ lwl(t3,
MemOperand(a1, 4, loadstore_chunk, MemOperand::offset_minus_one));
__ lwl(t4,
MemOperand(a1, 5, loadstore_chunk, MemOperand::offset_minus_one));
__ lwl(t5,
MemOperand(a1, 6, loadstore_chunk, MemOperand::offset_minus_one));
__ lwl(t6,
MemOperand(a1, 7, loadstore_chunk, MemOperand::offset_minus_one));
__ lwl(t7,
MemOperand(a1, 8, loadstore_chunk, MemOperand::offset_minus_one));
} else {
__ lwl(t0, MemOperand(a1));
__ lwl(t1, MemOperand(a1, 1, loadstore_chunk));
__ lwl(t2, MemOperand(a1, 2, loadstore_chunk));
__ lwl(t3, MemOperand(a1, 3, loadstore_chunk));
__ lwl(t4, MemOperand(a1, 4, loadstore_chunk));
__ lwl(t5, MemOperand(a1, 5, loadstore_chunk));
__ lwl(t6, MemOperand(a1, 6, loadstore_chunk));
__ lwl(t7, MemOperand(a1, 7, loadstore_chunk));
__ lwr(t0,
MemOperand(a1, 1, loadstore_chunk, MemOperand::offset_minus_one));
__ lwr(t1,
MemOperand(a1, 2, loadstore_chunk, MemOperand::offset_minus_one));
__ lwr(t2,
MemOperand(a1, 3, loadstore_chunk, MemOperand::offset_minus_one));
__ lwr(t3,
MemOperand(a1, 4, loadstore_chunk, MemOperand::offset_minus_one));
__ lwr(t4,
MemOperand(a1, 5, loadstore_chunk, MemOperand::offset_minus_one));
__ lwr(t5,
MemOperand(a1, 6, loadstore_chunk, MemOperand::offset_minus_one));
__ lwr(t6,
MemOperand(a1, 7, loadstore_chunk, MemOperand::offset_minus_one));
__ lwr(t7,
MemOperand(a1, 8, loadstore_chunk, MemOperand::offset_minus_one));
}
__ addiu(a1, a1, 8 * loadstore_chunk);
__ sw(t0, MemOperand(a0));
__ sw(t1, MemOperand(a0, 1, loadstore_chunk));
__ sw(t2, MemOperand(a0, 2, loadstore_chunk));
__ sw(t3, MemOperand(a0, 3, loadstore_chunk));
__ sw(t4, MemOperand(a0, 4, loadstore_chunk));
__ sw(t5, MemOperand(a0, 5, loadstore_chunk));
__ sw(t6, MemOperand(a0, 6, loadstore_chunk));
__ sw(t7, MemOperand(a0, 7, loadstore_chunk));
__ addiu(a0, a0, 8 * loadstore_chunk);
// Less than 32 bytes to copy. Set up for a loop to
// copy one word at a time.
__ bind(&ua_chk1w);
__ andi(a2, t8, loadstore_chunk - 1);
__ beq(a2, t8, &ua_smallCopy);
__ subu(a3, t8, a2); // In delay slot.
__ addu(a3, a0, a3);
__ bind(&ua_wordCopy_loop);
if (kArchEndian == kLittle) {
__ lwr(v1, MemOperand(a1));
__ lwl(v1,
MemOperand(a1, 1, loadstore_chunk, MemOperand::offset_minus_one));
} else {
__ lwl(v1, MemOperand(a1));
__ lwr(v1,
MemOperand(a1, 1, loadstore_chunk, MemOperand::offset_minus_one));
}
__ addiu(a0, a0, loadstore_chunk);
__ addiu(a1, a1, loadstore_chunk);
__ bne(a0, a3, &ua_wordCopy_loop);
__ sw(v1, MemOperand(a0, -1, loadstore_chunk)); // In delay slot.
// Copy the last 8 bytes.
__ bind(&ua_smallCopy);
__ beq(a2, zero_reg, &leave);
__ addu(a3, a0, a2); // In delay slot.
__ bind(&ua_smallCopy_loop);
__ lb(v1, MemOperand(a1));
__ addiu(a0, a0, 1);
__ addiu(a1, a1, 1);
__ bne(a0, a3, &ua_smallCopy_loop);
__ sb(v1, MemOperand(a0, -1)); // In delay slot.
__ jr(ra);
__ nop();
}
CodeDesc desc;
masm.GetCode(&desc);
ASSERT(!RelocInfo::RequiresRelocation(desc));
CpuFeatures::FlushICache(buffer, actual_size);
base::OS::ProtectCode(buffer, actual_size);
return FUNCTION_CAST<MemCopyUint8Function>(buffer);
#endif
}
#endif
UnaryMathFunction CreateSqrtFunction() {
#if defined(USE_SIMULATOR)
return &std::sqrt;
#else
size_t actual_size;
byte* buffer =
static_cast<byte*>(base::OS::Allocate(1 * KB, &actual_size, true));
if (buffer == NULL) return &std::sqrt;
MacroAssembler masm(NULL, buffer, static_cast<int>(actual_size));
__ MovFromFloatParameter(f12);
__ sqrt_d(f0, f12);
__ MovToFloatResult(f0);
__ Ret();
CodeDesc desc;
masm.GetCode(&desc);
ASSERT(!RelocInfo::RequiresRelocation(desc));
CpuFeatures::FlushICache(buffer, actual_size);
base::OS::ProtectCode(buffer, actual_size);
return FUNCTION_CAST<UnaryMathFunction>(buffer);
#endif
}
#undef __
// -------------------------------------------------------------------------
// Platform-specific RuntimeCallHelper functions.
void StubRuntimeCallHelper::BeforeCall(MacroAssembler* masm) const {
masm->EnterFrame(StackFrame::INTERNAL);
ASSERT(!masm->has_frame());
masm->set_has_frame(true);
}
void StubRuntimeCallHelper::AfterCall(MacroAssembler* masm) const {
masm->LeaveFrame(StackFrame::INTERNAL);
ASSERT(masm->has_frame());
masm->set_has_frame(false);
}
// -------------------------------------------------------------------------
// Code generators
#define __ ACCESS_MASM(masm)
void ElementsTransitionGenerator::GenerateMapChangeElementsTransition(
MacroAssembler* masm, AllocationSiteMode mode,
Label* allocation_memento_found) {
// ----------- S t a t e -------------
// -- a0 : value
// -- a1 : key
// -- a2 : receiver
// -- ra : return address
// -- a3 : target map, scratch for subsequent call
// -- t0 : scratch (elements)
// -----------------------------------
if (mode == TRACK_ALLOCATION_SITE) {
ASSERT(allocation_memento_found != NULL);
__ JumpIfJSArrayHasAllocationMemento(a2, t0, allocation_memento_found);
}
// Set transitioned map.
__ sw(a3, FieldMemOperand(a2, HeapObject::kMapOffset));
__ RecordWriteField(a2,
HeapObject::kMapOffset,
a3,
t5,
kRAHasNotBeenSaved,
kDontSaveFPRegs,
EMIT_REMEMBERED_SET,
OMIT_SMI_CHECK);
}
void ElementsTransitionGenerator::GenerateSmiToDouble(
MacroAssembler* masm, AllocationSiteMode mode, Label* fail) {
// ----------- S t a t e -------------
// -- a0 : value
// -- a1 : key
// -- a2 : receiver
// -- ra : return address
// -- a3 : target map, scratch for subsequent call
// -- t0 : scratch (elements)
// -----------------------------------
Label loop, entry, convert_hole, gc_required, only_change_map, done;
Register scratch = t6;
if (mode == TRACK_ALLOCATION_SITE) {
__ JumpIfJSArrayHasAllocationMemento(a2, t0, fail);
}
// Check for empty arrays, which only require a map transition and no changes
// to the backing store.
__ lw(t0, FieldMemOperand(a2, JSObject::kElementsOffset));
__ LoadRoot(at, Heap::kEmptyFixedArrayRootIndex);
__ Branch(&only_change_map, eq, at, Operand(t0));
__ push(ra);
__ lw(t1, FieldMemOperand(t0, FixedArray::kLengthOffset));
// t0: source FixedArray
// t1: number of elements (smi-tagged)
// Allocate new FixedDoubleArray.
__ sll(scratch, t1, 2);
__ Addu(scratch, scratch, FixedDoubleArray::kHeaderSize);
__ Allocate(scratch, t2, t3, t5, &gc_required, DOUBLE_ALIGNMENT);
// t2: destination FixedDoubleArray, not tagged as heap object
// Set destination FixedDoubleArray's length and map.
__ LoadRoot(t5, Heap::kFixedDoubleArrayMapRootIndex);
__ sw(t1, MemOperand(t2, FixedDoubleArray::kLengthOffset));
__ sw(t5, MemOperand(t2, HeapObject::kMapOffset));
// Update receiver's map.
__ sw(a3, FieldMemOperand(a2, HeapObject::kMapOffset));
__ RecordWriteField(a2,
HeapObject::kMapOffset,
a3,
t5,
kRAHasBeenSaved,
kDontSaveFPRegs,
OMIT_REMEMBERED_SET,
OMIT_SMI_CHECK);
// Replace receiver's backing store with newly created FixedDoubleArray.
__ Addu(a3, t2, Operand(kHeapObjectTag));
__ sw(a3, FieldMemOperand(a2, JSObject::kElementsOffset));
__ RecordWriteField(a2,
JSObject::kElementsOffset,
a3,
t5,
kRAHasBeenSaved,
kDontSaveFPRegs,
EMIT_REMEMBERED_SET,
OMIT_SMI_CHECK);
// Prepare for conversion loop.
__ Addu(a3, t0, Operand(FixedArray::kHeaderSize - kHeapObjectTag));
__ Addu(t3, t2, Operand(FixedDoubleArray::kHeaderSize));
__ sll(t2, t1, 2);
__ Addu(t2, t2, t3);
__ li(t0, Operand(kHoleNanLower32));
__ li(t1, Operand(kHoleNanUpper32));
// t0: kHoleNanLower32
// t1: kHoleNanUpper32
// t2: end of destination FixedDoubleArray, not tagged
// t3: begin of FixedDoubleArray element fields, not tagged
__ Branch(&entry);
__ bind(&only_change_map);
__ sw(a3, FieldMemOperand(a2, HeapObject::kMapOffset));
__ RecordWriteField(a2,
HeapObject::kMapOffset,
a3,
t5,
kRAHasNotBeenSaved,
kDontSaveFPRegs,
OMIT_REMEMBERED_SET,
OMIT_SMI_CHECK);
__ Branch(&done);
// Call into runtime if GC is required.
__ bind(&gc_required);
__ pop(ra);
__ Branch(fail);
// Convert and copy elements.
__ bind(&loop);
__ lw(t5, MemOperand(a3));
__ Addu(a3, a3, kIntSize);
// t5: current element
__ UntagAndJumpIfNotSmi(t5, t5, &convert_hole);
// Normal smi, convert to double and store.
__ mtc1(t5, f0);
__ cvt_d_w(f0, f0);
__ sdc1(f0, MemOperand(t3));
__ Addu(t3, t3, kDoubleSize);
__ Branch(&entry);
// Hole found, store the-hole NaN.
__ bind(&convert_hole);
if (FLAG_debug_code) {
// Restore a "smi-untagged" heap object.
__ SmiTag(t5);
__ Or(t5, t5, Operand(1));
__ LoadRoot(at, Heap::kTheHoleValueRootIndex);
__ Assert(eq, kObjectFoundInSmiOnlyArray, at, Operand(t5));
}
__ sw(t0, MemOperand(t3, Register::kMantissaOffset)); // mantissa
__ sw(t1, MemOperand(t3, Register::kExponentOffset)); // exponent
__ Addu(t3, t3, kDoubleSize);
__ bind(&entry);
__ Branch(&loop, lt, t3, Operand(t2));
__ pop(ra);
__ bind(&done);
}
void ElementsTransitionGenerator::GenerateDoubleToObject(
MacroAssembler* masm, AllocationSiteMode mode, Label* fail) {
// ----------- S t a t e -------------
// -- a0 : value
// -- a1 : key
// -- a2 : receiver
// -- ra : return address
// -- a3 : target map, scratch for subsequent call
// -- t0 : scratch (elements)
// -----------------------------------
Label entry, loop, convert_hole, gc_required, only_change_map;
if (mode == TRACK_ALLOCATION_SITE) {
__ JumpIfJSArrayHasAllocationMemento(a2, t0, fail);
}
// Check for empty arrays, which only require a map transition and no changes
// to the backing store.
__ lw(t0, FieldMemOperand(a2, JSObject::kElementsOffset));
__ LoadRoot(at, Heap::kEmptyFixedArrayRootIndex);
__ Branch(&only_change_map, eq, at, Operand(t0));
__ MultiPush(a0.bit() | a1.bit() | a2.bit() | a3.bit() | ra.bit());
__ lw(t1, FieldMemOperand(t0, FixedArray::kLengthOffset));
// t0: source FixedArray
// t1: number of elements (smi-tagged)
// Allocate new FixedArray.
__ sll(a0, t1, 1);
__ Addu(a0, a0, FixedDoubleArray::kHeaderSize);
__ Allocate(a0, t2, t3, t5, &gc_required, NO_ALLOCATION_FLAGS);
// t2: destination FixedArray, not tagged as heap object
// Set destination FixedDoubleArray's length and map.
__ LoadRoot(t5, Heap::kFixedArrayMapRootIndex);
__ sw(t1, MemOperand(t2, FixedDoubleArray::kLengthOffset));
__ sw(t5, MemOperand(t2, HeapObject::kMapOffset));
// Prepare for conversion loop.
__ Addu(t0, t0, Operand(
FixedDoubleArray::kHeaderSize - kHeapObjectTag
+ Register::kExponentOffset));
__ Addu(a3, t2, Operand(FixedArray::kHeaderSize));
__ Addu(t2, t2, Operand(kHeapObjectTag));
__ sll(t1, t1, 1);
__ Addu(t1, a3, t1);
__ LoadRoot(t3, Heap::kTheHoleValueRootIndex);
__ LoadRoot(t5, Heap::kHeapNumberMapRootIndex);
// Using offsetted addresses.
// a3: begin of destination FixedArray element fields, not tagged
// t0: begin of source FixedDoubleArray element fields, not tagged,
// points to the exponent
// t1: end of destination FixedArray, not tagged
// t2: destination FixedArray
// t3: the-hole pointer
// t5: heap number map
__ Branch(&entry);
// Call into runtime if GC is required.
__ bind(&gc_required);
__ MultiPop(a0.bit() | a1.bit() | a2.bit() | a3.bit() | ra.bit());
__ Branch(fail);
__ bind(&loop);
__ lw(a1, MemOperand(t0));
__ Addu(t0, t0, kDoubleSize);
// a1: current element's upper 32 bit
// t0: address of next element's upper 32 bit
__ Branch(&convert_hole, eq, a1, Operand(kHoleNanUpper32));
// Non-hole double, copy value into a heap number.
__ AllocateHeapNumber(a2, a0, t6, t5, &gc_required);
// a2: new heap number
// Load mantissa of current element, t0 point to exponent of next element.
__ lw(a0, MemOperand(t0, (Register::kMantissaOffset
- Register::kExponentOffset - kDoubleSize)));
__ sw(a0, FieldMemOperand(a2, HeapNumber::kMantissaOffset));
__ sw(a1, FieldMemOperand(a2, HeapNumber::kExponentOffset));
__ mov(a0, a3);
__ sw(a2, MemOperand(a3));
__ Addu(a3, a3, kIntSize);
__ RecordWrite(t2,
a0,
a2,
kRAHasBeenSaved,
kDontSaveFPRegs,
EMIT_REMEMBERED_SET,
OMIT_SMI_CHECK);
__ Branch(&entry);
// Replace the-hole NaN with the-hole pointer.
__ bind(&convert_hole);
__ sw(t3, MemOperand(a3));
__ Addu(a3, a3, kIntSize);
__ bind(&entry);
__ Branch(&loop, lt, a3, Operand(t1));
__ MultiPop(a2.bit() | a3.bit() | a0.bit() | a1.bit());
// Replace receiver's backing store with newly created and filled FixedArray.
__ sw(t2, FieldMemOperand(a2, JSObject::kElementsOffset));
__ RecordWriteField(a2,
JSObject::kElementsOffset,
t2,
t5,
kRAHasBeenSaved,
kDontSaveFPRegs,
EMIT_REMEMBERED_SET,
OMIT_SMI_CHECK);
__ pop(ra);
__ bind(&only_change_map);
// Update receiver's map.
__ sw(a3, FieldMemOperand(a2, HeapObject::kMapOffset));
__ RecordWriteField(a2,
HeapObject::kMapOffset,
a3,
t5,
kRAHasNotBeenSaved,
kDontSaveFPRegs,
OMIT_REMEMBERED_SET,
OMIT_SMI_CHECK);
}
void StringCharLoadGenerator::Generate(MacroAssembler* masm,
Register string,
Register index,
Register result,
Label* call_runtime) {
// Fetch the instance type of the receiver into result register.
__ lw(result, FieldMemOperand(string, HeapObject::kMapOffset));
__ lbu(result, FieldMemOperand(result, Map::kInstanceTypeOffset));
// We need special handling for indirect strings.
Label check_sequential;
__ And(at, result, Operand(kIsIndirectStringMask));
__ Branch(&check_sequential, eq, at, Operand(zero_reg));
// Dispatch on the indirect string shape: slice or cons.
Label cons_string;
__ And(at, result, Operand(kSlicedNotConsMask));
__ Branch(&cons_string, eq, at, Operand(zero_reg));
// Handle slices.
Label indirect_string_loaded;
__ lw(result, FieldMemOperand(string, SlicedString::kOffsetOffset));
__ lw(string, FieldMemOperand(string, SlicedString::kParentOffset));
__ sra(at, result, kSmiTagSize);
__ Addu(index, index, at);
__ jmp(&indirect_string_loaded);
// Handle cons strings.
// Check whether the right hand side is the empty string (i.e. if
// this is really a flat string in a cons string). If that is not
// the case we would rather go to the runtime system now to flatten
// the string.
__ bind(&cons_string);
__ lw(result, FieldMemOperand(string, ConsString::kSecondOffset));
__ LoadRoot(at, Heap::kempty_stringRootIndex);
__ Branch(call_runtime, ne, result, Operand(at));
// Get the first of the two strings and load its instance type.
__ lw(string, FieldMemOperand(string, ConsString::kFirstOffset));
__ bind(&indirect_string_loaded);
__ lw(result, FieldMemOperand(string, HeapObject::kMapOffset));
__ lbu(result, FieldMemOperand(result, Map::kInstanceTypeOffset));
// Distinguish sequential and external strings. Only these two string
// representations can reach here (slices and flat cons strings have been
// reduced to the underlying sequential or external string).
Label external_string, check_encoding;
__ bind(&check_sequential);
STATIC_ASSERT(kSeqStringTag == 0);
__ And(at, result, Operand(kStringRepresentationMask));
__ Branch(&external_string, ne, at, Operand(zero_reg));
// Prepare sequential strings
STATIC_ASSERT(SeqTwoByteString::kHeaderSize == SeqOneByteString::kHeaderSize);
__ Addu(string,
string,
SeqTwoByteString::kHeaderSize - kHeapObjectTag);
__ jmp(&check_encoding);