-
Notifications
You must be signed in to change notification settings - Fork 1.5k
/
Copy pathgemm_kernel_2x4_sse2.S
1790 lines (1572 loc) · 36.7 KB
/
gemm_kernel_2x4_sse2.S
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 2009, 2010 The University of Texas at Austin. */
/* All rights reserved. */
/* */
/* Redistribution and use in source and binary forms, with or */
/* without modification, are permitted provided that the following */
/* conditions are met: */
/* */
/* 1. Redistributions of source code must retain the above */
/* copyright notice, this list of conditions and the following */
/* disclaimer. */
/* */
/* 2. Redistributions in binary form must reproduce the above */
/* copyright notice, this list of conditions and the following */
/* disclaimer in the documentation and/or other materials */
/* provided with the distribution. */
/* */
/* THIS SOFTWARE IS PROVIDED BY THE UNIVERSITY OF TEXAS AT */
/* AUSTIN ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, */
/* INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF */
/* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE */
/* DISCLAIMED. IN NO EVENT SHALL THE UNIVERSITY OF TEXAS AT */
/* AUSTIN OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, */
/* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES */
/* (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE */
/* GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR */
/* BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF */
/* LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT */
/* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT */
/* OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE */
/* POSSIBILITY OF SUCH DAMAGE. */
/* */
/* The views and conclusions contained in the software and */
/* documentation are those of the authors and should not be */
/* interpreted as representing official policies, either expressed */
/* or implied, of The University of Texas at Austin. */
/*********************************************************************/
#define ASSEMBLER
#include "common.h"
#define STACK 16
#define ARGS 0
#define OLD_M 4 + STACK + ARGS(%esi)
#define OLD_N 8 + STACK + ARGS(%esi)
#define OLD_K 12 + STACK + ARGS(%esi)
#define OLD_ALPHA 16 + STACK + ARGS(%esi)
#define OLD_A 24 + STACK + ARGS(%esi)
#define OLD_B 28 + STACK + ARGS(%esi)
#define OLD_C 32 + STACK + ARGS(%esi)
#define OLD_LDC 36 + STACK + ARGS(%esi)
#define OLD_OFFT 40 + STACK + ARGS(%esi)
#define ALPHA 0(%esp)
#define K 16(%esp)
#define N 20(%esp)
#define M 24(%esp)
#define A 28(%esp)
#define C 32(%esp)
#define J 36(%esp)
#define BX 40(%esp)
#define OLD_STACK 44(%esp)
#define OFFSET 48(%esp)
#define KK 52(%esp)
#define KKK 56(%esp)
#define BUFFER 128(%esp)
#if defined(OPTERON) || defined(BARCELONA)
#define movsd movlpd
#endif
#if defined(OPTERON) || defined(BARCELONA)
#define PREFETCH prefetch
#define PREFETCHSIZE (8 * 10 + 4)
#endif
#define AA %edx
#define BB %ecx
#define LDC %ebp
#define KERNEL1(address) \
mulpd %xmm0, %xmm2; \
addpd %xmm2, %xmm4; \
PREFETCH (PREFETCHSIZE + 0) * SIZE + (address) * 1 * SIZE(AA); \
movapd 2 * SIZE + (address) * 4 * SIZE(BB), %xmm2; \
mulpd %xmm0, %xmm2; \
addpd %xmm2, %xmm5; \
movapd 4 * SIZE + (address) * 4 * SIZE(BB), %xmm2; \
mulpd %xmm0, %xmm2; \
mulpd 6 * SIZE + (address) * 4 * SIZE(BB), %xmm0; \
addpd %xmm2, %xmm6; \
movapd 16 * SIZE + (address) * 4 * SIZE(BB), %xmm2; \
addpd %xmm0, %xmm7; \
movapd 2 * SIZE + (address) * 1 * SIZE(AA), %xmm0
#define KERNEL2(address) \
mulpd %xmm0, %xmm3; \
addpd %xmm3, %xmm4; \
movapd 10 * SIZE + (address) * 4 * SIZE(BB), %xmm3; \
mulpd %xmm0, %xmm3; \
addpd %xmm3, %xmm5; \
movapd 12 * SIZE + (address) * 4 * SIZE(BB), %xmm3; \
mulpd %xmm0, %xmm3; \
mulpd 14 * SIZE + (address) * 4 * SIZE(BB), %xmm0; \
addpd %xmm3, %xmm6; \
movapd 24 * SIZE + (address) * 4 * SIZE(BB), %xmm3; \
addpd %xmm0, %xmm7; \
movapd 4 * SIZE + (address) * 1 * SIZE(AA), %xmm0
#define KERNEL3(address) \
mulpd %xmm0, %xmm2; \
addpd %xmm2, %xmm4; \
movapd 18 * SIZE + (address) * 4 * SIZE(BB), %xmm2; \
mulpd %xmm0, %xmm2; \
addpd %xmm2, %xmm5; \
movapd 20 * SIZE + (address) * 4 * SIZE(BB), %xmm2; \
mulpd %xmm0, %xmm2; \
mulpd 22 * SIZE + (address) * 4 * SIZE(BB), %xmm0; \
addpd %xmm2, %xmm6; \
movapd 32 * SIZE + (address) * 4 * SIZE(BB), %xmm2; \
addpd %xmm0, %xmm7; \
movapd 6 * SIZE + (address) * 1 * SIZE(AA), %xmm0
#define KERNEL4(address) \
mulpd %xmm0, %xmm3; \
addpd %xmm3, %xmm4; \
movapd 26 * SIZE + (address) * 4 * SIZE(BB), %xmm3; \
mulpd %xmm0, %xmm3; \
addpd %xmm3, %xmm5; \
movapd 28 * SIZE + (address) * 4 * SIZE(BB), %xmm3; \
mulpd %xmm0, %xmm3; \
mulpd 30 * SIZE + (address) * 4 * SIZE(BB), %xmm0; \
addpd %xmm3, %xmm6; \
movapd 40 * SIZE + (address) * 4 * SIZE(BB), %xmm3; \
addpd %xmm0, %xmm7; \
movapd 16 * SIZE + (address) * 1 * SIZE(AA), %xmm0
#define KERNEL5(address) \
PREFETCH (PREFETCHSIZE + 8) * SIZE + (address) * 1 * SIZE(AA); \
mulpd %xmm1, %xmm2; \
addpd %xmm2, %xmm4; \
movapd 34 * SIZE + (address) * 4 * SIZE(BB), %xmm2; \
mulpd %xmm1, %xmm2; \
addpd %xmm2, %xmm5; \
movapd 36 * SIZE + (address) * 4 * SIZE(BB), %xmm2; \
mulpd %xmm1, %xmm2; \
mulpd 38 * SIZE + (address) * 4 * SIZE(BB), %xmm1; \
addpd %xmm2, %xmm6; \
movapd 48 * SIZE + (address) * 4 * SIZE(BB), %xmm2; \
addpd %xmm1, %xmm7; \
movapd 10 * SIZE + (address) * 1 * SIZE(AA), %xmm1
#define KERNEL6(address) \
mulpd %xmm1, %xmm3; \
addpd %xmm3, %xmm4; \
movapd 42 * SIZE + (address) * 4 * SIZE(BB), %xmm3; \
mulpd %xmm1, %xmm3; \
addpd %xmm3, %xmm5; \
movapd 44 * SIZE + (address) * 4 * SIZE(BB), %xmm3; \
mulpd %xmm1, %xmm3; \
mulpd 46 * SIZE + (address) * 4 * SIZE(BB), %xmm1; \
addpd %xmm3, %xmm6; \
movapd 56 * SIZE + (address) * 4 * SIZE(BB), %xmm3; \
addpd %xmm1, %xmm7; \
movapd 12 * SIZE + (address) * 1 * SIZE(AA), %xmm1
#define KERNEL7(address) \
mulpd %xmm1, %xmm2; \
addpd %xmm2, %xmm4; \
movapd 50 * SIZE + (address) * 4 * SIZE(BB), %xmm2; \
mulpd %xmm1, %xmm2; \
addpd %xmm2, %xmm5; \
movapd 52 * SIZE + (address) * 4 * SIZE(BB), %xmm2; \
mulpd %xmm1, %xmm2; \
mulpd 54 * SIZE + (address) * 4 * SIZE(BB), %xmm1; \
addpd %xmm2, %xmm6; \
movapd 64 * SIZE + (address) * 4 * SIZE(BB), %xmm2; \
addpd %xmm1, %xmm7; \
movapd 14 * SIZE + (address) * 1 * SIZE(AA), %xmm1
#define KERNEL8(address) \
mulpd %xmm1, %xmm3; \
addpd %xmm3, %xmm4; \
movapd 58 * SIZE + (address) * 4 * SIZE(BB), %xmm3; \
mulpd %xmm1, %xmm3; \
addpd %xmm3, %xmm5; \
movapd 60 * SIZE + (address) * 4 * SIZE(BB), %xmm3; \
mulpd %xmm1, %xmm3; \
mulpd 62 * SIZE + (address) * 4 * SIZE(BB), %xmm1; \
addpd %xmm3, %xmm6; \
movapd 72 * SIZE + (address) * 4 * SIZE(BB), %xmm3; \
addpd %xmm1, %xmm7; \
movapd 24 * SIZE + (address) * 1 * SIZE(AA), %xmm1
PROLOGUE
pushl %ebp
pushl %edi
pushl %esi
pushl %ebx
PROFCODE
EMMS
movl %esp, %esi # save old stack
subl $128 + LOCAL_BUFFER_SIZE, %esp
andl $-1024, %esp # align stack
STACK_TOUCHING
movl OLD_M, %ebx
movl OLD_N, %eax
movl OLD_K, %ecx
movl OLD_A, %edx
movsd OLD_ALPHA, %xmm3
movl %ebx, M
movl %eax, N
movl %ecx, K
movl %edx, A
movl %esi, OLD_STACK
#ifdef TRMMKERNEL
movss OLD_OFFT, %xmm4
#endif
unpcklpd %xmm3, %xmm3
movl OLD_B, %edi
movl OLD_C, %ebx
movapd %xmm3, ALPHA
movl %ebx, C
movl OLD_LDC, LDC
#ifdef TRMMKERNEL
movss %xmm4, OFFSET
movss %xmm4, KK
#ifndef LEFT
negl KK
#endif
#endif
leal (, LDC, SIZE), LDC
sarl $2, %eax
movl %eax, J
jle .L30
ALIGN_2
.L01:
#if defined(TRMMKERNEL) && defined(LEFT)
movl OFFSET, %eax
movl %eax, KK
#endif
/* Copying to Sub Buffer */
movl K, %eax
leal BUFFER, %ecx
sarl $1, %eax
jle .L05
ALIGN_4
.L02:
#define COPYPREFETCH 40
prefetchnta (COPYPREFETCH) * SIZE(%edi)
movq 0 * SIZE(%edi), %mm0
movq 1 * SIZE(%edi), %mm1
movq 2 * SIZE(%edi), %mm2
movq 3 * SIZE(%edi), %mm3
movq 4 * SIZE(%edi), %mm4
movq 5 * SIZE(%edi), %mm5
movq 6 * SIZE(%edi), %mm6
movq 7 * SIZE(%edi), %mm7
movq %mm0, 0 * SIZE(%ecx)
movq %mm0, 1 * SIZE(%ecx)
movq %mm1, 2 * SIZE(%ecx)
movq %mm1, 3 * SIZE(%ecx)
movq %mm2, 4 * SIZE(%ecx)
movq %mm2, 5 * SIZE(%ecx)
movq %mm3, 6 * SIZE(%ecx)
movq %mm3, 7 * SIZE(%ecx)
movq %mm4, 8 * SIZE(%ecx)
movq %mm4, 9 * SIZE(%ecx)
movq %mm5, 10 * SIZE(%ecx)
movq %mm5, 11 * SIZE(%ecx)
movq %mm6, 12 * SIZE(%ecx)
movq %mm6, 13 * SIZE(%ecx)
movq %mm7, 14 * SIZE(%ecx)
movq %mm7, 15 * SIZE(%ecx)
addl $ 8 * SIZE, %edi
addl $16 * SIZE, %ecx
decl %eax
jne .L02
ALIGN_2
.L05:
movl K, %eax
andl $1, %eax
BRANCH
jle .L10
movq 0 * SIZE(%edi), %mm0
movq 1 * SIZE(%edi), %mm1
movq 2 * SIZE(%edi), %mm2
movq 3 * SIZE(%edi), %mm3
movq %mm0, 0 * SIZE(%ecx)
movq %mm0, 1 * SIZE(%ecx)
movq %mm1, 2 * SIZE(%ecx)
movq %mm1, 3 * SIZE(%ecx)
movq %mm2, 4 * SIZE(%ecx)
movq %mm2, 5 * SIZE(%ecx)
movq %mm3, 6 * SIZE(%ecx)
movq %mm3, 7 * SIZE(%ecx)
addl $4 * SIZE, %edi
ALIGN_4
.L10:
movl %edi, BX
movl C, %esi # coffset = c
movl A, AA # aoffset = a
movl M, %ebx
sarl $1, %ebx # i = (m >> 2)
jle .L20
ALIGN_4
.L11:
#if !defined(TRMMKERNEL) || \
(defined(TRMMKERNEL) && defined(LEFT) && defined(TRANSA)) || \
(defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
leal BUFFER, BB
#else
leal BUFFER, BB
movl KK, %eax
leal (, %eax, SIZE), %eax
leal (AA, %eax, 2), AA
leal (BB, %eax, 8), BB
#endif
movl BX, %eax
prefetchnta 0 * SIZE(%eax)
prefetchnta 8 * SIZE(%eax)
subl $-8 * SIZE, BX
pxor %xmm4, %xmm4
pxor %xmm5, %xmm5
pxor %xmm6, %xmm6
pxor %xmm7, %xmm7
movapd 0 * SIZE(AA), %xmm0
movapd 8 * SIZE(AA), %xmm1
movapd 0 * SIZE(BB), %xmm2
movapd 8 * SIZE(BB), %xmm3
leal (LDC, LDC, 2), %eax
prefetchw 1 * SIZE(%esi)
prefetchw 1 * SIZE(%esi, LDC)
prefetchw 1 * SIZE(%esi, LDC, 2)
prefetchw 1 * SIZE(%esi, %eax)
#ifndef TRMMKERNEL
movl K, %eax
#elif (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
movl K, %eax
subl KK, %eax
movl %eax, KKK
#else
movl KK, %eax
#ifdef LEFT
addl $2, %eax
#else
addl $4, %eax
#endif
movl %eax, KKK
#endif
#if 1
andl $-8, %eax
sall $4, %eax
je .L15
.L1X:
KERNEL1(16 * 0)
KERNEL2(16 * 0)
KERNEL3(16 * 0)
KERNEL4(16 * 0)
KERNEL5(16 * 0)
KERNEL6(16 * 0)
KERNEL7(16 * 0)
KERNEL8(16 * 0)
cmpl $128 * 1, %eax
jle .L12
KERNEL1(16 * 1)
KERNEL2(16 * 1)
KERNEL3(16 * 1)
KERNEL4(16 * 1)
KERNEL5(16 * 1)
KERNEL6(16 * 1)
KERNEL7(16 * 1)
KERNEL8(16 * 1)
cmpl $128 * 2, %eax
jle .L12
KERNEL1(16 * 2)
KERNEL2(16 * 2)
KERNEL3(16 * 2)
KERNEL4(16 * 2)
KERNEL5(16 * 2)
KERNEL6(16 * 2)
KERNEL7(16 * 2)
KERNEL8(16 * 2)
cmpl $128 * 3, %eax
jle .L12
KERNEL1(16 * 3)
KERNEL2(16 * 3)
KERNEL3(16 * 3)
KERNEL4(16 * 3)
KERNEL5(16 * 3)
KERNEL6(16 * 3)
KERNEL7(16 * 3)
KERNEL8(16 * 3)
cmpl $128 * 4, %eax
jle .L12
KERNEL1(16 * 4)
KERNEL2(16 * 4)
KERNEL3(16 * 4)
KERNEL4(16 * 4)
KERNEL5(16 * 4)
KERNEL6(16 * 4)
KERNEL7(16 * 4)
KERNEL8(16 * 4)
cmpl $128 * 5, %eax
jle .L12
KERNEL1(16 * 5)
KERNEL2(16 * 5)
KERNEL3(16 * 5)
KERNEL4(16 * 5)
KERNEL5(16 * 5)
KERNEL6(16 * 5)
KERNEL7(16 * 5)
KERNEL8(16 * 5)
cmpl $128 * 6, %eax
jle .L12
KERNEL1(16 * 6)
KERNEL2(16 * 6)
KERNEL3(16 * 6)
KERNEL4(16 * 6)
KERNEL5(16 * 6)
KERNEL6(16 * 6)
KERNEL7(16 * 6)
KERNEL8(16 * 6)
cmpl $128 * 7, %eax
jle .L12
KERNEL1(16 * 7)
KERNEL2(16 * 7)
KERNEL3(16 * 7)
KERNEL4(16 * 7)
KERNEL5(16 * 7)
KERNEL6(16 * 7)
KERNEL7(16 * 7)
KERNEL8(16 * 7)
addl $128 * 4 * SIZE, BB
addl $128 * 1 * SIZE, AA
subl $128 * 8, %eax
jg .L1X
jmp .L15
.L12:
leal (AA, %eax, 1), AA
leal (BB, %eax, 4), BB
ALIGN_4
#else
sarl $3, %eax
je .L15
ALIGN_4
.L12:
KERNEL1(16 * 0)
KERNEL2(16 * 0)
KERNEL3(16 * 0)
KERNEL4(16 * 0)
KERNEL5(16 * 0)
KERNEL6(16 * 0)
KERNEL7(16 * 0)
KERNEL8(16 * 0)
addl $64 * SIZE, BB
addl $16 * SIZE, AA
decl %eax
jne .L12
ALIGN_4
#endif
.L15:
#ifndef TRMMKERNEL
movl K, %eax
#else
movl KKK, %eax
#endif
movapd ALPHA, %xmm3
andl $7, %eax # if (k & 1)
BRANCH
je .L18
ALIGN_3
.L16:
mulpd %xmm0, %xmm2
addpd %xmm2, %xmm4
movapd 2 * SIZE(BB), %xmm2
mulpd %xmm0, %xmm2
addpd %xmm2, %xmm5
movapd 4 * SIZE(BB), %xmm2
mulpd %xmm0, %xmm2
mulpd 6 * SIZE(BB), %xmm0
addpd %xmm2, %xmm6
movapd 8 * SIZE(BB), %xmm2
addpd %xmm0, %xmm7
movapd 2 * SIZE(AA), %xmm0
addl $2 * SIZE, AA
addl $8 * SIZE, BB
decl %eax
jg .L16
ALIGN_4
.L18:
leal (LDC, LDC, 2), %eax
#ifndef TRMMKERNEL
mulpd %xmm3, %xmm4
movsd 0 * SIZE(%esi), %xmm0
movhpd 1 * SIZE(%esi), %xmm0
mulpd %xmm3, %xmm5
movsd 0 * SIZE(%esi, LDC, 1), %xmm1
movhpd 1 * SIZE(%esi, LDC, 1), %xmm1
mulpd %xmm3, %xmm6
movsd 0 * SIZE(%esi, LDC, 2), %xmm2
movhpd 1 * SIZE(%esi, LDC, 2), %xmm2
mulpd %xmm3, %xmm7
movsd 0 * SIZE(%esi, %eax, 1), %xmm3
movhpd 1 * SIZE(%esi, %eax, 1), %xmm3
addpd %xmm0, %xmm4
addpd %xmm1, %xmm5
addpd %xmm2, %xmm6
addpd %xmm3, %xmm7
#else
mulpd %xmm3, %xmm4
mulpd %xmm3, %xmm5
mulpd %xmm3, %xmm6
mulpd %xmm3, %xmm7
#endif
movsd %xmm4, 0 * SIZE(%esi)
movhpd %xmm4, 1 * SIZE(%esi)
movsd %xmm5, 0 * SIZE(%esi, LDC, 1)
movhpd %xmm5, 1 * SIZE(%esi, LDC, 1)
movsd %xmm6, 0 * SIZE(%esi, LDC, 2)
movhpd %xmm6, 1 * SIZE(%esi, LDC, 2)
movsd %xmm7, 0 * SIZE(%esi, %eax, 1)
movhpd %xmm7, 1 * SIZE(%esi, %eax, 1)
#if (defined(TRMMKERNEL) && defined(LEFT) && defined(TRANSA)) || \
(defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
movl K, %eax
subl KKK, %eax
leal (,%eax, SIZE), %eax
leal (AA, %eax, 2), AA
leal (BB, %eax, 8), BB
#endif
#if defined(TRMMKERNEL) && defined(LEFT)
addl $2, KK
#endif
addl $2 * SIZE, %esi # coffset += 2
decl %ebx # i --
jg .L11
ALIGN_4
.L20:
movl M, %ebx
testl $1, %ebx # i = (m >> 2)
jle .L29
#if !defined(TRMMKERNEL) || \
(defined(TRMMKERNEL) && defined(LEFT) && defined(TRANSA)) || \
(defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
leal BUFFER, BB
#else
leal BUFFER, BB
movl KK, %eax
leal (, %eax, SIZE), %eax
leal (AA, %eax, 1), AA
leal (BB, %eax, 8), BB
#endif
pxor %xmm4, %xmm4
pxor %xmm5, %xmm5
pxor %xmm6, %xmm6
pxor %xmm7, %xmm7
leal (LDC, LDC, 2), %eax
movsd 0 * SIZE(AA), %xmm0
movsd 4 * SIZE(AA), %xmm1
movsd 0 * SIZE(BB), %xmm2
movsd 8 * SIZE(BB), %xmm3
#ifndef TRMMKERNEL
movl K, %eax
#elif (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
movl K, %eax
subl KK, %eax
movl %eax, KKK
#else
movl KK, %eax
#ifdef LEFT
addl $1, %eax
#else
addl $4, %eax
#endif
movl %eax, KKK
#endif
sarl $3, %eax
je .L25
ALIGN_4
.L22:
mulsd %xmm0, %xmm2
addsd %xmm2, %xmm4
#if defined(OPTERON) || defined(BARCELONA)
PREFETCH (PREFETCHSIZE + 0) * SIZE(AA)
#endif
movsd 2 * SIZE(BB), %xmm2
mulsd %xmm0, %xmm2
addsd %xmm2, %xmm5
movsd 4 * SIZE(BB), %xmm2
mulsd %xmm0, %xmm2
mulsd 6 * SIZE(BB), %xmm0
addsd %xmm2, %xmm6
movsd 16 * SIZE(BB), %xmm2
addsd %xmm0, %xmm7
movsd 1 * SIZE(AA), %xmm0
mulsd %xmm0, %xmm3
addsd %xmm3, %xmm4
movsd 10 * SIZE(BB), %xmm3
mulsd %xmm0, %xmm3
addsd %xmm3, %xmm5
movsd 12 * SIZE(BB), %xmm3
mulsd %xmm0, %xmm3
mulsd 14 * SIZE(BB), %xmm0
addsd %xmm3, %xmm6
movsd 24 * SIZE(BB), %xmm3
addsd %xmm0, %xmm7
movsd 2 * SIZE(AA), %xmm0
mulsd %xmm0, %xmm2
addsd %xmm2, %xmm4
movsd 18 * SIZE(BB), %xmm2
mulsd %xmm0, %xmm2
addsd %xmm2, %xmm5
movsd 20 * SIZE(BB), %xmm2
mulsd %xmm0, %xmm2
mulsd 22 * SIZE(BB), %xmm0
addsd %xmm2, %xmm6
movsd 32 * SIZE(BB), %xmm2
addsd %xmm0, %xmm7
movsd 3 * SIZE(AA), %xmm0
mulsd %xmm0, %xmm3
addsd %xmm3, %xmm4
movsd 26 * SIZE(BB), %xmm3
mulsd %xmm0, %xmm3
addsd %xmm3, %xmm5
movsd 28 * SIZE(BB), %xmm3
mulsd %xmm0, %xmm3
mulsd 30 * SIZE(BB), %xmm0
addsd %xmm3, %xmm6
movsd 40 * SIZE(BB), %xmm3
addsd %xmm0, %xmm7
movsd 8 * SIZE(AA), %xmm0
#if defined(OPTERON) || defined(BARCELONA)
PREFETCH (PREFETCHSIZE + 8) * SIZE(AA)
#endif
mulsd %xmm1, %xmm2
addsd %xmm2, %xmm4
movsd 34 * SIZE(BB), %xmm2
mulsd %xmm1, %xmm2
addsd %xmm2, %xmm5
movsd 36 * SIZE(BB), %xmm2
mulsd %xmm1, %xmm2
mulsd 38 * SIZE(BB), %xmm1
addsd %xmm2, %xmm6
movsd 48 * SIZE(BB), %xmm2
addsd %xmm1, %xmm7
movsd 5 * SIZE(AA), %xmm1
mulsd %xmm1, %xmm3
addsd %xmm3, %xmm4
movsd 42 * SIZE(BB), %xmm3
mulsd %xmm1, %xmm3
addsd %xmm3, %xmm5
movsd 44 * SIZE(BB), %xmm3
mulsd %xmm1, %xmm3
mulsd 46 * SIZE(BB), %xmm1
addsd %xmm3, %xmm6
movsd 56 * SIZE(BB), %xmm3
addsd %xmm1, %xmm7
movsd 6 * SIZE(AA), %xmm1
mulsd %xmm1, %xmm2
addsd %xmm2, %xmm4
movsd 50 * SIZE(BB), %xmm2
mulsd %xmm1, %xmm2
addsd %xmm2, %xmm5
movsd 52 * SIZE(BB), %xmm2
mulsd %xmm1, %xmm2
mulsd 54 * SIZE(BB), %xmm1
addsd %xmm2, %xmm6
movsd 64 * SIZE(BB), %xmm2
addsd %xmm1, %xmm7
movsd 7 * SIZE(AA), %xmm1
mulsd %xmm1, %xmm3
addsd %xmm3, %xmm4
movsd 58 * SIZE(BB), %xmm3
mulsd %xmm1, %xmm3
addsd %xmm3, %xmm5
movsd 60 * SIZE(BB), %xmm3
mulsd %xmm1, %xmm3
mulsd 62 * SIZE(BB), %xmm1
addsd %xmm3, %xmm6
movsd 72 * SIZE(BB), %xmm3
addl $64 * SIZE, BB
addsd %xmm1, %xmm7
movsd 12 * SIZE(AA), %xmm1
addl $8 * SIZE, AA
decl %eax
jne .L22
ALIGN_4
.L25:
#ifndef TRMMKERNEL
movl K, %eax
#else
movl KKK, %eax
#endif
movsd ALPHA, %xmm3
andl $7, %eax # if (k & 1)
BRANCH
je .L28
.L26:
mulsd %xmm0, %xmm2
addsd %xmm2, %xmm4
movsd 2 * SIZE(BB), %xmm2
mulsd %xmm0, %xmm2
addsd %xmm2, %xmm5
movsd 4 * SIZE(BB), %xmm2
mulsd %xmm0, %xmm2
mulsd 6 * SIZE(BB), %xmm0
addsd %xmm2, %xmm6
movsd 8 * SIZE(BB), %xmm2
addsd %xmm0, %xmm7
movsd 1 * SIZE(AA), %xmm0
addl $1 * SIZE, AA
addl $8 * SIZE, BB
decl %eax
jg .L26
ALIGN_4
.L28:
leal (LDC, LDC, 2), %eax
#ifndef TRMMKERNEL
mulsd %xmm3, %xmm4
movsd 0 * SIZE(%esi), %xmm0
mulsd %xmm3, %xmm5
movsd 0 * SIZE(%esi, LDC, 1), %xmm1
mulsd %xmm3, %xmm6
movsd 0 * SIZE(%esi, LDC, 2), %xmm2
mulsd %xmm3, %xmm7
movsd 0 * SIZE(%esi, %eax, 1), %xmm3
addsd %xmm0, %xmm4
addsd %xmm1, %xmm5
addsd %xmm2, %xmm6
addsd %xmm3, %xmm7
#else
mulsd %xmm3, %xmm4
mulsd %xmm3, %xmm5
mulsd %xmm3, %xmm6
mulsd %xmm3, %xmm7
#endif
movsd %xmm4, 0 * SIZE(%esi)
movsd %xmm5, 0 * SIZE(%esi, LDC, 1)
movsd %xmm6, 0 * SIZE(%esi, LDC, 2)
movsd %xmm7, 0 * SIZE(%esi, %eax, 1)
#if (defined(TRMMKERNEL) && defined(LEFT) && defined(TRANSA)) || \
(defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
movl K, %eax
subl KKK, %eax
leal (,%eax, SIZE), %eax
leal (AA, %eax, 1), AA
leal (BB, %eax, 8), BB
#endif
#if defined(TRMMKERNEL) && defined(LEFT)
addl $1, KK
#endif
ALIGN_4
.L29:
#if defined(TRMMKERNEL) && !defined(LEFT)
addl $4, KK
#endif
leal (, LDC, 4), %eax
addl %eax, C # c += 4 * ldc
decl J # j --
jg .L01
ALIGN_4
.L30:
testl $2, N
je .L60
ALIGN_2
.L31:
#if defined(TRMMKERNEL) && defined(LEFT)
movl OFFSET, %eax
movl %eax, KK
#endif
/* Copying to Sub Buffer */
movl K, %eax
leal BUFFER, %ecx
sarl $2, %eax
jle .L35
ALIGN_4
.L32:
#ifdef PENTIUM4
#ifdef HAVE_SSE3
movddup 0 * SIZE(%edi), %xmm0
movddup 1 * SIZE(%edi), %xmm1
movddup 2 * SIZE(%edi), %xmm2
movddup 3 * SIZE(%edi), %xmm3
movddup 4 * SIZE(%edi), %xmm4
movddup 5 * SIZE(%edi), %xmm5
movddup 6 * SIZE(%edi), %xmm6
movddup 7 * SIZE(%edi), %xmm7
movapd %xmm0, 0 * SIZE(%ecx)
movapd %xmm1, 2 * SIZE(%ecx)
movapd %xmm2, 4 * SIZE(%ecx)
movapd %xmm3, 6 * SIZE(%ecx)
movapd %xmm4, 8 * SIZE(%ecx)
movapd %xmm5, 10 * SIZE(%ecx)
movapd %xmm6, 12 * SIZE(%ecx)
movapd %xmm7, 14 * SIZE(%ecx)
#else
movsd 0 * SIZE(%edi), %xmm0
movsd 1 * SIZE(%edi), %xmm1
movsd 2 * SIZE(%edi), %xmm2
movsd 3 * SIZE(%edi), %xmm3
movsd 4 * SIZE(%edi), %xmm4
movsd 5 * SIZE(%edi), %xmm5
movsd 6 * SIZE(%edi), %xmm6
movsd 7 * SIZE(%edi), %xmm7
unpcklpd %xmm0, %xmm0
unpckhpd %xmm1, %xmm1
unpcklpd %xmm2, %xmm2
unpckhpd %xmm3, %xmm3
unpcklpd %xmm4, %xmm4
unpckhpd %xmm5, %xmm5
unpcklpd %xmm6, %xmm6
unpckhpd %xmm7, %xmm7
movapd %xmm0, 0 * SIZE(%ecx)
movapd %xmm1, 2 * SIZE(%ecx)
movapd %xmm2, 4 * SIZE(%ecx)
movapd %xmm3, 6 * SIZE(%ecx)
movapd %xmm4, 8 * SIZE(%ecx)
movapd %xmm5, 10 * SIZE(%ecx)
movapd %xmm6, 12 * SIZE(%ecx)
movapd %xmm7, 14 * SIZE(%ecx)
#endif
prefetcht0 80 * SIZE(%edi)
prefetcht1 112 * SIZE(%ecx)
#endif
#if defined(OPTERON) || defined(BARCELONA)
#define COPYPREFETCH 40
prefetchnta (COPYPREFETCH) * SIZE(%edi)
movq 0 * SIZE(%edi), %mm0
movq 1 * SIZE(%edi), %mm1
movq 2 * SIZE(%edi), %mm2
movq 3 * SIZE(%edi), %mm3
movq 4 * SIZE(%edi), %mm4
movq 5 * SIZE(%edi), %mm5
movq 6 * SIZE(%edi), %mm6
movq 7 * SIZE(%edi), %mm7
movq %mm0, 0 * SIZE(%ecx)
movq %mm0, 1 * SIZE(%ecx)
movq %mm1, 2 * SIZE(%ecx)
movq %mm1, 3 * SIZE(%ecx)
movq %mm2, 4 * SIZE(%ecx)
movq %mm2, 5 * SIZE(%ecx)
movq %mm3, 6 * SIZE(%ecx)
movq %mm3, 7 * SIZE(%ecx)
movq %mm4, 8 * SIZE(%ecx)
movq %mm4, 9 * SIZE(%ecx)
movq %mm5, 10 * SIZE(%ecx)
movq %mm5, 11 * SIZE(%ecx)
movq %mm6, 12 * SIZE(%ecx)
movq %mm6, 13 * SIZE(%ecx)
movq %mm7, 14 * SIZE(%ecx)
movq %mm7, 15 * SIZE(%ecx)
#endif
addl $ 8 * SIZE, %edi
addl $16 * SIZE, %ecx
decl %eax
jne .L32
ALIGN_2
.L35:
movl K, %eax
andl $3, %eax
BRANCH
jle .L40
ALIGN_2
.L36:
#ifdef PENTIUM4
#ifdef HAVE_SSE3
movddup 0 * SIZE(%edi), %xmm0
movddup 1 * SIZE(%edi), %xmm1
movapd %xmm0, 0 * SIZE(%ecx)
movapd %xmm1, 2 * SIZE(%ecx)
#else
movsd 0 * SIZE(%edi), %xmm0
movsd 1 * SIZE(%edi), %xmm1
unpcklpd %xmm0, %xmm0
unpckhpd %xmm1, %xmm1
movapd %xmm0, 0 * SIZE(%ecx)
movapd %xmm1, 2 * SIZE(%ecx)
#endif
#endif
#if defined(OPTERON) || defined(BARCELONA)
movq 0 * SIZE(%edi), %mm0
movq 1 * SIZE(%edi), %mm1
movq %mm0, 0 * SIZE(%ecx)
movq %mm0, 1 * SIZE(%ecx)
movq %mm1, 2 * SIZE(%ecx)
movq %mm1, 3 * SIZE(%ecx)
#endif
addl $2 * SIZE, %edi
addl $4 * SIZE, %ecx
decl %eax
jne .L36
ALIGN_4
.L40:
movl C, %esi # coffset = c
movl A, AA # aoffset = a
movl M, %ebx
sarl $1, %ebx # i = (m >> 2)
jle .L50
ALIGN_4
.L41:
#if !defined(TRMMKERNEL) || \
(defined(TRMMKERNEL) && defined(LEFT) && defined(TRANSA)) || \
(defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
leal BUFFER, BB
#else
leal BUFFER, BB