/
msa_macro.h
1392 lines (1275 loc) · 62.7 KB
/
msa_macro.h
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 2016 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// MSA common macros
//
// Author(s): Prashant Patil (prashant.patil@imgtec.com)
#ifndef WEBP_DSP_MSA_MACRO_H_
#define WEBP_DSP_MSA_MACRO_H_
#include <stdint.h>
#include <msa.h>
#if defined(__clang__)
#define CLANG_BUILD
#endif
#ifdef CLANG_BUILD
#define ALPHAVAL (-1)
#define ADDVI_H(a, b) __msa_addvi_h((v8i16)a, b)
#define ADDVI_W(a, b) __msa_addvi_w((v4i32)a, b)
#define SRAI_B(a, b) __msa_srai_b((v16i8)a, b)
#define SRAI_H(a, b) __msa_srai_h((v8i16)a, b)
#define SRAI_W(a, b) __msa_srai_w((v4i32)a, b)
#define SRLI_H(a, b) __msa_srli_h((v8i16)a, b)
#define SLLI_B(a, b) __msa_slli_b((v4i32)a, b)
#define ANDI_B(a, b) __msa_andi_b((v16u8)a, b)
#define ORI_B(a, b) __msa_ori_b((v16u8)a, b)
#else
#define ALPHAVAL (0xff)
#define ADDVI_H(a, b) (a + b)
#define ADDVI_W(a, b) (a + b)
#define SRAI_B(a, b) (a >> b)
#define SRAI_H(a, b) (a >> b)
#define SRAI_W(a, b) (a >> b)
#define SRLI_H(a, b) (a << b)
#define SLLI_B(a, b) (a << b)
#define ANDI_B(a, b) (a & b)
#define ORI_B(a, b) (a | b)
#endif
#define LD_B(RTYPE, psrc) *((RTYPE*)(psrc))
#define LD_UB(...) LD_B(v16u8, __VA_ARGS__)
#define LD_SB(...) LD_B(v16i8, __VA_ARGS__)
#define LD_H(RTYPE, psrc) *((RTYPE*)(psrc))
#define LD_UH(...) LD_H(v8u16, __VA_ARGS__)
#define LD_SH(...) LD_H(v8i16, __VA_ARGS__)
#define LD_W(RTYPE, psrc) *((RTYPE*)(psrc))
#define LD_UW(...) LD_W(v4u32, __VA_ARGS__)
#define LD_SW(...) LD_W(v4i32, __VA_ARGS__)
#define ST_B(RTYPE, in, pdst) *((RTYPE*)(pdst)) = in
#define ST_UB(...) ST_B(v16u8, __VA_ARGS__)
#define ST_SB(...) ST_B(v16i8, __VA_ARGS__)
#define ST_H(RTYPE, in, pdst) *((RTYPE*)(pdst)) = in
#define ST_UH(...) ST_H(v8u16, __VA_ARGS__)
#define ST_SH(...) ST_H(v8i16, __VA_ARGS__)
#define ST_W(RTYPE, in, pdst) *((RTYPE*)(pdst)) = in
#define ST_UW(...) ST_W(v4u32, __VA_ARGS__)
#define ST_SW(...) ST_W(v4i32, __VA_ARGS__)
#define MSA_LOAD_FUNC(TYPE, INSTR, FUNC_NAME) \
static inline TYPE FUNC_NAME(const void* const psrc) { \
const uint8_t* const psrc_m = (const uint8_t*)psrc; \
TYPE val_m; \
asm volatile ( \
"" #INSTR " %[val_m], %[psrc_m] \n\t" \
: [val_m] "=r" (val_m) \
: [psrc_m] "m" (*psrc_m)); \
return val_m; \
}
#define MSA_LOAD(psrc, FUNC_NAME) FUNC_NAME(psrc)
#define MSA_STORE_FUNC(TYPE, INSTR, FUNC_NAME) \
static inline void FUNC_NAME(TYPE val, void* const pdst) { \
uint8_t* const pdst_m = (uint8_t*)pdst; \
TYPE val_m = val; \
asm volatile ( \
" " #INSTR " %[val_m], %[pdst_m] \n\t" \
: [pdst_m] "=m" (*pdst_m) \
: [val_m] "r" (val_m)); \
}
#define MSA_STORE(val, pdst, FUNC_NAME) FUNC_NAME(val, pdst)
#if (__mips_isa_rev >= 6)
MSA_LOAD_FUNC(uint16_t, lh, msa_lh);
#define LH(psrc) MSA_LOAD(psrc, msa_lh)
MSA_LOAD_FUNC(uint32_t, lw, msa_lw);
#define LW(psrc) MSA_LOAD(psrc, msa_lw)
#if (__mips == 64)
MSA_LOAD_FUNC(uint64_t, ld, msa_ld);
#define LD(psrc) MSA_LOAD(psrc, msa_ld)
#else // !(__mips == 64)
#define LD(psrc) ((((uint64_t)MSA_LOAD(psrc + 4, msa_lw)) << 32) | \
MSA_LOAD(psrc, msa_lw))
#endif // (__mips == 64)
MSA_STORE_FUNC(uint16_t, sh, msa_sh);
#define SH(val, pdst) MSA_STORE(val, pdst, msa_sh)
MSA_STORE_FUNC(uint32_t, sw, msa_sw);
#define SW(val, pdst) MSA_STORE(val, pdst, msa_sw)
MSA_STORE_FUNC(uint64_t, sd, msa_sd);
#define SD(val, pdst) MSA_STORE(val, pdst, msa_sd)
#else // !(__mips_isa_rev >= 6)
MSA_LOAD_FUNC(uint16_t, ulh, msa_ulh);
#define LH(psrc) MSA_LOAD(psrc, msa_ulh)
MSA_LOAD_FUNC(uint32_t, ulw, msa_ulw);
#define LW(psrc) MSA_LOAD(psrc, msa_ulw)
#if (__mips == 64)
MSA_LOAD_FUNC(uint64_t, uld, msa_uld);
#define LD(psrc) MSA_LOAD(psrc, msa_uld)
#else // !(__mips == 64)
#define LD(psrc) ((((uint64_t)MSA_LOAD(psrc + 4, msa_ulw)) << 32) | \
MSA_LOAD(psrc, msa_ulw))
#endif // (__mips == 64)
MSA_STORE_FUNC(uint16_t, ush, msa_ush);
#define SH(val, pdst) MSA_STORE(val, pdst, msa_ush)
MSA_STORE_FUNC(uint32_t, usw, msa_usw);
#define SW(val, pdst) MSA_STORE(val, pdst, msa_usw)
#define SD(val, pdst) do { \
uint8_t* const pdst_sd_m = (uint8_t*)(pdst); \
const uint32_t val0_m = (uint32_t)(val & 0x00000000FFFFFFFF); \
const uint32_t val1_m = (uint32_t)((val >> 32) & 0x00000000FFFFFFFF); \
SW(val0_m, pdst_sd_m); \
SW(val1_m, pdst_sd_m + 4); \
} while (0)
#endif // (__mips_isa_rev >= 6)
/* Description : Load 4 words with stride
* Arguments : Inputs - psrc, stride
* Outputs - out0, out1, out2, out3
* Details : Load word in 'out0' from (psrc)
* Load word in 'out1' from (psrc + stride)
* Load word in 'out2' from (psrc + 2 * stride)
* Load word in 'out3' from (psrc + 3 * stride)
*/
#define LW4(psrc, stride, out0, out1, out2, out3) do { \
const uint8_t* ptmp = (const uint8_t*)psrc; \
out0 = LW(ptmp); \
ptmp += stride; \
out1 = LW(ptmp); \
ptmp += stride; \
out2 = LW(ptmp); \
ptmp += stride; \
out3 = LW(ptmp); \
} while (0)
/* Description : Store words with stride
* Arguments : Inputs - in0, in1, in2, in3, pdst, stride
* Details : Store word from 'in0' to (pdst)
* Store word from 'in1' to (pdst + stride)
* Store word from 'in2' to (pdst + 2 * stride)
* Store word from 'in3' to (pdst + 3 * stride)
*/
#define SW4(in0, in1, in2, in3, pdst, stride) do { \
uint8_t* ptmp = (uint8_t*)pdst; \
SW(in0, ptmp); \
ptmp += stride; \
SW(in1, ptmp); \
ptmp += stride; \
SW(in2, ptmp); \
ptmp += stride; \
SW(in3, ptmp); \
} while (0)
#define SW3(in0, in1, in2, pdst, stride) do { \
uint8_t* ptmp = (uint8_t*)pdst; \
SW(in0, ptmp); \
ptmp += stride; \
SW(in1, ptmp); \
ptmp += stride; \
SW(in2, ptmp); \
} while (0)
#define SW2(in0, in1, pdst, stride) do { \
uint8_t* ptmp = (uint8_t*)pdst; \
SW(in0, ptmp); \
ptmp += stride; \
SW(in1, ptmp); \
} while (0)
/* Description : Store 4 double words with stride
* Arguments : Inputs - in0, in1, in2, in3, pdst, stride
* Details : Store double word from 'in0' to (pdst)
* Store double word from 'in1' to (pdst + stride)
* Store double word from 'in2' to (pdst + 2 * stride)
* Store double word from 'in3' to (pdst + 3 * stride)
*/
#define SD4(in0, in1, in2, in3, pdst, stride) do { \
uint8_t* ptmp = (uint8_t*)pdst; \
SD(in0, ptmp); \
ptmp += stride; \
SD(in1, ptmp); \
ptmp += stride; \
SD(in2, ptmp); \
ptmp += stride; \
SD(in3, ptmp); \
} while (0)
/* Description : Load vectors with 16 byte elements with stride
* Arguments : Inputs - psrc, stride
* Outputs - out0, out1
* Return Type - as per RTYPE
* Details : Load 16 byte elements in 'out0' from (psrc)
* Load 16 byte elements in 'out1' from (psrc + stride)
*/
#define LD_B2(RTYPE, psrc, stride, out0, out1) do { \
out0 = LD_B(RTYPE, psrc); \
out1 = LD_B(RTYPE, psrc + stride); \
} while (0)
#define LD_UB2(...) LD_B2(v16u8, __VA_ARGS__)
#define LD_SB2(...) LD_B2(v16i8, __VA_ARGS__)
#define LD_B3(RTYPE, psrc, stride, out0, out1, out2) do { \
LD_B2(RTYPE, psrc, stride, out0, out1); \
out2 = LD_B(RTYPE, psrc + 2 * stride); \
} while (0)
#define LD_UB3(...) LD_B3(v16u8, __VA_ARGS__)
#define LD_SB3(...) LD_B3(v16i8, __VA_ARGS__)
#define LD_B4(RTYPE, psrc, stride, out0, out1, out2, out3) do { \
LD_B2(RTYPE, psrc, stride, out0, out1); \
LD_B2(RTYPE, psrc + 2 * stride , stride, out2, out3); \
} while (0)
#define LD_UB4(...) LD_B4(v16u8, __VA_ARGS__)
#define LD_SB4(...) LD_B4(v16i8, __VA_ARGS__)
#define LD_B8(RTYPE, psrc, stride, \
out0, out1, out2, out3, out4, out5, out6, out7) do { \
LD_B4(RTYPE, psrc, stride, out0, out1, out2, out3); \
LD_B4(RTYPE, psrc + 4 * stride, stride, out4, out5, out6, out7); \
} while (0)
#define LD_UB8(...) LD_B8(v16u8, __VA_ARGS__)
#define LD_SB8(...) LD_B8(v16i8, __VA_ARGS__)
/* Description : Load vectors with 8 halfword elements with stride
* Arguments : Inputs - psrc, stride
* Outputs - out0, out1
* Details : Load 8 halfword elements in 'out0' from (psrc)
* Load 8 halfword elements in 'out1' from (psrc + stride)
*/
#define LD_H2(RTYPE, psrc, stride, out0, out1) do { \
out0 = LD_H(RTYPE, psrc); \
out1 = LD_H(RTYPE, psrc + stride); \
} while (0)
#define LD_UH2(...) LD_H2(v8u16, __VA_ARGS__)
#define LD_SH2(...) LD_H2(v8i16, __VA_ARGS__)
/* Description : Load vectors with 4 word elements with stride
* Arguments : Inputs - psrc, stride
* Outputs - out0, out1, out2, out3
* Details : Load 4 word elements in 'out0' from (psrc + 0 * stride)
* Load 4 word elements in 'out1' from (psrc + 1 * stride)
* Load 4 word elements in 'out2' from (psrc + 2 * stride)
* Load 4 word elements in 'out3' from (psrc + 3 * stride)
*/
#define LD_W2(RTYPE, psrc, stride, out0, out1) do { \
out0 = LD_W(RTYPE, psrc); \
out1 = LD_W(RTYPE, psrc + stride); \
} while (0)
#define LD_UW2(...) LD_W2(v4u32, __VA_ARGS__)
#define LD_SW2(...) LD_W2(v4i32, __VA_ARGS__)
#define LD_W3(RTYPE, psrc, stride, out0, out1, out2) do { \
LD_W2(RTYPE, psrc, stride, out0, out1); \
out2 = LD_W(RTYPE, psrc + 2 * stride); \
} while (0)
#define LD_UW3(...) LD_W3(v4u32, __VA_ARGS__)
#define LD_SW3(...) LD_W3(v4i32, __VA_ARGS__)
#define LD_W4(RTYPE, psrc, stride, out0, out1, out2, out3) do { \
LD_W2(RTYPE, psrc, stride, out0, out1); \
LD_W2(RTYPE, psrc + 2 * stride, stride, out2, out3); \
} while (0)
#define LD_UW4(...) LD_W4(v4u32, __VA_ARGS__)
#define LD_SW4(...) LD_W4(v4i32, __VA_ARGS__)
/* Description : Store vectors of 16 byte elements with stride
* Arguments : Inputs - in0, in1, pdst, stride
* Details : Store 16 byte elements from 'in0' to (pdst)
* Store 16 byte elements from 'in1' to (pdst + stride)
*/
#define ST_B2(RTYPE, in0, in1, pdst, stride) do { \
ST_B(RTYPE, in0, pdst); \
ST_B(RTYPE, in1, pdst + stride); \
} while (0)
#define ST_UB2(...) ST_B2(v16u8, __VA_ARGS__)
#define ST_SB2(...) ST_B2(v16i8, __VA_ARGS__)
#define ST_B4(RTYPE, in0, in1, in2, in3, pdst, stride) do { \
ST_B2(RTYPE, in0, in1, pdst, stride); \
ST_B2(RTYPE, in2, in3, pdst + 2 * stride, stride); \
} while (0)
#define ST_UB4(...) ST_B4(v16u8, __VA_ARGS__)
#define ST_SB4(...) ST_B4(v16i8, __VA_ARGS__)
#define ST_B8(RTYPE, in0, in1, in2, in3, in4, in5, in6, in7, \
pdst, stride) do { \
ST_B4(RTYPE, in0, in1, in2, in3, pdst, stride); \
ST_B4(RTYPE, in4, in5, in6, in7, pdst + 4 * stride, stride); \
} while (0)
#define ST_UB8(...) ST_B8(v16u8, __VA_ARGS__)
/* Description : Store vectors of 4 word elements with stride
* Arguments : Inputs - in0, in1, in2, in3, pdst, stride
* Details : Store 4 word elements from 'in0' to (pdst + 0 * stride)
* Store 4 word elements from 'in1' to (pdst + 1 * stride)
* Store 4 word elements from 'in2' to (pdst + 2 * stride)
* Store 4 word elements from 'in3' to (pdst + 3 * stride)
*/
#define ST_W2(RTYPE, in0, in1, pdst, stride) do { \
ST_W(RTYPE, in0, pdst); \
ST_W(RTYPE, in1, pdst + stride); \
} while (0)
#define ST_UW2(...) ST_W2(v4u32, __VA_ARGS__)
#define ST_SW2(...) ST_W2(v4i32, __VA_ARGS__)
#define ST_W3(RTYPE, in0, in1, in2, pdst, stride) do { \
ST_W2(RTYPE, in0, in1, pdst, stride); \
ST_W(RTYPE, in2, pdst + 2 * stride); \
} while (0)
#define ST_UW3(...) ST_W3(v4u32, __VA_ARGS__)
#define ST_SW3(...) ST_W3(v4i32, __VA_ARGS__)
#define ST_W4(RTYPE, in0, in1, in2, in3, pdst, stride) do { \
ST_W2(RTYPE, in0, in1, pdst, stride); \
ST_W2(RTYPE, in2, in3, pdst + 2 * stride, stride); \
} while (0)
#define ST_UW4(...) ST_W4(v4u32, __VA_ARGS__)
#define ST_SW4(...) ST_W4(v4i32, __VA_ARGS__)
/* Description : Store vectors of 8 halfword elements with stride
* Arguments : Inputs - in0, in1, pdst, stride
* Details : Store 8 halfword elements from 'in0' to (pdst)
* Store 8 halfword elements from 'in1' to (pdst + stride)
*/
#define ST_H2(RTYPE, in0, in1, pdst, stride) do { \
ST_H(RTYPE, in0, pdst); \
ST_H(RTYPE, in1, pdst + stride); \
} while (0)
#define ST_UH2(...) ST_H2(v8u16, __VA_ARGS__)
#define ST_SH2(...) ST_H2(v8i16, __VA_ARGS__)
/* Description : Store 2x4 byte block to destination memory from input vector
* Arguments : Inputs - in, stidx, pdst, stride
* Details : Index 'stidx' halfword element from 'in' vector is copied to
* the GP register and stored to (pdst)
* Index 'stidx+1' halfword element from 'in' vector is copied to
* the GP register and stored to (pdst + stride)
* Index 'stidx+2' halfword element from 'in' vector is copied to
* the GP register and stored to (pdst + 2 * stride)
* Index 'stidx+3' halfword element from 'in' vector is copied to
* the GP register and stored to (pdst + 3 * stride)
*/
#define ST2x4_UB(in, stidx, pdst, stride) do { \
uint8_t* pblk_2x4_m = (uint8_t*)pdst; \
const uint16_t out0_m = __msa_copy_s_h((v8i16)in, stidx); \
const uint16_t out1_m = __msa_copy_s_h((v8i16)in, stidx + 1); \
const uint16_t out2_m = __msa_copy_s_h((v8i16)in, stidx + 2); \
const uint16_t out3_m = __msa_copy_s_h((v8i16)in, stidx + 3); \
SH(out0_m, pblk_2x4_m); \
pblk_2x4_m += stride; \
SH(out1_m, pblk_2x4_m); \
pblk_2x4_m += stride; \
SH(out2_m, pblk_2x4_m); \
pblk_2x4_m += stride; \
SH(out3_m, pblk_2x4_m); \
} while (0)
/* Description : Store 4x4 byte block to destination memory from input vector
* Arguments : Inputs - in0, in1, pdst, stride
* Details : 'Idx0' word element from input vector 'in0' is copied to the
* GP register and stored to (pdst)
* 'Idx1' word element from input vector 'in0' is copied to the
* GP register and stored to (pdst + stride)
* 'Idx2' word element from input vector 'in0' is copied to the
* GP register and stored to (pdst + 2 * stride)
* 'Idx3' word element from input vector 'in0' is copied to the
* GP register and stored to (pdst + 3 * stride)
*/
#define ST4x4_UB(in0, in1, idx0, idx1, idx2, idx3, pdst, stride) do { \
uint8_t* const pblk_4x4_m = (uint8_t*)pdst; \
const uint32_t out0_m = __msa_copy_s_w((v4i32)in0, idx0); \
const uint32_t out1_m = __msa_copy_s_w((v4i32)in0, idx1); \
const uint32_t out2_m = __msa_copy_s_w((v4i32)in1, idx2); \
const uint32_t out3_m = __msa_copy_s_w((v4i32)in1, idx3); \
SW4(out0_m, out1_m, out2_m, out3_m, pblk_4x4_m, stride); \
} while (0)
#define ST4x8_UB(in0, in1, pdst, stride) do { \
uint8_t* const pblk_4x8 = (uint8_t*)pdst; \
ST4x4_UB(in0, in0, 0, 1, 2, 3, pblk_4x8, stride); \
ST4x4_UB(in1, in1, 0, 1, 2, 3, pblk_4x8 + 4 * stride, stride); \
} while (0)
/* Description : Immediate number of elements to slide
* Arguments : Inputs - in0, in1, slide_val
* Outputs - out
* Return Type - as per RTYPE
* Details : Byte elements from 'in1' vector are slid into 'in0' by
* value specified in the 'slide_val'
*/
#define SLDI_B(RTYPE, in0, in1, slide_val) \
(RTYPE)__msa_sldi_b((v16i8)in0, (v16i8)in1, slide_val) \
#define SLDI_UB(...) SLDI_B(v16u8, __VA_ARGS__)
#define SLDI_SB(...) SLDI_B(v16i8, __VA_ARGS__)
#define SLDI_SH(...) SLDI_B(v8i16, __VA_ARGS__)
/* Description : Shuffle byte vector elements as per mask vector
* Arguments : Inputs - in0, in1, in2, in3, mask0, mask1
* Outputs - out0, out1
* Return Type - as per RTYPE
* Details : Byte elements from 'in0' & 'in1' are copied selectively to
* 'out0' as per control vector 'mask0'
*/
#define VSHF_B(RTYPE, in0, in1, mask) \
(RTYPE)__msa_vshf_b((v16i8)mask, (v16i8)in1, (v16i8)in0)
#define VSHF_UB(...) VSHF_B(v16u8, __VA_ARGS__)
#define VSHF_SB(...) VSHF_B(v16i8, __VA_ARGS__)
#define VSHF_UH(...) VSHF_B(v8u16, __VA_ARGS__)
#define VSHF_SH(...) VSHF_B(v8i16, __VA_ARGS__)
#define VSHF_B2(RTYPE, in0, in1, in2, in3, mask0, mask1, out0, out1) do { \
out0 = VSHF_B(RTYPE, in0, in1, mask0); \
out1 = VSHF_B(RTYPE, in2, in3, mask1); \
} while (0)
#define VSHF_B2_UB(...) VSHF_B2(v16u8, __VA_ARGS__)
#define VSHF_B2_SB(...) VSHF_B2(v16i8, __VA_ARGS__)
#define VSHF_B2_UH(...) VSHF_B2(v8u16, __VA_ARGS__)
#define VSHF_B2_SH(...) VSHF_B2(v8i16, __VA_ARGS__)
/* Description : Shuffle halfword vector elements as per mask vector
* Arguments : Inputs - in0, in1, in2, in3, mask0, mask1
* Outputs - out0, out1
* Return Type - as per RTYPE
* Details : halfword elements from 'in0' & 'in1' are copied selectively to
* 'out0' as per control vector 'mask0'
*/
#define VSHF_H2(RTYPE, in0, in1, in2, in3, mask0, mask1, out0, out1) do { \
out0 = (RTYPE)__msa_vshf_h((v8i16)mask0, (v8i16)in1, (v8i16)in0); \
out1 = (RTYPE)__msa_vshf_h((v8i16)mask1, (v8i16)in3, (v8i16)in2); \
} while (0)
#define VSHF_H2_UH(...) VSHF_H2(v8u16, __VA_ARGS__)
#define VSHF_H2_SH(...) VSHF_H2(v8i16, __VA_ARGS__)
/* Description : Dot product of byte vector elements
* Arguments : Inputs - mult0, mult1, cnst0, cnst1
* Outputs - out0, out1
* Return Type - as per RTYPE
* Details : Signed byte elements from 'mult0' are multiplied with
* signed byte elements from 'cnst0' producing a result
* twice the size of input i.e. signed halfword.
* The multiplication result of adjacent odd-even elements
* are added together and written to the 'out0' vector
*/
#define DOTP_SB2(RTYPE, mult0, mult1, cnst0, cnst1, out0, out1) do { \
out0 = (RTYPE)__msa_dotp_s_h((v16i8)mult0, (v16i8)cnst0); \
out1 = (RTYPE)__msa_dotp_s_h((v16i8)mult1, (v16i8)cnst1); \
} while (0)
#define DOTP_SB2_SH(...) DOTP_SB2(v8i16, __VA_ARGS__)
/* Description : Dot product of halfword vector elements
* Arguments : Inputs - mult0, mult1, cnst0, cnst1
* Outputs - out0, out1
* Return Type - as per RTYPE
* Details : Signed halfword elements from 'mult0' are multiplied with
* signed halfword elements from 'cnst0' producing a result
* twice the size of input i.e. signed word.
* The multiplication result of adjacent odd-even elements
* are added together and written to the 'out0' vector
*/
#define DOTP_SH2(RTYPE, mult0, mult1, cnst0, cnst1, out0, out1) do { \
out0 = (RTYPE)__msa_dotp_s_w((v8i16)mult0, (v8i16)cnst0); \
out1 = (RTYPE)__msa_dotp_s_w((v8i16)mult1, (v8i16)cnst1); \
} while (0)
#define DOTP_SH2_SW(...) DOTP_SH2(v4i32, __VA_ARGS__)
/* Description : Dot product of unsigned word vector elements
* Arguments : Inputs - mult0, mult1, cnst0, cnst1
* Outputs - out0, out1
* Return Type - as per RTYPE
* Details : Unsigned word elements from 'mult0' are multiplied with
* unsigned word elements from 'cnst0' producing a result
* twice the size of input i.e. unsigned double word.
* The multiplication result of adjacent odd-even elements
* are added together and written to the 'out0' vector
*/
#define DOTP_UW2(RTYPE, mult0, mult1, cnst0, cnst1, out0, out1) do { \
out0 = (RTYPE)__msa_dotp_u_d((v4u32)mult0, (v4u32)cnst0); \
out1 = (RTYPE)__msa_dotp_u_d((v4u32)mult1, (v4u32)cnst1); \
} while (0)
#define DOTP_UW2_UD(...) DOTP_UW2(v2u64, __VA_ARGS__)
/* Description : Dot product & addition of halfword vector elements
* Arguments : Inputs - mult0, mult1, cnst0, cnst1
* Outputs - out0, out1
* Return Type - as per RTYPE
* Details : Signed halfword elements from 'mult0' are multiplied with
* signed halfword elements from 'cnst0' producing a result
* twice the size of input i.e. signed word.
* The multiplication result of adjacent odd-even elements
* are added to the 'out0' vector
*/
#define DPADD_SH2(RTYPE, mult0, mult1, cnst0, cnst1, out0, out1) do { \
out0 = (RTYPE)__msa_dpadd_s_w((v4i32)out0, (v8i16)mult0, (v8i16)cnst0); \
out1 = (RTYPE)__msa_dpadd_s_w((v4i32)out1, (v8i16)mult1, (v8i16)cnst1); \
} while (0)
#define DPADD_SH2_SW(...) DPADD_SH2(v4i32, __VA_ARGS__)
/* Description : Clips all signed halfword elements of input vector
* between 0 & 255
* Arguments : Input/output - val
* Return Type - signed halfword
*/
#define CLIP_SH_0_255(val) do { \
const v8i16 max_m = __msa_ldi_h(255); \
val = __msa_maxi_s_h((v8i16)val, 0); \
val = __msa_min_s_h(max_m, (v8i16)val); \
} while (0)
#define CLIP_SH2_0_255(in0, in1) do { \
CLIP_SH_0_255(in0); \
CLIP_SH_0_255(in1); \
} while (0)
#define CLIP_SH4_0_255(in0, in1, in2, in3) do { \
CLIP_SH2_0_255(in0, in1); \
CLIP_SH2_0_255(in2, in3); \
} while (0)
/* Description : Clips all unsigned halfword elements of input vector
* between 0 & 255
* Arguments : Input - in
* Output - out_m
* Return Type - unsigned halfword
*/
#define CLIP_UH_0_255(in) do { \
const v8u16 max_m = (v8u16)__msa_ldi_h(255); \
in = __msa_maxi_u_h((v8u16) in, 0); \
in = __msa_min_u_h((v8u16) max_m, (v8u16) in); \
} while (0)
#define CLIP_UH2_0_255(in0, in1) do { \
CLIP_UH_0_255(in0); \
CLIP_UH_0_255(in1); \
} while (0)
/* Description : Clips all signed word elements of input vector
* between 0 & 255
* Arguments : Input/output - val
* Return Type - signed word
*/
#define CLIP_SW_0_255(val) do { \
const v4i32 max_m = __msa_ldi_w(255); \
val = __msa_maxi_s_w((v4i32)val, 0); \
val = __msa_min_s_w(max_m, (v4i32)val); \
} while (0)
#define CLIP_SW4_0_255(in0, in1, in2, in3) do { \
CLIP_SW_0_255(in0); \
CLIP_SW_0_255(in1); \
CLIP_SW_0_255(in2); \
CLIP_SW_0_255(in3); \
} while (0)
/* Description : Horizontal addition of 4 signed word elements of input vector
* Arguments : Input - in (signed word vector)
* Output - sum_m (i32 sum)
* Return Type - signed word (GP)
* Details : 4 signed word elements of 'in' vector are added together and
* the resulting integer sum is returned
*/
static WEBP_INLINE int32_t func_hadd_sw_s32(v4i32 in) {
const v2i64 res0_m = __msa_hadd_s_d((v4i32)in, (v4i32)in);
const v2i64 res1_m = __msa_splati_d(res0_m, 1);
const v2i64 out = res0_m + res1_m;
int32_t sum_m = __msa_copy_s_w((v4i32)out, 0);
return sum_m;
}
#define HADD_SW_S32(in) func_hadd_sw_s32(in)
/* Description : Horizontal addition of 8 signed halfword elements
* Arguments : Input - in (signed halfword vector)
* Output - sum_m (s32 sum)
* Return Type - signed word
* Details : 8 signed halfword elements of input vector are added
* together and the resulting integer sum is returned
*/
static WEBP_INLINE int32_t func_hadd_sh_s32(v8i16 in) {
const v4i32 res = __msa_hadd_s_w(in, in);
const v2i64 res0 = __msa_hadd_s_d(res, res);
const v2i64 res1 = __msa_splati_d(res0, 1);
const v2i64 res2 = res0 + res1;
const int32_t sum_m = __msa_copy_s_w((v4i32)res2, 0);
return sum_m;
}
#define HADD_SH_S32(in) func_hadd_sh_s32(in)
/* Description : Horizontal addition of 8 unsigned halfword elements
* Arguments : Input - in (unsigned halfword vector)
* Output - sum_m (u32 sum)
* Return Type - unsigned word
* Details : 8 unsigned halfword elements of input vector are added
* together and the resulting integer sum is returned
*/
static WEBP_INLINE uint32_t func_hadd_uh_u32(v8u16 in) {
uint32_t sum_m;
const v4u32 res_m = __msa_hadd_u_w(in, in);
v2u64 res0_m = __msa_hadd_u_d(res_m, res_m);
v2u64 res1_m = (v2u64)__msa_splati_d((v2i64)res0_m, 1);
res0_m = res0_m + res1_m;
sum_m = __msa_copy_s_w((v4i32)res0_m, 0);
return sum_m;
}
#define HADD_UH_U32(in) func_hadd_uh_u32(in)
/* Description : Horizontal addition of signed half word vector elements
Arguments : Inputs - in0, in1
Outputs - out0, out1
Return Type - as per RTYPE
Details : Each signed odd half word element from 'in0' is added to
even signed half word element from 'in0' (pairwise) and the
halfword result is written in 'out0'
*/
#define HADD_SH2(RTYPE, in0, in1, out0, out1) do { \
out0 = (RTYPE)__msa_hadd_s_w((v8i16)in0, (v8i16)in0); \
out1 = (RTYPE)__msa_hadd_s_w((v8i16)in1, (v8i16)in1); \
} while (0)
#define HADD_SH2_SW(...) HADD_SH2(v4i32, __VA_ARGS__)
#define HADD_SH4(RTYPE, in0, in1, in2, in3, out0, out1, out2, out3) do { \
HADD_SH2(RTYPE, in0, in1, out0, out1); \
HADD_SH2(RTYPE, in2, in3, out2, out3); \
} while (0)
#define HADD_SH4_SW(...) HADD_SH4(v4i32, __VA_ARGS__)
/* Description : Horizontal subtraction of unsigned byte vector elements
* Arguments : Inputs - in0, in1
* Outputs - out0, out1
* Return Type - as per RTYPE
* Details : Each unsigned odd byte element from 'in0' is subtracted from
* even unsigned byte element from 'in0' (pairwise) and the
* halfword result is written to 'out0'
*/
#define HSUB_UB2(RTYPE, in0, in1, out0, out1) do { \
out0 = (RTYPE)__msa_hsub_u_h((v16u8)in0, (v16u8)in0); \
out1 = (RTYPE)__msa_hsub_u_h((v16u8)in1, (v16u8)in1); \
} while (0)
#define HSUB_UB2_UH(...) HSUB_UB2(v8u16, __VA_ARGS__)
#define HSUB_UB2_SH(...) HSUB_UB2(v8i16, __VA_ARGS__)
#define HSUB_UB2_SW(...) HSUB_UB2(v4i32, __VA_ARGS__)
/* Description : Set element n input vector to GPR value
* Arguments : Inputs - in0, in1, in2, in3
* Output - out
* Return Type - as per RTYPE
* Details : Set element 0 in vector 'out' to value specified in 'in0'
*/
#define INSERT_W2(RTYPE, in0, in1, out) do { \
out = (RTYPE)__msa_insert_w((v4i32)out, 0, in0); \
out = (RTYPE)__msa_insert_w((v4i32)out, 1, in1); \
} while (0)
#define INSERT_W2_UB(...) INSERT_W2(v16u8, __VA_ARGS__)
#define INSERT_W2_SB(...) INSERT_W2(v16i8, __VA_ARGS__)
#define INSERT_W4(RTYPE, in0, in1, in2, in3, out) do { \
out = (RTYPE)__msa_insert_w((v4i32)out, 0, in0); \
out = (RTYPE)__msa_insert_w((v4i32)out, 1, in1); \
out = (RTYPE)__msa_insert_w((v4i32)out, 2, in2); \
out = (RTYPE)__msa_insert_w((v4i32)out, 3, in3); \
} while (0)
#define INSERT_W4_UB(...) INSERT_W4(v16u8, __VA_ARGS__)
#define INSERT_W4_SB(...) INSERT_W4(v16i8, __VA_ARGS__)
#define INSERT_W4_SW(...) INSERT_W4(v4i32, __VA_ARGS__)
/* Description : Set element n of double word input vector to GPR value
* Arguments : Inputs - in0, in1
* Output - out
* Return Type - as per RTYPE
* Details : Set element 0 in vector 'out' to GPR value specified in 'in0'
* Set element 1 in vector 'out' to GPR value specified in 'in1'
*/
#define INSERT_D2(RTYPE, in0, in1, out) do { \
out = (RTYPE)__msa_insert_d((v2i64)out, 0, in0); \
out = (RTYPE)__msa_insert_d((v2i64)out, 1, in1); \
} while (0)
#define INSERT_D2_UB(...) INSERT_D2(v16u8, __VA_ARGS__)
#define INSERT_D2_SB(...) INSERT_D2(v16i8, __VA_ARGS__)
/* Description : Interleave even byte elements from vectors
* Arguments : Inputs - in0, in1, in2, in3
* Outputs - out0, out1
* Return Type - as per RTYPE
* Details : Even byte elements of 'in0' and 'in1' are interleaved
* and written to 'out0'
*/
#define ILVEV_B2(RTYPE, in0, in1, in2, in3, out0, out1) do { \
out0 = (RTYPE)__msa_ilvev_b((v16i8)in1, (v16i8)in0); \
out1 = (RTYPE)__msa_ilvev_b((v16i8)in3, (v16i8)in2); \
} while (0)
#define ILVEV_B2_UB(...) ILVEV_B2(v16u8, __VA_ARGS__)
#define ILVEV_B2_SB(...) ILVEV_B2(v16i8, __VA_ARGS__)
#define ILVEV_B2_UH(...) ILVEV_B2(v8u16, __VA_ARGS__)
#define ILVEV_B2_SH(...) ILVEV_B2(v8i16, __VA_ARGS__)
#define ILVEV_B2_SD(...) ILVEV_B2(v2i64, __VA_ARGS__)
/* Description : Interleave odd byte elements from vectors
* Arguments : Inputs - in0, in1, in2, in3
* Outputs - out0, out1
* Return Type - as per RTYPE
* Details : Odd byte elements of 'in0' and 'in1' are interleaved
* and written to 'out0'
*/
#define ILVOD_B2(RTYPE, in0, in1, in2, in3, out0, out1) do { \
out0 = (RTYPE)__msa_ilvod_b((v16i8)in1, (v16i8)in0); \
out1 = (RTYPE)__msa_ilvod_b((v16i8)in3, (v16i8)in2); \
} while (0)
#define ILVOD_B2_UB(...) ILVOD_B2(v16u8, __VA_ARGS__)
#define ILVOD_B2_SB(...) ILVOD_B2(v16i8, __VA_ARGS__)
#define ILVOD_B2_UH(...) ILVOD_B2(v8u16, __VA_ARGS__)
#define ILVOD_B2_SH(...) ILVOD_B2(v8i16, __VA_ARGS__)
#define ILVOD_B2_SD(...) ILVOD_B2(v2i64, __VA_ARGS__)
/* Description : Interleave even halfword elements from vectors
* Arguments : Inputs - in0, in1, in2, in3
* Outputs - out0, out1
* Return Type - as per RTYPE
* Details : Even halfword elements of 'in0' and 'in1' are interleaved
* and written to 'out0'
*/
#define ILVEV_H2(RTYPE, in0, in1, in2, in3, out0, out1) do { \
out0 = (RTYPE)__msa_ilvev_h((v8i16)in1, (v8i16)in0); \
out1 = (RTYPE)__msa_ilvev_h((v8i16)in3, (v8i16)in2); \
} while (0)
#define ILVEV_H2_UB(...) ILVEV_H2(v16u8, __VA_ARGS__)
#define ILVEV_H2_UH(...) ILVEV_H2(v8u16, __VA_ARGS__)
#define ILVEV_H2_SH(...) ILVEV_H2(v8i16, __VA_ARGS__)
#define ILVEV_H2_SW(...) ILVEV_H2(v4i32, __VA_ARGS__)
/* Description : Interleave odd halfword elements from vectors
* Arguments : Inputs - in0, in1, in2, in3
* Outputs - out0, out1
* Return Type - as per RTYPE
* Details : Odd halfword elements of 'in0' and 'in1' are interleaved
* and written to 'out0'
*/
#define ILVOD_H2(RTYPE, in0, in1, in2, in3, out0, out1) do { \
out0 = (RTYPE)__msa_ilvod_h((v8i16)in1, (v8i16)in0); \
out1 = (RTYPE)__msa_ilvod_h((v8i16)in3, (v8i16)in2); \
} while (0)
#define ILVOD_H2_UB(...) ILVOD_H2(v16u8, __VA_ARGS__)
#define ILVOD_H2_UH(...) ILVOD_H2(v8u16, __VA_ARGS__)
#define ILVOD_H2_SH(...) ILVOD_H2(v8i16, __VA_ARGS__)
#define ILVOD_H2_SW(...) ILVOD_H2(v4i32, __VA_ARGS__)
/* Description : Interleave even word elements from vectors
* Arguments : Inputs - in0, in1, in2, in3
* Outputs - out0, out1
* Return Type - as per RTYPE
* Details : Even word elements of 'in0' and 'in1' are interleaved
* and written to 'out0'
*/
#define ILVEV_W2(RTYPE, in0, in1, in2, in3, out0, out1) do { \
out0 = (RTYPE)__msa_ilvev_w((v4i32)in1, (v4i32)in0); \
out1 = (RTYPE)__msa_ilvev_w((v4i32)in3, (v4i32)in2); \
} while (0)
#define ILVEV_W2_UB(...) ILVEV_W2(v16u8, __VA_ARGS__)
#define ILVEV_W2_SB(...) ILVEV_W2(v16i8, __VA_ARGS__)
#define ILVEV_W2_UH(...) ILVEV_W2(v8u16, __VA_ARGS__)
#define ILVEV_W2_SD(...) ILVEV_W2(v2i64, __VA_ARGS__)
/* Description : Interleave even-odd word elements from vectors
* Arguments : Inputs - in0, in1, in2, in3
* Outputs - out0, out1
* Return Type - as per RTYPE
* Details : Even word elements of 'in0' and 'in1' are interleaved
* and written to 'out0'
* Odd word elements of 'in2' and 'in3' are interleaved
* and written to 'out1'
*/
#define ILVEVOD_W2(RTYPE, in0, in1, in2, in3, out0, out1) do { \
out0 = (RTYPE)__msa_ilvev_w((v4i32)in1, (v4i32)in0); \
out1 = (RTYPE)__msa_ilvod_w((v4i32)in3, (v4i32)in2); \
} while (0)
#define ILVEVOD_W2_UB(...) ILVEVOD_W2(v16u8, __VA_ARGS__)
#define ILVEVOD_W2_UH(...) ILVEVOD_W2(v8u16, __VA_ARGS__)
#define ILVEVOD_W2_SH(...) ILVEVOD_W2(v8i16, __VA_ARGS__)
#define ILVEVOD_W2_SW(...) ILVEVOD_W2(v4i32, __VA_ARGS__)
/* Description : Interleave even-odd half-word elements from vectors
* Arguments : Inputs - in0, in1, in2, in3
* Outputs - out0, out1
* Return Type - as per RTYPE
* Details : Even half-word elements of 'in0' and 'in1' are interleaved
* and written to 'out0'
* Odd half-word elements of 'in2' and 'in3' are interleaved
* and written to 'out1'
*/
#define ILVEVOD_H2(RTYPE, in0, in1, in2, in3, out0, out1) do { \
out0 = (RTYPE)__msa_ilvev_h((v8i16)in1, (v8i16)in0); \
out1 = (RTYPE)__msa_ilvod_h((v8i16)in3, (v8i16)in2); \
} while (0)
#define ILVEVOD_H2_UB(...) ILVEVOD_H2(v16u8, __VA_ARGS__)
#define ILVEVOD_H2_UH(...) ILVEVOD_H2(v8u16, __VA_ARGS__)
#define ILVEVOD_H2_SH(...) ILVEVOD_H2(v8i16, __VA_ARGS__)
#define ILVEVOD_H2_SW(...) ILVEVOD_H2(v4i32, __VA_ARGS__)
/* Description : Interleave even double word elements from vectors
* Arguments : Inputs - in0, in1, in2, in3
* Outputs - out0, out1
* Return Type - as per RTYPE
* Details : Even double word elements of 'in0' and 'in1' are interleaved
* and written to 'out0'
*/
#define ILVEV_D2(RTYPE, in0, in1, in2, in3, out0, out1) do { \
out0 = (RTYPE)__msa_ilvev_d((v2i64)in1, (v2i64)in0); \
out1 = (RTYPE)__msa_ilvev_d((v2i64)in3, (v2i64)in2); \
} while (0)
#define ILVEV_D2_UB(...) ILVEV_D2(v16u8, __VA_ARGS__)
#define ILVEV_D2_SB(...) ILVEV_D2(v16i8, __VA_ARGS__)
#define ILVEV_D2_SW(...) ILVEV_D2(v4i32, __VA_ARGS__)
#define ILVEV_D2_SD(...) ILVEV_D2(v2i64, __VA_ARGS__)
/* Description : Interleave left half of byte elements from vectors
* Arguments : Inputs - in0, in1, in2, in3
* Outputs - out0, out1
* Return Type - as per RTYPE
* Details : Left half of byte elements of 'in0' and 'in1' are interleaved
* and written to 'out0'.
*/
#define ILVL_B2(RTYPE, in0, in1, in2, in3, out0, out1) do { \
out0 = (RTYPE)__msa_ilvl_b((v16i8)in0, (v16i8)in1); \
out1 = (RTYPE)__msa_ilvl_b((v16i8)in2, (v16i8)in3); \
} while (0)
#define ILVL_B2_UB(...) ILVL_B2(v16u8, __VA_ARGS__)
#define ILVL_B2_SB(...) ILVL_B2(v16i8, __VA_ARGS__)
#define ILVL_B2_UH(...) ILVL_B2(v8u16, __VA_ARGS__)
#define ILVL_B2_SH(...) ILVL_B2(v8i16, __VA_ARGS__)
#define ILVL_B2_SW(...) ILVL_B2(v4i32, __VA_ARGS__)
/* Description : Interleave right half of byte elements from vectors
* Arguments : Inputs - in0, in1, in2, in3
* Outputs - out0, out1
* Return Type - as per RTYPE
* Details : Right half of byte elements of 'in0' and 'in1' are interleaved
* and written to out0.
*/
#define ILVR_B2(RTYPE, in0, in1, in2, in3, out0, out1) do { \
out0 = (RTYPE)__msa_ilvr_b((v16i8)in0, (v16i8)in1); \
out1 = (RTYPE)__msa_ilvr_b((v16i8)in2, (v16i8)in3); \
} while (0)
#define ILVR_B2_UB(...) ILVR_B2(v16u8, __VA_ARGS__)
#define ILVR_B2_SB(...) ILVR_B2(v16i8, __VA_ARGS__)
#define ILVR_B2_UH(...) ILVR_B2(v8u16, __VA_ARGS__)
#define ILVR_B2_SH(...) ILVR_B2(v8i16, __VA_ARGS__)
#define ILVR_B2_SW(...) ILVR_B2(v4i32, __VA_ARGS__)
#define ILVR_B4(RTYPE, in0, in1, in2, in3, in4, in5, in6, in7, \
out0, out1, out2, out3) do { \
ILVR_B2(RTYPE, in0, in1, in2, in3, out0, out1); \
ILVR_B2(RTYPE, in4, in5, in6, in7, out2, out3); \
} while (0)
#define ILVR_B4_UB(...) ILVR_B4(v16u8, __VA_ARGS__)
#define ILVR_B4_SB(...) ILVR_B4(v16i8, __VA_ARGS__)
#define ILVR_B4_UH(...) ILVR_B4(v8u16, __VA_ARGS__)
#define ILVR_B4_SH(...) ILVR_B4(v8i16, __VA_ARGS__)
#define ILVR_B4_SW(...) ILVR_B4(v4i32, __VA_ARGS__)
/* Description : Interleave right half of halfword elements from vectors
* Arguments : Inputs - in0, in1, in2, in3
* Outputs - out0, out1
* Return Type - as per RTYPE
* Details : Right half of halfword elements of 'in0' and 'in1' are
* interleaved and written to 'out0'.
*/
#define ILVR_H2(RTYPE, in0, in1, in2, in3, out0, out1) do { \
out0 = (RTYPE)__msa_ilvr_h((v8i16)in0, (v8i16)in1); \
out1 = (RTYPE)__msa_ilvr_h((v8i16)in2, (v8i16)in3); \
} while (0)
#define ILVR_H2_UB(...) ILVR_H2(v16u8, __VA_ARGS__)
#define ILVR_H2_SH(...) ILVR_H2(v8i16, __VA_ARGS__)
#define ILVR_H2_SW(...) ILVR_H2(v4i32, __VA_ARGS__)
#define ILVR_H4(RTYPE, in0, in1, in2, in3, in4, in5, in6, in7, \
out0, out1, out2, out3) do { \
ILVR_H2(RTYPE, in0, in1, in2, in3, out0, out1); \
ILVR_H2(RTYPE, in4, in5, in6, in7, out2, out3); \
} while (0)
#define ILVR_H4_UB(...) ILVR_H4(v16u8, __VA_ARGS__)
#define ILVR_H4_SH(...) ILVR_H4(v8i16, __VA_ARGS__)
#define ILVR_H4_SW(...) ILVR_H4(v4i32, __VA_ARGS__)
/* Description : Interleave right half of double word elements from vectors
* Arguments : Inputs - in0, in1, in2, in3
* Outputs - out0, out1
* Return Type - as per RTYPE
* Details : Right half of double word elements of 'in0' and 'in1' are
* interleaved and written to 'out0'.
*/
#define ILVR_D2(RTYPE, in0, in1, in2, in3, out0, out1) do { \
out0 = (RTYPE)__msa_ilvr_d((v2i64)in0, (v2i64)in1); \
out1 = (RTYPE)__msa_ilvr_d((v2i64)in2, (v2i64)in3); \
} while (0)
#define ILVR_D2_UB(...) ILVR_D2(v16u8, __VA_ARGS__)
#define ILVR_D2_SB(...) ILVR_D2(v16i8, __VA_ARGS__)
#define ILVR_D2_SH(...) ILVR_D2(v8i16, __VA_ARGS__)
#define ILVR_D4(RTYPE, in0, in1, in2, in3, in4, in5, in6, in7, \
out0, out1, out2, out3) do { \
ILVR_D2(RTYPE, in0, in1, in2, in3, out0, out1); \
ILVR_D2(RTYPE, in4, in5, in6, in7, out2, out3); \
} while (0)
#define ILVR_D4_SB(...) ILVR_D4(v16i8, __VA_ARGS__)
#define ILVR_D4_UB(...) ILVR_D4(v16u8, __VA_ARGS__)
/* Description : Interleave both left and right half of input vectors
* Arguments : Inputs - in0, in1
* Outputs - out0, out1
* Return Type - as per RTYPE
* Details : Right half of byte elements from 'in0' and 'in1' are
* interleaved and written to 'out0'
*/
#define ILVRL_B2(RTYPE, in0, in1, out0, out1) do { \
out0 = (RTYPE)__msa_ilvr_b((v16i8)in0, (v16i8)in1); \
out1 = (RTYPE)__msa_ilvl_b((v16i8)in0, (v16i8)in1); \
} while (0)
#define ILVRL_B2_UB(...) ILVRL_B2(v16u8, __VA_ARGS__)
#define ILVRL_B2_SB(...) ILVRL_B2(v16i8, __VA_ARGS__)
#define ILVRL_B2_UH(...) ILVRL_B2(v8u16, __VA_ARGS__)
#define ILVRL_B2_SH(...) ILVRL_B2(v8i16, __VA_ARGS__)
#define ILVRL_B2_SW(...) ILVRL_B2(v4i32, __VA_ARGS__)
#define ILVRL_H2(RTYPE, in0, in1, out0, out1) do { \
out0 = (RTYPE)__msa_ilvr_h((v8i16)in0, (v8i16)in1); \
out1 = (RTYPE)__msa_ilvl_h((v8i16)in0, (v8i16)in1); \
} while (0)
#define ILVRL_H2_UB(...) ILVRL_H2(v16u8, __VA_ARGS__)
#define ILVRL_H2_SB(...) ILVRL_H2(v16i8, __VA_ARGS__)
#define ILVRL_H2_SH(...) ILVRL_H2(v8i16, __VA_ARGS__)
#define ILVRL_H2_SW(...) ILVRL_H2(v4i32, __VA_ARGS__)
#define ILVRL_H2_UW(...) ILVRL_H2(v4u32, __VA_ARGS__)
#define ILVRL_W2(RTYPE, in0, in1, out0, out1) do { \
out0 = (RTYPE)__msa_ilvr_w((v4i32)in0, (v4i32)in1); \
out1 = (RTYPE)__msa_ilvl_w((v4i32)in0, (v4i32)in1); \
} while (0)
#define ILVRL_W2_UB(...) ILVRL_W2(v16u8, __VA_ARGS__)
#define ILVRL_W2_SH(...) ILVRL_W2(v8i16, __VA_ARGS__)
#define ILVRL_W2_SW(...) ILVRL_W2(v4i32, __VA_ARGS__)
#define ILVRL_W2_UW(...) ILVRL_W2(v4u32, __VA_ARGS__)
/* Description : Pack even byte elements of vector pairs
* Arguments : Inputs - in0, in1, in2, in3
* Outputs - out0, out1
* Return Type - as per RTYPE
* Details : Even byte elements of 'in0' are copied to the left half of
* 'out0' & even byte elements of 'in1' are copied to the right
* half of 'out0'.
*/
#define PCKEV_B2(RTYPE, in0, in1, in2, in3, out0, out1) do { \
out0 = (RTYPE)__msa_pckev_b((v16i8)in0, (v16i8)in1); \
out1 = (RTYPE)__msa_pckev_b((v16i8)in2, (v16i8)in3); \
} while (0)
#define PCKEV_B2_SB(...) PCKEV_B2(v16i8, __VA_ARGS__)
#define PCKEV_B2_UB(...) PCKEV_B2(v16u8, __VA_ARGS__)
#define PCKEV_B2_SH(...) PCKEV_B2(v8i16, __VA_ARGS__)
#define PCKEV_B2_SW(...) PCKEV_B2(v4i32, __VA_ARGS__)
#define PCKEV_B4(RTYPE, in0, in1, in2, in3, in4, in5, in6, in7, \
out0, out1, out2, out3) do { \
PCKEV_B2(RTYPE, in0, in1, in2, in3, out0, out1); \
PCKEV_B2(RTYPE, in4, in5, in6, in7, out2, out3); \
} while (0)
#define PCKEV_B4_SB(...) PCKEV_B4(v16i8, __VA_ARGS__)
#define PCKEV_B4_UB(...) PCKEV_B4(v16u8, __VA_ARGS__)
#define PCKEV_B4_SH(...) PCKEV_B4(v8i16, __VA_ARGS__)
#define PCKEV_B4_SW(...) PCKEV_B4(v4i32, __VA_ARGS__)
/* Description : Pack even halfword elements of vector pairs
* Arguments : Inputs - in0, in1, in2, in3
* Outputs - out0, out1
* Return Type - as per RTYPE
* Details : Even halfword elements of 'in0' are copied to the left half of
* 'out0' & even halfword elements of 'in1' are copied to the
* right half of 'out0'.