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
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
|
/*
* Copyright © 2009 Nokia Corporation
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice (including the next
* paragraph) shall be included in all copies or substantial portions of the
* Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
* DEALINGS IN THE SOFTWARE.
*
* Author: Siarhei Siamashka (siarhei.siamashka@nokia.com)
*/
/*
* This file contains implementations of NEON optimized pixel processing
* functions. There is no full and detailed tutorial, but some functions
* (those which are exposing some new or interesting features) are
* extensively commented and can be used as examples.
*
* You may want to have a look at the comments for following functions:
* - pixman_composite_over_8888_0565_asm_neon
* - pixman_composite_over_n_8_0565_asm_neon
*/
/* Prevent the stack from becoming executable for no reason... */
#if defined(__linux__) && defined(__ELF__)
.section .note.GNU-stack,"",%progbits
#endif
.text
.fpu neon
.arch armv7a
.altmacro
#include "pixman-arm-neon-asm.h"
/* Global configuration options and preferences */
/*
* The code can optionally make use of unaligned memory accesses to improve
* performance of handling leading/trailing pixels for each scanline.
* Configuration variable RESPECT_STRICT_ALIGNMENT can be set to 0 for
* example in linux if unaligned memory accesses are not configured to
* generate.exceptions.
*/
.set RESPECT_STRICT_ALIGNMENT, 1
/*
* Set default prefetch type. There is a choice between the following options:
*
* PREFETCH_TYPE_NONE (may be useful for the ARM cores where PLD is set to work
* as NOP to workaround some HW bugs or for whatever other reason)
*
* PREFETCH_TYPE_SIMPLE (may be useful for simple single-issue ARM cores where
* advanced prefetch intruduces heavy overhead)
*
* PREFETCH_TYPE_ADVANCED (useful for superscalar cores such as ARM Cortex-A8
* which can run ARM and NEON instructions simultaneously so that extra ARM
* instructions do not add (many) extra cycles, but improve prefetch efficiency)
*
* Note: some types of function can't support advanced prefetch and fallback
* to simple one (those which handle 24bpp pixels)
*/
.set PREFETCH_TYPE_DEFAULT, PREFETCH_TYPE_ADVANCED
/* Prefetch distance in pixels for simple prefetch */
.set PREFETCH_DISTANCE_SIMPLE, 64
/*
* Implementation of pixman_composite_over_8888_0565_asm_neon
*
* This function takes a8r8g8b8 source buffer, r5g6b5 destination buffer and
* performs OVER compositing operation. Function fast_composite_over_8888_0565
* from pixman-fast-path.c does the same in C and can be used as a reference.
*
* First we need to have some NEON assembly code which can do the actual
* operation on the pixels and provide it to the template macro.
*
* Template macro quite conveniently takes care of emitting all the necessary
* code for memory reading and writing (including quite tricky cases of
* handling unaligned leading/trailing pixels), so we only need to deal with
* the data in NEON registers.
*
* NEON registers allocation in general is recommented to be the following:
* d0, d1, d2, d3 - contain loaded source pixel data
* d4, d5, d6, d7 - contain loaded destination pixels (if they are needed)
* d24, d25, d26, d27 - contain loading mask pixel data (if mask is used)
* d28, d29, d30, d31 - place for storing the result (destination pixels)
*
* As can be seen above, four 64-bit NEON registers are used for keeping
* intermediate pixel data and up to 8 pixels can be processed in one step
* for 32bpp formats (16 pixels for 16bpp, 32 pixels for 8bpp).
*
* This particular function uses the following registers allocation:
* d0, d1, d2, d3 - contain loaded source pixel data
* d4, d5 - contain loaded destination pixels (they are needed)
* d28, d29 - place for storing the result (destination pixels)
*/
/*
* Step one. We need to have some code to do some arithmetics on pixel data.
* This is implemented as a pair of macros: '*_head' and '*_tail'. When used
* back-to-back, they take pixel data from {d0, d1, d2, d3} and {d4, d5},
* perform all the needed calculations and write the result to {d28, d29}.
* The rationale for having two macros and not just one will be explained
* later. In practice, any single monolitic function which does the work can
* be split into two parts in any arbitrary way without affecting correctness.
*
* There is one special trick here too. Common template macro can optionally
* make our life a bit easier by doing R, G, B, A color components
* deinterleaving for 32bpp pixel formats (and this feature is used in
* 'pixman_composite_over_8888_0565_asm_neon' function). So it means that
* instead of having 8 packed pixels in {d0, d1, d2, d3} registers, we
* actually use d0 register for blue channel (a vector of eight 8-bit
* values), d1 register for green, d2 for red and d3 for alpha. This
* simple conversion can be also done with a few NEON instructions:
*
* Packed to planar conversion:
* vuzp.8 d0, d1
* vuzp.8 d2, d3
* vuzp.8 d1, d3
* vuzp.8 d0, d2
*
* Planar to packed conversion:
* vzip.8 d0, d2
* vzip.8 d1, d3
* vzip.8 d2, d3
* vzip.8 d0, d1
*
* But pixel can be loaded directly in planar format using VLD4.8 NEON
* instruction. It is 1 cycle slower than VLD1.32, so this is not always
* desirable, that's why deinterleaving is optional.
*
* But anyway, here is the code:
*/
.macro pixman_composite_over_8888_0565_process_pixblock_head
/* convert 8 r5g6b5 pixel data from {d4, d5} to planar 8-bit format
and put data into d6 - red, d7 - green, d30 - blue */
vshrn.u16 d6, q2, #8
vshrn.u16 d7, q2, #3
vsli.u16 q2, q2, #5
vsri.u8 d6, d6, #5
vmvn.8 d3, d3 /* invert source alpha */
vsri.u8 d7, d7, #6
vshrn.u16 d30, q2, #2
/* now do alpha blending, storing results in 8-bit planar format
into d16 - red, d19 - green, d18 - blue */
vmull.u8 q10, d3, d6
vmull.u8 q11, d3, d7
vmull.u8 q12, d3, d30
vrshr.u16 q13, q10, #8
vrshr.u16 q3, q11, #8
vrshr.u16 q15, q12, #8
vraddhn.u16 d20, q10, q13
vraddhn.u16 d23, q11, q3
vraddhn.u16 d22, q12, q15
.endm
.macro pixman_composite_over_8888_0565_process_pixblock_tail
/* ... continue alpha blending */
vqadd.u8 d16, d2, d20
vqadd.u8 q9, q0, q11
/* convert the result to r5g6b5 and store it into {d28, d29} */
vshll.u8 q14, d16, #8
vshll.u8 q8, d19, #8
vshll.u8 q9, d18, #8
vsri.u16 q14, q8, #5
vsri.u16 q14, q9, #11
.endm
/*
* OK, now we got almost everything that we need. Using the above two
* macros, the work can be done right. But now we want to optimize
* it a bit. ARM Cortex-A8 is an in-order core, and benefits really
* a lot from good code scheduling and software pipelining.
*
* Let's construct some code, which will run in the core main loop.
* Some pseudo-code of the main loop will look like this:
* head
* while (...) {
* tail
* head
* }
* tail
*
* It may look a bit weird, but this setup allows to hide instruction
* latencies better and also utilize dual-issue capability more
* efficiently (make pairs of load-store and ALU instructions).
*
* So what we need now is a '*_tail_head' macro, which will be used
* in the core main loop. A trivial straightforward implementation
* of this macro would look like this:
*
* pixman_composite_over_8888_0565_process_pixblock_tail
* vst1.16 {d28, d29}, [DST_W, :128]!
* vld1.16 {d4, d5}, [DST_R, :128]!
* vld4.32 {d0, d1, d2, d3}, [SRC]!
* pixman_composite_over_8888_0565_process_pixblock_head
* cache_preload 8, 8
*
* Now it also got some VLD/VST instructions. We simply can't move from
* processing one block of pixels to the other one with just arithmetics.
* The previously processed data needs to be written to memory and new
* data needs to be fetched. Fortunately, this main loop does not deal
* with partial leading/trailing pixels and can load/store a full block
* of pixels in a bulk. Additionally, destination buffer is already
* 16 bytes aligned here (which is good for performance).
*
* New things here are DST_R, DST_W, SRC and MASK identifiers. These
* are the aliases for ARM registers which are used as pointers for
* accessing data. We maintain separate pointers for reading and writing
* destination buffer (DST_R and DST_W).
*
* Another new thing is 'cache_preload' macro. It is used for prefetching
* data into CPU L2 cache and improve performance when dealing with large
* images which are far larger than cache size. It uses one argument
* (actually two, but they need to be the same here) - number of pixels
* in a block. Looking into 'pixman-arm-neon-asm.h' can provide some
* details about this macro. Moreover, if good performance is needed
* the code from this macro needs to be copied into '*_tail_head' macro
* and mixed with the rest of code for optimal instructions scheduling.
* We are actually doing it below.
*
* Now after all the explanations, here is the optimized code.
* Different instruction streams (originaling from '*_head', '*_tail'
* and 'cache_preload' macro) use different indentation levels for
* better readability. Actually taking the code from one of these
* indentation levels and ignoring a few VLD/VST instructions would
* result in exactly the code from '*_head', '*_tail' or 'cache_preload'
* macro!
*/
#if 1
.macro pixman_composite_over_8888_0565_process_pixblock_tail_head
vqadd.u8 d16, d2, d20
vld1.16 {d4, d5}, [DST_R, :128]!
vqadd.u8 q9, q0, q11
vshrn.u16 d6, q2, #8
vld4.8 {d0, d1, d2, d3}, [SRC]!
vshrn.u16 d7, q2, #3
vsli.u16 q2, q2, #5
vshll.u8 q14, d16, #8
PF add PF_X, PF_X, #8
vshll.u8 q8, d19, #8
PF tst PF_CTL, #0xF
vsri.u8 d6, d6, #5
PF addne PF_X, PF_X, #8
vmvn.8 d3, d3
PF subne PF_CTL, PF_CTL, #1
vsri.u8 d7, d7, #6
vshrn.u16 d30, q2, #2
vmull.u8 q10, d3, d6
PF pld, [PF_SRC, PF_X, lsl #src_bpp_shift]
vmull.u8 q11, d3, d7
vmull.u8 q12, d3, d30
PF pld, [PF_DST, PF_X, lsl #dst_bpp_shift]
vsri.u16 q14, q8, #5
PF cmp PF_X, ORIG_W
vshll.u8 q9, d18, #8
vrshr.u16 q13, q10, #8
PF subge PF_X, PF_X, ORIG_W
vrshr.u16 q3, q11, #8
vrshr.u16 q15, q12, #8
PF subges PF_CTL, PF_CTL, #0x10
vsri.u16 q14, q9, #11
PF ldrgeb DUMMY, [PF_SRC, SRC_STRIDE, lsl #src_bpp_shift]!
vraddhn.u16 d20, q10, q13
vraddhn.u16 d23, q11, q3
PF ldrgeb DUMMY, [PF_DST, DST_STRIDE, lsl #dst_bpp_shift]!
vraddhn.u16 d22, q12, q15
vst1.16 {d28, d29}, [DST_W, :128]!
.endm
#else
/* If we did not care much about the performance, we would just use this... */
.macro pixman_composite_over_8888_0565_process_pixblock_tail_head
pixman_composite_over_8888_0565_process_pixblock_tail
vst1.16 {d28, d29}, [DST_W, :128]!
vld1.16 {d4, d5}, [DST_R, :128]!
vld4.32 {d0, d1, d2, d3}, [SRC]!
pixman_composite_over_8888_0565_process_pixblock_head
cache_preload 8, 8
.endm
#endif
/*
* And now the final part. We are using 'generate_composite_function' macro
* to put all the stuff together. We are specifying the name of the function
* which we want to get, number of bits per pixel for the source, mask and
* destination (0 if unused, like mask in this case). Next come some bit
* flags:
* FLAG_DST_READWRITE - tells that the destination buffer is both read
* and written, for write-only buffer we would use
* FLAG_DST_WRITEONLY flag instead
* FLAG_DEINTERLEAVE_32BPP - tells that we prefer to work with planar data
* and separate color channels for 32bpp format.
* The next things are:
* - the number of pixels processed per iteration (8 in this case, because
* that's the maximum what can fit into four 64-bit NEON registers).
* - prefetch distance, measured in pixel blocks. In this case it is 5 times
* by 8 pixels. That would be 40 pixels, or up to 160 bytes. Optimal
* prefetch distance can be selected by running some benchmarks.
*
* After that we specify some macros, these are 'default_init',
* 'default_cleanup' here which are empty (but it is possible to have custom
* init/cleanup macros to be able to save/restore some extra NEON registers
* like d8-d15 or do anything else) followed by
* 'pixman_composite_over_8888_0565_process_pixblock_head',
* 'pixman_composite_over_8888_0565_process_pixblock_tail' and
* 'pixman_composite_over_8888_0565_process_pixblock_tail_head'
* which we got implemented above.
*
* The last part is the NEON registers allocation scheme.
*/
generate_composite_function \
pixman_composite_over_8888_0565_asm_neon, 32, 0, 16, \
FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
8, /* number of pixels, processed in a single block */ \
5, /* prefetch distance */ \
default_init, \
default_cleanup, \
pixman_composite_over_8888_0565_process_pixblock_head, \
pixman_composite_over_8888_0565_process_pixblock_tail, \
pixman_composite_over_8888_0565_process_pixblock_tail_head, \
28, /* dst_w_basereg */ \
4, /* dst_r_basereg */ \
0, /* src_basereg */ \
24 /* mask_basereg */
/******************************************************************************/
.macro pixman_composite_over_n_0565_process_pixblock_head
/* convert 8 r5g6b5 pixel data from {d4, d5} to planar 8-bit format
and put data into d6 - red, d7 - green, d30 - blue */
vshrn.u16 d6, q2, #8
vshrn.u16 d7, q2, #3
vsli.u16 q2, q2, #5
vsri.u8 d6, d6, #5
vsri.u8 d7, d7, #6
vshrn.u16 d30, q2, #2
/* now do alpha blending, storing results in 8-bit planar format
into d16 - red, d19 - green, d18 - blue */
vmull.u8 q10, d3, d6
vmull.u8 q11, d3, d7
vmull.u8 q12, d3, d30
vrshr.u16 q13, q10, #8
vrshr.u16 q3, q11, #8
vrshr.u16 q15, q12, #8
vraddhn.u16 d20, q10, q13
vraddhn.u16 d23, q11, q3
vraddhn.u16 d22, q12, q15
.endm
.macro pixman_composite_over_n_0565_process_pixblock_tail
/* ... continue alpha blending */
vqadd.u8 d16, d2, d20
vqadd.u8 q9, q0, q11
/* convert the result to r5g6b5 and store it into {d28, d29} */
vshll.u8 q14, d16, #8
vshll.u8 q8, d19, #8
vshll.u8 q9, d18, #8
vsri.u16 q14, q8, #5
vsri.u16 q14, q9, #11
.endm
/* TODO: expand macros and do better instructions scheduling */
.macro pixman_composite_over_n_0565_process_pixblock_tail_head
pixman_composite_over_n_0565_process_pixblock_tail
vld1.16 {d4, d5}, [DST_R, :128]!
vst1.16 {d28, d29}, [DST_W, :128]!
pixman_composite_over_n_0565_process_pixblock_head
.endm
.macro pixman_composite_over_n_0565_init
add DUMMY, sp, #ARGS_STACK_OFFSET
vld1.32 {d3[0]}, [DUMMY]
vdup.8 d0, d3[0]
vdup.8 d1, d3[1]
vdup.8 d2, d3[2]
vdup.8 d3, d3[3]
vmvn.8 d3, d3 /* invert source alpha */
.endm
generate_composite_function \
pixman_composite_over_n_0565_asm_neon, 0, 0, 16, \
FLAG_DST_READWRITE, \
8, /* number of pixels, processed in a single block */ \
5, /* prefetch distance */ \
pixman_composite_over_n_0565_init, \
default_cleanup, \
pixman_composite_over_n_0565_process_pixblock_head, \
pixman_composite_over_n_0565_process_pixblock_tail, \
pixman_composite_over_n_0565_process_pixblock_tail_head, \
28, /* dst_w_basereg */ \
4, /* dst_r_basereg */ \
0, /* src_basereg */ \
24 /* mask_basereg */
/******************************************************************************/
.macro pixman_composite_src_8888_0565_process_pixblock_head
vshll.u8 q8, d1, #8
vshll.u8 q14, d2, #8
vshll.u8 q9, d0, #8
.endm
.macro pixman_composite_src_8888_0565_process_pixblock_tail
vsri.u16 q14, q8, #5
vsri.u16 q14, q9, #11
.endm
.macro pixman_composite_src_8888_0565_process_pixblock_tail_head
vsri.u16 q14, q8, #5
PF add PF_X, PF_X, #8
PF tst PF_CTL, #0xF
vld4.8 {d0, d1, d2, d3}, [SRC]!
PF addne PF_X, PF_X, #8
PF subne PF_CTL, PF_CTL, #1
vsri.u16 q14, q9, #11
PF cmp PF_X, ORIG_W
PF pld, [PF_SRC, PF_X, lsl #src_bpp_shift]
vshll.u8 q8, d1, #8
vst1.16 {d28, d29}, [DST_W, :128]!
PF subge PF_X, PF_X, ORIG_W
PF subges PF_CTL, PF_CTL, #0x10
vshll.u8 q14, d2, #8
PF ldrgeb DUMMY, [PF_SRC, SRC_STRIDE, lsl #src_bpp_shift]!
vshll.u8 q9, d0, #8
.endm
generate_composite_function \
pixman_composite_src_8888_0565_asm_neon, 32, 0, 16, \
FLAG_DST_WRITEONLY | FLAG_DEINTERLEAVE_32BPP, \
8, /* number of pixels, processed in a single block */ \
10, /* prefetch distance */ \
default_init, \
default_cleanup, \
pixman_composite_src_8888_0565_process_pixblock_head, \
pixman_composite_src_8888_0565_process_pixblock_tail, \
pixman_composite_src_8888_0565_process_pixblock_tail_head
/******************************************************************************/
.macro pixman_composite_src_0565_8888_process_pixblock_head
vshrn.u16 d30, q0, #8
vshrn.u16 d29, q0, #3
vsli.u16 q0, q0, #5
vmov.u8 d31, #255
vsri.u8 d30, d30, #5
vsri.u8 d29, d29, #6
vshrn.u16 d28, q0, #2
.endm
.macro pixman_composite_src_0565_8888_process_pixblock_tail
.endm
/* TODO: expand macros and do better instructions scheduling */
.macro pixman_composite_src_0565_8888_process_pixblock_tail_head
pixman_composite_src_0565_8888_process_pixblock_tail
vst4.8 {d28, d29, d30, d31}, [DST_W, :128]!
vld1.16 {d0, d1}, [SRC]!
pixman_composite_src_0565_8888_process_pixblock_head
cache_preload 8, 8
.endm
generate_composite_function \
pixman_composite_src_0565_8888_asm_neon, 16, 0, 32, \
FLAG_DST_WRITEONLY | FLAG_DEINTERLEAVE_32BPP, \
8, /* number of pixels, processed in a single block */ \
10, /* prefetch distance */ \
default_init, \
default_cleanup, \
pixman_composite_src_0565_8888_process_pixblock_head, \
pixman_composite_src_0565_8888_process_pixblock_tail, \
pixman_composite_src_0565_8888_process_pixblock_tail_head
/******************************************************************************/
.macro pixman_composite_add_8000_8000_process_pixblock_head
vqadd.u8 q14, q0, q2
vqadd.u8 q15, q1, q3
.endm
.macro pixman_composite_add_8000_8000_process_pixblock_tail
.endm
.macro pixman_composite_add_8000_8000_process_pixblock_tail_head
vld1.8 {d0, d1, d2, d3}, [SRC]!
PF add PF_X, PF_X, #32
PF tst PF_CTL, #0xF
vld1.8 {d4, d5, d6, d7}, [DST_R, :128]!
PF addne PF_X, PF_X, #32
PF subne PF_CTL, PF_CTL, #1
vst1.8 {d28, d29, d30, d31}, [DST_W, :128]!
PF cmp PF_X, ORIG_W
PF pld, [PF_SRC, PF_X, lsl #src_bpp_shift]
PF pld, [PF_DST, PF_X, lsl #dst_bpp_shift]
PF subge PF_X, PF_X, ORIG_W
PF subges PF_CTL, PF_CTL, #0x10
vqadd.u8 q14, q0, q2
PF ldrgeb DUMMY, [PF_SRC, SRC_STRIDE, lsl #src_bpp_shift]!
PF ldrgeb DUMMY, [PF_DST, DST_STRIDE, lsl #dst_bpp_shift]!
vqadd.u8 q15, q1, q3
.endm
generate_composite_function \
pixman_composite_add_8000_8000_asm_neon, 8, 0, 8, \
FLAG_DST_READWRITE, \
32, /* number of pixels, processed in a single block */ \
10, /* prefetch distance */ \
default_init, \
default_cleanup, \
pixman_composite_add_8000_8000_process_pixblock_head, \
pixman_composite_add_8000_8000_process_pixblock_tail, \
pixman_composite_add_8000_8000_process_pixblock_tail_head
/******************************************************************************/
.macro pixman_composite_add_8888_8888_process_pixblock_tail_head
vld1.8 {d0, d1, d2, d3}, [SRC]!
PF add PF_X, PF_X, #8
PF tst PF_CTL, #0xF
vld1.8 {d4, d5, d6, d7}, [DST_R, :128]!
PF addne PF_X, PF_X, #8
PF subne PF_CTL, PF_CTL, #1
vst1.8 {d28, d29, d30, d31}, [DST_W, :128]!
PF cmp PF_X, ORIG_W
PF pld, [PF_SRC, PF_X, lsl #src_bpp_shift]
PF pld, [PF_DST, PF_X, lsl #dst_bpp_shift]
PF subge PF_X, PF_X, ORIG_W
PF subges PF_CTL, PF_CTL, #0x10
vqadd.u8 q14, q0, q2
PF ldrgeb DUMMY, [PF_SRC, SRC_STRIDE, lsl #src_bpp_shift]!
PF ldrgeb DUMMY, [PF_DST, DST_STRIDE, lsl #dst_bpp_shift]!
vqadd.u8 q15, q1, q3
.endm
generate_composite_function \
pixman_composite_add_8888_8888_asm_neon, 32, 0, 32, \
FLAG_DST_READWRITE, \
8, /* number of pixels, processed in a single block */ \
10, /* prefetch distance */ \
default_init, \
default_cleanup, \
pixman_composite_add_8000_8000_process_pixblock_head, \
pixman_composite_add_8000_8000_process_pixblock_tail, \
pixman_composite_add_8888_8888_process_pixblock_tail_head
generate_composite_function_single_scanline \
pixman_composite_scanline_add_asm_neon, 32, 0, 32, \
FLAG_DST_READWRITE, \
8, /* number of pixels, processed in a single block */ \
default_init, \
default_cleanup, \
pixman_composite_add_8000_8000_process_pixblock_head, \
pixman_composite_add_8000_8000_process_pixblock_tail, \
pixman_composite_add_8888_8888_process_pixblock_tail_head
/******************************************************************************/
.macro pixman_composite_over_8888_8888_process_pixblock_head
vmvn.8 d24, d3 /* get inverted alpha */
/* do alpha blending */
vmull.u8 q8, d24, d4
vmull.u8 q9, d24, d5
vmull.u8 q10, d24, d6
vmull.u8 q11, d24, d7
.endm
.macro pixman_composite_over_8888_8888_process_pixblock_tail
vrshr.u16 q14, q8, #8
vrshr.u16 q15, q9, #8
vrshr.u16 q12, q10, #8
vrshr.u16 q13, q11, #8
vraddhn.u16 d28, q14, q8
vraddhn.u16 d29, q15, q9
vraddhn.u16 d30, q12, q10
vraddhn.u16 d31, q13, q11
vqadd.u8 q14, q0, q14
vqadd.u8 q15, q1, q15
.endm
.macro pixman_composite_over_8888_8888_process_pixblock_tail_head
vld4.8 {d4, d5, d6, d7}, [DST_R, :128]!
vrshr.u16 q14, q8, #8
PF add PF_X, PF_X, #8
PF tst PF_CTL, #0xF
vrshr.u16 q15, q9, #8
vrshr.u16 q12, q10, #8
vrshr.u16 q13, q11, #8
PF addne PF_X, PF_X, #8
PF subne PF_CTL, PF_CTL, #1
vraddhn.u16 d28, q14, q8
vraddhn.u16 d29, q15, q9
PF cmp PF_X, ORIG_W
vraddhn.u16 d30, q12, q10
vraddhn.u16 d31, q13, q11
vqadd.u8 q14, q0, q14
vqadd.u8 q15, q1, q15
vld4.8 {d0, d1, d2, d3}, [SRC]!
PF pld, [PF_SRC, PF_X, lsl #src_bpp_shift]
vmvn.8 d22, d3
PF pld, [PF_DST, PF_X, lsl #dst_bpp_shift]
vst4.8 {d28, d29, d30, d31}, [DST_W, :128]!
PF subge PF_X, PF_X, ORIG_W
vmull.u8 q8, d22, d4
PF subges PF_CTL, PF_CTL, #0x10
vmull.u8 q9, d22, d5
PF ldrgeb DUMMY, [PF_SRC, SRC_STRIDE, lsl #src_bpp_shift]!
vmull.u8 q10, d22, d6
PF ldrgeb DUMMY, [PF_DST, DST_STRIDE, lsl #dst_bpp_shift]!
vmull.u8 q11, d22, d7
.endm
generate_composite_function \
pixman_composite_over_8888_8888_asm_neon, 32, 0, 32, \
FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
8, /* number of pixels, processed in a single block */ \
5, /* prefetch distance */ \
default_init, \
default_cleanup, \
pixman_composite_over_8888_8888_process_pixblock_head, \
pixman_composite_over_8888_8888_process_pixblock_tail, \
pixman_composite_over_8888_8888_process_pixblock_tail_head
generate_composite_function_single_scanline \
pixman_composite_scanline_over_asm_neon, 32, 0, 32, \
FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
8, /* number of pixels, processed in a single block */ \
default_init, \
default_cleanup, \
pixman_composite_over_8888_8888_process_pixblock_head, \
pixman_composite_over_8888_8888_process_pixblock_tail, \
pixman_composite_over_8888_8888_process_pixblock_tail_head
/******************************************************************************/
/* TODO: expand macros and do better instructions scheduling */
.macro pixman_composite_over_n_8888_process_pixblock_tail_head
pixman_composite_over_8888_8888_process_pixblock_tail
vld4.8 {d4, d5, d6, d7}, [DST_R, :128]!
vst4.8 {d28, d29, d30, d31}, [DST_W, :128]!
pixman_composite_over_8888_8888_process_pixblock_head
.endm
.macro pixman_composite_over_n_8888_init
add DUMMY, sp, #ARGS_STACK_OFFSET
vld1.32 {d3[0]}, [DUMMY]
vdup.8 d0, d3[0]
vdup.8 d1, d3[1]
vdup.8 d2, d3[2]
vdup.8 d3, d3[3]
.endm
generate_composite_function \
pixman_composite_over_n_8888_asm_neon, 0, 0, 32, \
FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
8, /* number of pixels, processed in a single block */ \
5, /* prefetch distance */ \
pixman_composite_over_n_8888_init, \
default_cleanup, \
pixman_composite_over_8888_8888_process_pixblock_head, \
pixman_composite_over_8888_8888_process_pixblock_tail, \
pixman_composite_over_n_8888_process_pixblock_tail_head
/******************************************************************************/
.macro pixman_composite_over_n_8_0565_process_pixblock_head
/* in */
vmull.u8 q0, d24, d8
vmull.u8 q1, d24, d9
vmull.u8 q6, d24, d10
vmull.u8 q7, d24, d11
vrshr.u16 q10, q0, #8
vrshr.u16 q11, q1, #8
vrshr.u16 q12, q6, #8
vrshr.u16 q13, q7, #8
vraddhn.u16 d0, q0, q10
vraddhn.u16 d1, q1, q11
vraddhn.u16 d2, q6, q12
vraddhn.u16 d3, q7, q13
vshrn.u16 d6, q2, #8
vshrn.u16 d7, q2, #3
vsli.u16 q2, q2, #5
vsri.u8 d6, d6, #5
vmvn.8 d3, d3
vsri.u8 d7, d7, #6
vshrn.u16 d30, q2, #2
/* now do alpha blending */
vmull.u8 q10, d3, d6
vmull.u8 q11, d3, d7
vmull.u8 q12, d3, d30
vrshr.u16 q13, q10, #8
vrshr.u16 q3, q11, #8
vrshr.u16 q15, q12, #8
vraddhn.u16 d20, q10, q13
vraddhn.u16 d23, q11, q3
vraddhn.u16 d22, q12, q15
.endm
.macro pixman_composite_over_n_8_0565_process_pixblock_tail
vqadd.u8 d16, d2, d20
vqadd.u8 q9, q0, q11
/* convert to r5g6b5 */
vshll.u8 q14, d16, #8
vshll.u8 q8, d19, #8
vshll.u8 q9, d18, #8
vsri.u16 q14, q8, #5
vsri.u16 q14, q9, #11
.endm
/* TODO: expand macros and do better instructions scheduling */
.macro pixman_composite_over_n_8_0565_process_pixblock_tail_head
pixman_composite_over_n_8_0565_process_pixblock_tail
vst1.16 {d28, d29}, [DST_W, :128]!
vld1.16 {d4, d5}, [DST_R, :128]!
vld1.8 {d24}, [MASK]!
cache_preload 8, 8
pixman_composite_over_n_8_0565_process_pixblock_head
.endm
/*
* This function needs a special initialization of solid mask.
* Solid source pixel data is fetched from stack at ARGS_STACK_OFFSET
* offset, split into color components and replicated in d8-d11
* registers. Additionally, this function needs all the NEON registers,
* so it has to save d8-d15 registers which are callee saved according
* to ABI. These registers are restored from 'cleanup' macro. All the
* other NEON registers are caller saved, so can be clobbered freely
* without introducing any problems.
*/
.macro pixman_composite_over_n_8_0565_init
add DUMMY, sp, #ARGS_STACK_OFFSET
vpush {d8-d15}
vld1.32 {d11[0]}, [DUMMY]
vdup.8 d8, d11[0]
vdup.8 d9, d11[1]
vdup.8 d10, d11[2]
vdup.8 d11, d11[3]
.endm
.macro pixman_composite_over_n_8_0565_cleanup
vpop {d8-d15}
.endm
generate_composite_function \
pixman_composite_over_n_8_0565_asm_neon, 0, 8, 16, \
FLAG_DST_READWRITE, \
8, /* number of pixels, processed in a single block */ \
5, /* prefetch distance */ \
pixman_composite_over_n_8_0565_init, \
pixman_composite_over_n_8_0565_cleanup, \
pixman_composite_over_n_8_0565_process_pixblock_head, \
pixman_composite_over_n_8_0565_process_pixblock_tail, \
pixman_composite_over_n_8_0565_process_pixblock_tail_head
/******************************************************************************/
.macro pixman_composite_src_0565_0565_process_pixblock_head
.endm
.macro pixman_composite_src_0565_0565_process_pixblock_tail
.endm
.macro pixman_composite_src_0565_0565_process_pixblock_tail_head
vst1.16 {d0, d1, d2, d3}, [DST_W, :128]!
vld1.16 {d0, d1, d2, d3}, [SRC]!
cache_preload 16, 16
.endm
generate_composite_function \
pixman_composite_src_0565_0565_asm_neon, 16, 0, 16, \
FLAG_DST_WRITEONLY, \
16, /* number of pixels, processed in a single block */ \
10, /* prefetch distance */ \
default_init, \
default_cleanup, \
pixman_composite_src_0565_0565_process_pixblock_head, \
pixman_composite_src_0565_0565_process_pixblock_tail, \
pixman_composite_src_0565_0565_process_pixblock_tail_head, \
0, /* dst_w_basereg */ \
0, /* dst_r_basereg */ \
0, /* src_basereg */ \
0 /* mask_basereg */
/******************************************************************************/
.macro pixman_composite_src_n_8_process_pixblock_head
.endm
.macro pixman_composite_src_n_8_process_pixblock_tail
.endm
.macro pixman_composite_src_n_8_process_pixblock_tail_head
vst1.8 {d0, d1, d2, d3}, [DST_W, :128]!
.endm
.macro pixman_composite_src_n_8_init
add DUMMY, sp, #ARGS_STACK_OFFSET
vld1.32 {d0[0]}, [DUMMY]
vsli.u64 d0, d0, #8
vsli.u64 d0, d0, #16
vsli.u64 d0, d0, #32
vmov d1, d0
vmov q1, q0
.endm
.macro pixman_composite_src_n_8_cleanup
.endm
generate_composite_function \
pixman_composite_src_n_8_asm_neon, 0, 0, 8, \
FLAG_DST_WRITEONLY, \
32, /* number of pixels, processed in a single block */ \
0, /* prefetch distance */ \
pixman_composite_src_n_8_init, \
pixman_composite_src_n_8_cleanup, \
pixman_composite_src_n_8_process_pixblock_head, \
pixman_composite_src_n_8_process_pixblock_tail, \
pixman_composite_src_n_8_process_pixblock_tail_head, \
0, /* dst_w_basereg */ \
0, /* dst_r_basereg */ \
0, /* src_basereg */ \
0 /* mask_basereg */
/******************************************************************************/
.macro pixman_composite_src_n_0565_process_pixblock_head
.endm
.macro pixman_composite_src_n_0565_process_pixblock_tail
.endm
.macro pixman_composite_src_n_0565_process_pixblock_tail_head
vst1.16 {d0, d1, d2, d3}, [DST_W, :128]!
.endm
.macro pixman_composite_src_n_0565_init
add DUMMY, sp, #ARGS_STACK_OFFSET
vld1.32 {d0[0]}, [DUMMY]
vsli.u64 d0, d0, #16
vsli.u64 d0, d0, #32
vmov d1, d0
vmov q1, q0
.endm
.macro pixman_composite_src_n_0565_cleanup
.endm
generate_composite_function \
pixman_composite_src_n_0565_asm_neon, 0, 0, 16, \
FLAG_DST_WRITEONLY, \
16, /* number of pixels, processed in a single block */ \
0, /* prefetch distance */ \
pixman_composite_src_n_0565_init, \
pixman_composite_src_n_0565_cleanup, \
pixman_composite_src_n_0565_process_pixblock_head, \
pixman_composite_src_n_0565_process_pixblock_tail, \
pixman_composite_src_n_0565_process_pixblock_tail_head, \
0, /* dst_w_basereg */ \
0, /* dst_r_basereg */ \
0, /* src_basereg */ \
0 /* mask_basereg */
/******************************************************************************/
.macro pixman_composite_src_n_8888_process_pixblock_head
.endm
.macro pixman_composite_src_n_8888_process_pixblock_tail
.endm
.macro pixman_composite_src_n_8888_process_pixblock_tail_head
vst1.32 {d0, d1, d2, d3}, [DST_W, :128]!
.endm
.macro pixman_composite_src_n_8888_init
add DUMMY, sp, #ARGS_STACK_OFFSET
vld1.32 {d0[0]}, [DUMMY]
vsli.u64 d0, d0, #32
vmov d1, d0
vmov q1, q0
.endm
.macro pixman_composite_src_n_8888_cleanup
.endm
generate_composite_function \
pixman_composite_src_n_8888_asm_neon, 0, 0, 32, \
FLAG_DST_WRITEONLY, \
8, /* number of pixels, processed in a single block */ \
0, /* prefetch distance */ \
pixman_composite_src_n_8888_init, \
pixman_composite_src_n_8888_cleanup, \
pixman_composite_src_n_8888_process_pixblock_head, \
pixman_composite_src_n_8888_process_pixblock_tail, \
pixman_composite_src_n_8888_process_pixblock_tail_head, \
0, /* dst_w_basereg */ \
0, /* dst_r_basereg */ \
0, /* src_basereg */ \
0 /* mask_basereg */
/******************************************************************************/
.macro pixman_composite_src_8888_8888_process_pixblock_head
.endm
.macro pixman_composite_src_8888_8888_process_pixblock_tail
.endm
.macro pixman_composite_src_8888_8888_process_pixblock_tail_head
vst1.32 {d0, d1, d2, d3}, [DST_W, :128]!
vld1.32 {d0, d1, d2, d3}, [SRC]!
cache_preload 8, 8
.endm
generate_composite_function \
pixman_composite_src_8888_8888_asm_neon, 32, 0, 32, \
FLAG_DST_WRITEONLY, \
8, /* number of pixels, processed in a single block */ \
10, /* prefetch distance */ \
default_init, \
default_cleanup, \
pixman_composite_src_8888_8888_process_pixblock_head, \
pixman_composite_src_8888_8888_process_pixblock_tail, \
pixman_composite_src_8888_8888_process_pixblock_tail_head, \
0, /* dst_w_basereg */ \
0, /* dst_r_basereg */ \
0, /* src_basereg */ \
0 /* mask_basereg */
/******************************************************************************/
.macro pixman_composite_over_n_8_8888_process_pixblock_head
/* expecting deinterleaved source data in {d8, d9, d10, d11} */
/* d8 - blue, d9 - green, d10 - red, d11 - alpha */
/* and destination data in {d4, d5, d6, d7} */
/* mask is in d24 (d25, d26, d27 are unused) */
/* in */
vmull.u8 q0, d24, d8
vmull.u8 q1, d24, d9
vmull.u8 q6, d24, d10
vmull.u8 q7, d24, d11
vrshr.u16 q10, q0, #8
vrshr.u16 q11, q1, #8
vrshr.u16 q12, q6, #8
vrshr.u16 q13, q7, #8
vraddhn.u16 d0, q0, q10
vraddhn.u16 d1, q1, q11
vraddhn.u16 d2, q6, q12
vraddhn.u16 d3, q7, q13
vmvn.8 d24, d3 /* get inverted alpha */
/* source: d0 - blue, d1 - green, d2 - red, d3 - alpha */
/* destination: d4 - blue, d5 - green, d6 - red, d7 - alpha */
/* now do alpha blending */
vmull.u8 q8, d24, d4
vmull.u8 q9, d24, d5
vmull.u8 q10, d24, d6
vmull.u8 q11, d24, d7
.endm
.macro pixman_composite_over_n_8_8888_process_pixblock_tail
vrshr.u16 q14, q8, #8
vrshr.u16 q15, q9, #8
vrshr.u16 q12, q10, #8
vrshr.u16 q13, q11, #8
vraddhn.u16 d28, q14, q8
vraddhn.u16 d29, q15, q9
vraddhn.u16 d30, q12, q10
vraddhn.u16 d31, q13, q11
vqadd.u8 q14, q0, q14
vqadd.u8 q15, q1, q15
.endm
/* TODO: expand macros and do better instructions scheduling */
.macro pixman_composite_over_n_8_8888_process_pixblock_tail_head
pixman_composite_over_n_8_8888_process_pixblock_tail
vst4.8 {d28, d29, d30, d31}, [DST_W, :128]!
vld4.8 {d4, d5, d6, d7}, [DST_R, :128]!
vld1.8 {d24}, [MASK]!
cache_preload 8, 8
pixman_composite_over_n_8_8888_process_pixblock_head
.endm
.macro pixman_composite_over_n_8_8888_init
add DUMMY, sp, #ARGS_STACK_OFFSET
vpush {d8-d15}
vld1.32 {d11[0]}, [DUMMY]
vdup.8 d8, d11[0]
vdup.8 d9, d11[1]
vdup.8 d10, d11[2]
vdup.8 d11, d11[3]
.endm
.macro pixman_composite_over_n_8_8888_cleanup
vpop {d8-d15}
.endm
generate_composite_function \
pixman_composite_over_n_8_8888_asm_neon, 0, 8, 32, \
FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
8, /* number of pixels, processed in a single block */ \
5, /* prefetch distance */ \
pixman_composite_over_n_8_8888_init, \
pixman_composite_over_n_8_8888_cleanup, \
pixman_composite_over_n_8_8888_process_pixblock_head, \
pixman_composite_over_n_8_8888_process_pixblock_tail, \
pixman_composite_over_n_8_8888_process_pixblock_tail_head
/******************************************************************************/
.macro pixman_composite_add_n_8_8_process_pixblock_head
/* expecting source data in {d8, d9, d10, d11} */
/* d8 - blue, d9 - green, d10 - red, d11 - alpha */
/* and destination data in {d4, d5, d6, d7} */
/* mask is in d24, d25, d26, d27 */
vmull.u8 q0, d24, d11
vmull.u8 q1, d25, d11
vmull.u8 q6, d26, d11
vmull.u8 q7, d27, d11
vrshr.u16 q10, q0, #8
vrshr.u16 q11, q1, #8
vrshr.u16 q12, q6, #8
vrshr.u16 q13, q7, #8
vraddhn.u16 d0, q0, q10
vraddhn.u16 d1, q1, q11
vraddhn.u16 d2, q6, q12
vraddhn.u16 d3, q7, q13
vqadd.u8 q14, q0, q2
vqadd.u8 q15, q1, q3
.endm
.macro pixman_composite_add_n_8_8_process_pixblock_tail
.endm
/* TODO: expand macros and do better instructions scheduling */
.macro pixman_composite_add_n_8_8_process_pixblock_tail_head
pixman_composite_add_n_8_8_process_pixblock_tail
vst1.8 {d28, d29, d30, d31}, [DST_W, :128]!
vld1.8 {d4, d5, d6, d7}, [DST_R, :128]!
vld1.8 {d24, d25, d26, d27}, [MASK]!
cache_preload 32, 32
pixman_composite_add_n_8_8_process_pixblock_head
.endm
.macro pixman_composite_add_n_8_8_init
add DUMMY, sp, #ARGS_STACK_OFFSET
vpush {d8-d15}
vld1.32 {d11[0]}, [DUMMY]
vdup.8 d11, d11[3]
.endm
.macro pixman_composite_add_n_8_8_cleanup
vpop {d8-d15}
.endm
generate_composite_function \
pixman_composite_add_n_8_8_asm_neon, 0, 8, 8, \
FLAG_DST_READWRITE, \
32, /* number of pixels, processed in a single block */ \
5, /* prefetch distance */ \
pixman_composite_add_n_8_8_init, \
pixman_composite_add_n_8_8_cleanup, \
pixman_composite_add_n_8_8_process_pixblock_head, \
pixman_composite_add_n_8_8_process_pixblock_tail, \
pixman_composite_add_n_8_8_process_pixblock_tail_head
/******************************************************************************/
.macro pixman_composite_add_8_8_8_process_pixblock_head
/* expecting source data in {d0, d1, d2, d3} */
/* destination data in {d4, d5, d6, d7} */
/* mask in {d24, d25, d26, d27} */
vmull.u8 q8, d24, d0
vmull.u8 q9, d25, d1
vmull.u8 q10, d26, d2
vmull.u8 q11, d27, d3
vrshr.u16 q0, q8, #8
vrshr.u16 q1, q9, #8
vrshr.u16 q12, q10, #8
vrshr.u16 q13, q11, #8
vraddhn.u16 d0, q0, q8
vraddhn.u16 d1, q1, q9
vraddhn.u16 d2, q12, q10
vraddhn.u16 d3, q13, q11
vqadd.u8 q14, q0, q2
vqadd.u8 q15, q1, q3
.endm
.macro pixman_composite_add_8_8_8_process_pixblock_tail
.endm
/* TODO: expand macros and do better instructions scheduling */
.macro pixman_composite_add_8_8_8_process_pixblock_tail_head
pixman_composite_add_8_8_8_process_pixblock_tail
vst1.8 {d28, d29, d30, d31}, [DST_W, :128]!
vld1.8 {d4, d5, d6, d7}, [DST_R, :128]!
vld1.8 {d24, d25, d26, d27}, [MASK]!
vld1.8 {d0, d1, d2, d3}, [SRC]!
cache_preload 32, 32
pixman_composite_add_8_8_8_process_pixblock_head
.endm
.macro pixman_composite_add_8_8_8_init
.endm
.macro pixman_composite_add_8_8_8_cleanup
.endm
generate_composite_function \
pixman_composite_add_8_8_8_asm_neon, 8, 8, 8, \
FLAG_DST_READWRITE, \
32, /* number of pixels, processed in a single block */ \
5, /* prefetch distance */ \
pixman_composite_add_8_8_8_init, \
pixman_composite_add_8_8_8_cleanup, \
pixman_composite_add_8_8_8_process_pixblock_head, \
pixman_composite_add_8_8_8_process_pixblock_tail, \
pixman_composite_add_8_8_8_process_pixblock_tail_head
/******************************************************************************/
.macro pixman_composite_add_8888_8888_8888_process_pixblock_head
/* expecting source data in {d0, d1, d2, d3} */
/* destination data in {d4, d5, d6, d7} */
/* mask in {d24, d25, d26, d27} */
vmull.u8 q8, d27, d0
vmull.u8 q9, d27, d1
vmull.u8 q10, d27, d2
vmull.u8 q11, d27, d3
vrshr.u16 q0, q8, #8
vrshr.u16 q1, q9, #8
vrshr.u16 q12, q10, #8
vrshr.u16 q13, q11, #8
vraddhn.u16 d0, q0, q8
vraddhn.u16 d1, q1, q9
vraddhn.u16 d2, q12, q10
vraddhn.u16 d3, q13, q11
vqadd.u8 q14, q0, q2
vqadd.u8 q15, q1, q3
.endm
.macro pixman_composite_add_8888_8888_8888_process_pixblock_tail
.endm
/* TODO: expand macros and do better instructions scheduling */
.macro pixman_composite_add_8888_8888_8888_process_pixblock_tail_head
pixman_composite_add_8888_8888_8888_process_pixblock_tail
vst4.8 {d28, d29, d30, d31}, [DST_W, :128]!
vld4.8 {d4, d5, d6, d7}, [DST_R, :128]!
vld4.8 {d24, d25, d26, d27}, [MASK]!
vld4.8 {d0, d1, d2, d3}, [SRC]!
cache_preload 8, 8
pixman_composite_add_8888_8888_8888_process_pixblock_head
.endm
generate_composite_function \
pixman_composite_add_8888_8888_8888_asm_neon, 32, 32, 32, \
FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
8, /* number of pixels, processed in a single block */ \
10, /* prefetch distance */ \
default_init, \
default_cleanup, \
pixman_composite_add_8888_8888_8888_process_pixblock_head, \
pixman_composite_add_8888_8888_8888_process_pixblock_tail, \
pixman_composite_add_8888_8888_8888_process_pixblock_tail_head
generate_composite_function_single_scanline \
pixman_composite_scanline_add_mask_asm_neon, 32, 32, 32, \
FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
8, /* number of pixels, processed in a single block */ \
default_init, \
default_cleanup, \
pixman_composite_add_8888_8888_8888_process_pixblock_head, \
pixman_composite_add_8888_8888_8888_process_pixblock_tail, \
pixman_composite_add_8888_8888_8888_process_pixblock_tail_head
/******************************************************************************/
.macro pixman_composite_over_8888_n_8888_process_pixblock_head
/* expecting source data in {d0, d1, d2, d3} */
/* destination data in {d4, d5, d6, d7} */
/* solid mask is in d15 */
/* 'in' */
vmull.u8 q8, d15, d3
vmull.u8 q6, d15, d2
vmull.u8 q5, d15, d1
vmull.u8 q4, d15, d0
vrshr.u16 q13, q8, #8
vrshr.u16 q12, q6, #8
vrshr.u16 q11, q5, #8
vrshr.u16 q10, q4, #8
vraddhn.u16 d3, q8, q13
vraddhn.u16 d2, q6, q12
vraddhn.u16 d1, q5, q11
vraddhn.u16 d0, q4, q10
vmvn.8 d24, d3 /* get inverted alpha */
/* now do alpha blending */
vmull.u8 q8, d24, d4
vmull.u8 q9, d24, d5
vmull.u8 q10, d24, d6
vmull.u8 q11, d24, d7
.endm
.macro pixman_composite_over_8888_n_8888_process_pixblock_tail
vrshr.u16 q14, q8, #8
vrshr.u16 q15, q9, #8
vrshr.u16 q12, q10, #8
vrshr.u16 q13, q11, #8
vraddhn.u16 d28, q14, q8
vraddhn.u16 d29, q15, q9
vraddhn.u16 d30, q12, q10
vraddhn.u16 d31, q13, q11
vqadd.u8 q14, q0, q14
vqadd.u8 q15, q1, q15
.endm
/* TODO: expand macros and do better instructions scheduling */
.macro pixman_composite_over_8888_n_8888_process_pixblock_tail_head
vld4.8 {d4, d5, d6, d7}, [DST_R, :128]!
pixman_composite_over_8888_n_8888_process_pixblock_tail
vld4.8 {d0, d1, d2, d3}, [SRC]!
cache_preload 8, 8
pixman_composite_over_8888_n_8888_process_pixblock_head
vst4.8 {d28, d29, d30, d31}, [DST_W, :128]!
.endm
.macro pixman_composite_over_8888_n_8888_init
add DUMMY, sp, #48
vpush {d8-d15}
vld1.32 {d15[0]}, [DUMMY]
vdup.8 d15, d15[3]
.endm
.macro pixman_composite_over_8888_n_8888_cleanup
vpop {d8-d15}
.endm
generate_composite_function \
pixman_composite_over_8888_n_8888_asm_neon, 32, 0, 32, \
FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
8, /* number of pixels, processed in a single block */ \
5, /* prefetch distance */ \
pixman_composite_over_8888_n_8888_init, \
pixman_composite_over_8888_n_8888_cleanup, \
pixman_composite_over_8888_n_8888_process_pixblock_head, \
pixman_composite_over_8888_n_8888_process_pixblock_tail, \
pixman_composite_over_8888_n_8888_process_pixblock_tail_head
/******************************************************************************/
/* TODO: expand macros and do better instructions scheduling */
.macro pixman_composite_over_8888_8888_8888_process_pixblock_tail_head
vld4.8 {d4, d5, d6, d7}, [DST_R, :128]!
pixman_composite_over_8888_n_8888_process_pixblock_tail
vld4.8 {d0, d1, d2, d3}, [SRC]!
cache_preload 8, 8
vld4.8 {d12, d13, d14, d15}, [MASK]!
pixman_composite_over_8888_n_8888_process_pixblock_head
vst4.8 {d28, d29, d30, d31}, [DST_W, :128]!
.endm
.macro pixman_composite_over_8888_8888_8888_init
vpush {d8-d15}
.endm
.macro pixman_composite_over_8888_8888_8888_cleanup
vpop {d8-d15}
.endm
generate_composite_function \
pixman_composite_over_8888_8888_8888_asm_neon, 32, 32, 32, \
FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
8, /* number of pixels, processed in a single block */ \
5, /* prefetch distance */ \
pixman_composite_over_8888_8888_8888_init, \
pixman_composite_over_8888_8888_8888_cleanup, \
pixman_composite_over_8888_n_8888_process_pixblock_head, \
pixman_composite_over_8888_n_8888_process_pixblock_tail, \
pixman_composite_over_8888_8888_8888_process_pixblock_tail_head \
28, /* dst_w_basereg */ \
4, /* dst_r_basereg */ \
0, /* src_basereg */ \
12 /* mask_basereg */
generate_composite_function_single_scanline \
pixman_composite_scanline_over_mask_asm_neon, 32, 32, 32, \
FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
8, /* number of pixels, processed in a single block */ \
pixman_composite_over_8888_8888_8888_init, \
pixman_composite_over_8888_8888_8888_cleanup, \
pixman_composite_over_8888_n_8888_process_pixblock_head, \
pixman_composite_over_8888_n_8888_process_pixblock_tail, \
pixman_composite_over_8888_8888_8888_process_pixblock_tail_head \
28, /* dst_w_basereg */ \
4, /* dst_r_basereg */ \
0, /* src_basereg */ \
12 /* mask_basereg */
/******************************************************************************/
/* TODO: expand macros and do better instructions scheduling */
.macro pixman_composite_over_8888_8_8888_process_pixblock_tail_head
vld4.8 {d4, d5, d6, d7}, [DST_R, :128]!
pixman_composite_over_8888_n_8888_process_pixblock_tail
vld4.8 {d0, d1, d2, d3}, [SRC]!
cache_preload 8, 8
vld1.8 {d15}, [MASK]!
pixman_composite_over_8888_n_8888_process_pixblock_head
vst4.8 {d28, d29, d30, d31}, [DST_W, :128]!
.endm
.macro pixman_composite_over_8888_8_8888_init
vpush {d8-d15}
.endm
.macro pixman_composite_over_8888_8_8888_cleanup
vpop {d8-d15}
.endm
generate_composite_function \
pixman_composite_over_8888_8_8888_asm_neon, 32, 8, 32, \
FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
8, /* number of pixels, processed in a single block */ \
5, /* prefetch distance */ \
pixman_composite_over_8888_8_8888_init, \
pixman_composite_over_8888_8_8888_cleanup, \
pixman_composite_over_8888_n_8888_process_pixblock_head, \
pixman_composite_over_8888_n_8888_process_pixblock_tail, \
pixman_composite_over_8888_8_8888_process_pixblock_tail_head \
28, /* dst_w_basereg */ \
4, /* dst_r_basereg */ \
0, /* src_basereg */ \
15 /* mask_basereg */
/******************************************************************************/
.macro pixman_composite_src_0888_0888_process_pixblock_head
.endm
.macro pixman_composite_src_0888_0888_process_pixblock_tail
.endm
.macro pixman_composite_src_0888_0888_process_pixblock_tail_head
vst3.8 {d0, d1, d2}, [DST_W]!
vld3.8 {d0, d1, d2}, [SRC]!
cache_preload 8, 8
.endm
generate_composite_function \
pixman_composite_src_0888_0888_asm_neon, 24, 0, 24, \
FLAG_DST_WRITEONLY, \
8, /* number of pixels, processed in a single block */ \
10, /* prefetch distance */ \
default_init, \
default_cleanup, \
pixman_composite_src_0888_0888_process_pixblock_head, \
pixman_composite_src_0888_0888_process_pixblock_tail, \
pixman_composite_src_0888_0888_process_pixblock_tail_head, \
0, /* dst_w_basereg */ \
0, /* dst_r_basereg */ \
0, /* src_basereg */ \
0 /* mask_basereg */
/******************************************************************************/
.macro pixman_composite_src_0888_8888_rev_process_pixblock_head
vswp d0, d2
.endm
.macro pixman_composite_src_0888_8888_rev_process_pixblock_tail
.endm
.macro pixman_composite_src_0888_8888_rev_process_pixblock_tail_head
vst4.8 {d0, d1, d2, d3}, [DST_W]!
vld3.8 {d0, d1, d2}, [SRC]!
vswp d0, d2
cache_preload 8, 8
.endm
.macro pixman_composite_src_0888_8888_rev_init
veor d3, d3, d3
.endm
generate_composite_function \
pixman_composite_src_0888_8888_rev_asm_neon, 24, 0, 32, \
FLAG_DST_WRITEONLY | FLAG_DEINTERLEAVE_32BPP, \
8, /* number of pixels, processed in a single block */ \
10, /* prefetch distance */ \
pixman_composite_src_0888_8888_rev_init, \
default_cleanup, \
pixman_composite_src_0888_8888_rev_process_pixblock_head, \
pixman_composite_src_0888_8888_rev_process_pixblock_tail, \
pixman_composite_src_0888_8888_rev_process_pixblock_tail_head, \
0, /* dst_w_basereg */ \
0, /* dst_r_basereg */ \
0, /* src_basereg */ \
0 /* mask_basereg */
/******************************************************************************/
.macro pixman_composite_src_0888_0565_rev_process_pixblock_head
vshll.u8 q8, d1, #8
vshll.u8 q9, d2, #8
.endm
.macro pixman_composite_src_0888_0565_rev_process_pixblock_tail
vshll.u8 q14, d0, #8
vsri.u16 q14, q8, #5
vsri.u16 q14, q9, #11
.endm
.macro pixman_composite_src_0888_0565_rev_process_pixblock_tail_head
vshll.u8 q14, d0, #8
vld3.8 {d0, d1, d2}, [SRC]!
vsri.u16 q14, q8, #5
vsri.u16 q14, q9, #11
vshll.u8 q8, d1, #8
vst1.16 {d28, d29}, [DST_W, :128]!
vshll.u8 q9, d2, #8
.endm
generate_composite_function \
pixman_composite_src_0888_0565_rev_asm_neon, 24, 0, 16, \
FLAG_DST_WRITEONLY, \
8, /* number of pixels, processed in a single block */ \
10, /* prefetch distance */ \
default_init, \
default_cleanup, \
pixman_composite_src_0888_0565_rev_process_pixblock_head, \
pixman_composite_src_0888_0565_rev_process_pixblock_tail, \
pixman_composite_src_0888_0565_rev_process_pixblock_tail_head, \
28, /* dst_w_basereg */ \
0, /* dst_r_basereg */ \
0, /* src_basereg */ \
0 /* mask_basereg */
/******************************************************************************/
.macro pixman_composite_src_pixbuf_8888_process_pixblock_head
vmull.u8 q8, d3, d0
vmull.u8 q9, d3, d1
vmull.u8 q10, d3, d2
.endm
.macro pixman_composite_src_pixbuf_8888_process_pixblock_tail
vrshr.u16 q11, q8, #8
vswp d3, d31
vrshr.u16 q12, q9, #8
vrshr.u16 q13, q10, #8
vraddhn.u16 d30, q11, q8
vraddhn.u16 d29, q12, q9
vraddhn.u16 d28, q13, q10
.endm
.macro pixman_composite_src_pixbuf_8888_process_pixblock_tail_head
vrshr.u16 q11, q8, #8
vswp d3, d31
vrshr.u16 q12, q9, #8
vrshr.u16 q13, q10, #8
vld4.8 {d0, d1, d2, d3}, [SRC]!
vraddhn.u16 d30, q11, q8
PF add PF_X, PF_X, #8
PF tst PF_CTL, #0xF
PF addne PF_X, PF_X, #8
PF subne PF_CTL, PF_CTL, #1
vraddhn.u16 d29, q12, q9
vraddhn.u16 d28, q13, q10
vmull.u8 q8, d3, d0
vmull.u8 q9, d3, d1
vmull.u8 q10, d3, d2
vst4.8 {d28, d29, d30, d31}, [DST_W, :128]!
PF cmp PF_X, ORIG_W
PF pld, [PF_SRC, PF_X, lsl #src_bpp_shift]
PF subge PF_X, PF_X, ORIG_W
PF subges PF_CTL, PF_CTL, #0x10
PF ldrgeb DUMMY, [PF_SRC, SRC_STRIDE, lsl #src_bpp_shift]!
.endm
generate_composite_function \
pixman_composite_src_pixbuf_8888_asm_neon, 32, 0, 32, \
FLAG_DST_WRITEONLY | FLAG_DEINTERLEAVE_32BPP, \
8, /* number of pixels, processed in a single block */ \
10, /* prefetch distance */ \
default_init, \
default_cleanup, \
pixman_composite_src_pixbuf_8888_process_pixblock_head, \
pixman_composite_src_pixbuf_8888_process_pixblock_tail, \
pixman_composite_src_pixbuf_8888_process_pixblock_tail_head, \
28, /* dst_w_basereg */ \
0, /* dst_r_basereg */ \
0, /* src_basereg */ \
0 /* mask_basereg */
|