filter_msa_intrinsics.c
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
/* filter_msa_intrinsics.c - MSA optimised filter functions
*
* Copyright (c) 2016 Glenn Randers-Pehrson
* Written by Mandar Sahastrabuddhe, August 2016.
* Last changed in libpng 1.6.25 [September 1, 2016]
*
* This code is released under the libpng license.
* For conditions of distribution and use, see the disclaimer
* and license in png.h
*/
#include <stdio.h>
#include <stdint.h>
#include "../pngpriv.h"
#ifdef PNG_READ_SUPPORTED
/* This code requires -mfpu=msa on the command line: */
#if PNG_MIPS_MSA_IMPLEMENTATION == 1 /* intrinsics code from pngpriv.h */
#include <msa.h>
/* libpng row pointers are not necessarily aligned to any particular boundary,
* however this code will only work with appropriate alignment. mips/mips_init.c
* checks for this (and will not compile unless it is done). This code uses
* variants of png_aligncast to avoid compiler warnings.
*/
#define png_ptr(type,pointer) png_aligncast(type *,pointer)
#define png_ptrc(type,pointer) png_aligncastconst(const type *,pointer)
/* The following relies on a variable 'temp_pointer' being declared with type
* 'type'. This is written this way just to hide the GCC strict aliasing
* warning; note that the code is safe because there never is an alias between
* the input and output pointers.
*/
#define png_ldr(type,pointer)\
(temp_pointer = png_ptr(type,pointer), *temp_pointer)
#if PNG_MIPS_MSA_OPT > 0
#ifdef CLANG_BUILD
#define MSA_SRLI_B(a, b) __msa_srli_b((v16i8) a, b)
#define LW(psrc) \
( { \
uint8_t *psrc_lw_m = (uint8_t *) (psrc); \
uint32_t val_m; \
\
asm volatile ( \
"lw %[val_m], %[psrc_lw_m] \n\t" \
\
: [val_m] "=r" (val_m) \
: [psrc_lw_m] "m" (*psrc_lw_m) \
); \
\
val_m; \
} )
#define SH(val, pdst) \
{ \
uint8_t *pdst_sh_m = (uint8_t *) (pdst); \
uint16_t val_m = (val); \
\
asm volatile ( \
"sh %[val_m], %[pdst_sh_m] \n\t" \
\
: [pdst_sh_m] "=m" (*pdst_sh_m) \
: [val_m] "r" (val_m) \
); \
}
#define SW(val, pdst) \
{ \
uint8_t *pdst_sw_m = (uint8_t *) (pdst); \
uint32_t val_m = (val); \
\
asm volatile ( \
"sw %[val_m], %[pdst_sw_m] \n\t" \
\
: [pdst_sw_m] "=m" (*pdst_sw_m) \
: [val_m] "r" (val_m) \
); \
}
#if (__mips == 64)
#define SD(val, pdst) \
{ \
uint8_t *pdst_sd_m = (uint8_t *) (pdst); \
uint64_t val_m = (val); \
\
asm volatile ( \
"sd %[val_m], %[pdst_sd_m] \n\t" \
\
: [pdst_sd_m] "=m" (*pdst_sd_m) \
: [val_m] "r" (val_m) \
); \
}
#else
#define SD(val, pdst) \
{ \
uint8_t *pdst_sd_m = (uint8_t *) (pdst); \
uint32_t val0_m, val1_m; \
\
val0_m = (uint32_t) ((val) & 0x00000000FFFFFFFF); \
val1_m = (uint32_t) (((val) >> 32) & 0x00000000FFFFFFFF); \
\
SW(val0_m, pdst_sd_m); \
SW(val1_m, pdst_sd_m + 4); \
}
#endif
#else
#define MSA_SRLI_B(a, b) (a >> b)
#if (__mips_isa_rev >= 6)
#define LW(psrc) \
( { \
uint8_t *psrc_lw_m = (uint8_t *) (psrc); \
uint32_t val_m; \
\
asm volatile ( \
"lw %[val_m], %[psrc_lw_m] \n\t" \
\
: [val_m] "=r" (val_m) \
: [psrc_lw_m] "m" (*psrc_lw_m) \
); \
\
val_m; \
} )
#define SH(val, pdst) \
{ \
uint8_t *pdst_sh_m = (uint8_t *) (pdst); \
uint16_t val_m = (val); \
\
asm volatile ( \
"sh %[val_m], %[pdst_sh_m] \n\t" \
\
: [pdst_sh_m] "=m" (*pdst_sh_m) \
: [val_m] "r" (val_m) \
); \
}
#define SW(val, pdst) \
{ \
uint8_t *pdst_sw_m = (uint8_t *) (pdst); \
uint32_t val_m = (val); \
\
asm volatile ( \
"sw %[val_m], %[pdst_sw_m] \n\t" \
\
: [pdst_sw_m] "=m" (*pdst_sw_m) \
: [val_m] "r" (val_m) \
); \
}
#if (__mips == 64)
#define SD(val, pdst) \
{ \
uint8_t *pdst_sd_m = (uint8_t *) (pdst); \
uint64_t val_m = (val); \
\
asm volatile ( \
"sd %[val_m], %[pdst_sd_m] \n\t" \
\
: [pdst_sd_m] "=m" (*pdst_sd_m) \
: [val_m] "r" (val_m) \
); \
}
#else
#define SD(val, pdst) \
{ \
uint8_t *pdst_sd_m = (uint8_t *) (pdst); \
uint32_t val0_m, val1_m; \
\
val0_m = (uint32_t) ((val) & 0x00000000FFFFFFFF); \
val1_m = (uint32_t) (((val) >> 32) & 0x00000000FFFFFFFF); \
\
SW(val0_m, pdst_sd_m); \
SW(val1_m, pdst_sd_m + 4); \
}
#endif
#else // !(__mips_isa_rev >= 6)
#define LW(psrc) \
( { \
uint8_t *psrc_lw_m = (uint8_t *) (psrc); \
uint32_t val_m; \
\
asm volatile ( \
"ulw %[val_m], %[psrc_lw_m] \n\t" \
\
: [val_m] "=r" (val_m) \
: [psrc_lw_m] "m" (*psrc_lw_m) \
); \
\
val_m; \
} )
#define SH(val, pdst) \
{ \
uint8_t *pdst_sh_m = (uint8_t *) (pdst); \
uint16_t val_m = (val); \
\
asm volatile ( \
"ush %[val_m], %[pdst_sh_m] \n\t" \
\
: [pdst_sh_m] "=m" (*pdst_sh_m) \
: [val_m] "r" (val_m) \
); \
}
#define SW(val, pdst) \
{ \
uint8_t *pdst_sw_m = (uint8_t *) (pdst); \
uint32_t val_m = (val); \
\
asm volatile ( \
"usw %[val_m], %[pdst_sw_m] \n\t" \
\
: [pdst_sw_m] "=m" (*pdst_sw_m) \
: [val_m] "r" (val_m) \
); \
}
#define SD(val, pdst) \
{ \
uint8_t *pdst_sd_m = (uint8_t *) (pdst); \
uint32_t val0_m, val1_m; \
\
val0_m = (uint32_t) ((val) & 0x00000000FFFFFFFF); \
val1_m = (uint32_t) (((val) >> 32) & 0x00000000FFFFFFFF); \
\
SW(val0_m, pdst_sd_m); \
SW(val1_m, pdst_sd_m + 4); \
}
#define SW_ZERO(pdst) \
{ \
uint8_t *pdst_m = (uint8_t *) (pdst); \
\
asm volatile ( \
"usw $0, %[pdst_m] \n\t" \
\
: [pdst_m] "=m" (*pdst_m) \
: \
); \
}
#endif // (__mips_isa_rev >= 6)
#endif
#define LD_B(RTYPE, psrc) *((RTYPE *) (psrc))
#define LD_UB(...) LD_B(v16u8, __VA_ARGS__)
#define LD_B2(RTYPE, psrc, stride, out0, out1) \
{ \
out0 = LD_B(RTYPE, (psrc)); \
out1 = LD_B(RTYPE, (psrc) + stride); \
}
#define LD_UB2(...) LD_B2(v16u8, __VA_ARGS__)
#define LD_B4(RTYPE, psrc, stride, out0, out1, out2, out3) \
{ \
LD_B2(RTYPE, (psrc), stride, out0, out1); \
LD_B2(RTYPE, (psrc) + 2 * stride , stride, out2, out3); \
}
#define LD_UB4(...) LD_B4(v16u8, __VA_ARGS__)
#define ST_B(RTYPE, in, pdst) *((RTYPE *) (pdst)) = (in)
#define ST_UB(...) ST_B(v16u8, __VA_ARGS__)
#define ST_B2(RTYPE, in0, in1, pdst, stride) \
{ \
ST_B(RTYPE, in0, (pdst)); \
ST_B(RTYPE, in1, (pdst) + stride); \
}
#define ST_UB2(...) ST_B2(v16u8, __VA_ARGS__)
#define ST_B4(RTYPE, in0, in1, in2, in3, pdst, stride) \
{ \
ST_B2(RTYPE, in0, in1, (pdst), stride); \
ST_B2(RTYPE, in2, in3, (pdst) + 2 * stride, stride); \
}
#define ST_UB4(...) ST_B4(v16u8, __VA_ARGS__)
#define ADD2(in0, in1, in2, in3, out0, out1) \
{ \
out0 = in0 + in1; \
out1 = in2 + in3; \
}
#define ADD3(in0, in1, in2, in3, in4, in5, \
out0, out1, out2) \
{ \
ADD2(in0, in1, in2, in3, out0, out1); \
out2 = in4 + in5; \
}
#define ADD4(in0, in1, in2, in3, in4, in5, in6, in7, \
out0, out1, out2, out3) \
{ \
ADD2(in0, in1, in2, in3, out0, out1); \
ADD2(in4, in5, in6, in7, out2, out3); \
}
#define ILVR_B2(RTYPE, in0, in1, in2, in3, out0, out1) \
{ \
out0 = (RTYPE) __msa_ilvr_b((v16i8) in0, (v16i8) in1); \
out1 = (RTYPE) __msa_ilvr_b((v16i8) in2, (v16i8) in3); \
}
#define ILVR_B2_SH(...) ILVR_B2(v8i16, __VA_ARGS__)
#define HSUB_UB2(RTYPE, in0, in1, out0, out1) \
{ \
out0 = (RTYPE) __msa_hsub_u_h((v16u8) in0, (v16u8) in0); \
out1 = (RTYPE) __msa_hsub_u_h((v16u8) in1, (v16u8) in1); \
}
#define HSUB_UB2_SH(...) HSUB_UB2(v8i16, __VA_ARGS__)
#define SLDI_B2_0(RTYPE, in0, in1, out0, out1, slide_val) \
{ \
v16i8 zero_m = { 0 }; \
out0 = (RTYPE) __msa_sldi_b((v16i8) zero_m, (v16i8) in0, slide_val); \
out1 = (RTYPE) __msa_sldi_b((v16i8) zero_m, (v16i8) in1, slide_val); \
}
#define SLDI_B2_0_UB(...) SLDI_B2_0(v16u8, __VA_ARGS__)
#define SLDI_B3_0(RTYPE, in0, in1, in2, out0, out1, out2, slide_val) \
{ \
v16i8 zero_m = { 0 }; \
SLDI_B2_0(RTYPE, in0, in1, out0, out1, slide_val); \
out2 = (RTYPE) __msa_sldi_b((v16i8) zero_m, (v16i8) in2, slide_val); \
}
#define SLDI_B3_0_UB(...) SLDI_B3_0(v16u8, __VA_ARGS__)
#define ILVEV_W2(RTYPE, in0, in1, in2, in3, out0, out1) \
{ \
out0 = (RTYPE) __msa_ilvev_w((v4i32) in1, (v4i32) in0); \
out1 = (RTYPE) __msa_ilvev_w((v4i32) in3, (v4i32) in2); \
}
#define ILVEV_W2_UB(...) ILVEV_W2(v16u8, __VA_ARGS__)
#define ADD_ABS_H3(RTYPE, in0, in1, in2, out0, out1, out2) \
{ \
RTYPE zero = {0}; \
\
out0 = __msa_add_a_h((v8i16) zero, in0); \
out1 = __msa_add_a_h((v8i16) zero, in1); \
out2 = __msa_add_a_h((v8i16) zero, in2); \
}
#define ADD_ABS_H3_SH(...) ADD_ABS_H3(v8i16, __VA_ARGS__)
#define VSHF_B2(RTYPE, in0, in1, in2, in3, mask0, mask1, out0, out1) \
{ \
out0 = (RTYPE) __msa_vshf_b((v16i8) mask0, (v16i8) in1, (v16i8) in0); \
out1 = (RTYPE) __msa_vshf_b((v16i8) mask1, (v16i8) in3, (v16i8) in2); \
}
#define VSHF_B2_UB(...) VSHF_B2(v16u8, __VA_ARGS__)
#define CMP_AND_SELECT(inp0, inp1, inp2, inp3, inp4, inp5, out0) \
{ \
v8i16 _sel_h0, _sel_h1; \
v16u8 _sel_b0, _sel_b1; \
_sel_h0 = (v8i16) __msa_clt_u_h((v8u16) inp1, (v8u16) inp0); \
_sel_b0 = (v16u8) __msa_pckev_b((v16i8) _sel_h0, (v16i8) _sel_h0); \
inp0 = (v8i16) __msa_bmnz_v((v16u8) inp0, (v16u8) inp1, (v16u8) _sel_h0); \
inp4 = (v16u8) __msa_bmnz_v(inp3, inp4, _sel_b0); \
_sel_h1 = (v8i16) __msa_clt_u_h((v8u16) inp2, (v8u16) inp0); \
_sel_b1 = (v16u8) __msa_pckev_b((v16i8) _sel_h1, (v16i8) _sel_h1); \
inp4 = (v16u8) __msa_bmnz_v(inp4, inp5, _sel_b1); \
out0 += inp4; \
}
void png_read_filter_row_up_msa(png_row_infop row_info, png_bytep row,
png_const_bytep prev_row)
{
369
370
size_t i, cnt, cnt16, cnt32;
size_t istop = row_info->rowbytes;
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
png_bytep rp = row;
png_const_bytep pp = prev_row;
v16u8 src0, src1, src2, src3, src4, src5, src6, src7;
for (i = 0; i < (istop >> 6); i++)
{
LD_UB4(rp, 16, src0, src1, src2, src3);
LD_UB4(pp, 16, src4, src5, src6, src7);
pp += 64;
ADD4(src0, src4, src1, src5, src2, src6, src3, src7,
src0, src1, src2, src3);
ST_UB4(src0, src1, src2, src3, rp, 16);
rp += 64;
}
if (istop & 0x3F)
{
cnt32 = istop & 0x20;
cnt16 = istop & 0x10;
cnt = istop & 0xF;
if(cnt32)
{
if (cnt16 && cnt)
{
LD_UB4(rp, 16, src0, src1, src2, src3);
LD_UB4(pp, 16, src4, src5, src6, src7);
ADD4(src0, src4, src1, src5, src2, src6, src3, src7,
src0, src1, src2, src3);
ST_UB4(src0, src1, src2, src3, rp, 16);
rp += 64;
}
else if (cnt16 || cnt)
{
LD_UB2(rp, 16, src0, src1);
LD_UB2(pp, 16, src4, src5);
pp += 32;
src2 = LD_UB(rp + 32);
src6 = LD_UB(pp);
ADD3(src0, src4, src1, src5, src2, src6, src0, src1, src2);
ST_UB2(src0, src1, rp, 16);
rp += 32;
ST_UB(src2, rp);
rp += 16;
}
else
{
LD_UB2(rp, 16, src0, src1);
LD_UB2(pp, 16, src4, src5);
ADD2(src0, src4, src1, src5, src0, src1);
ST_UB2(src0, src1, rp, 16);
rp += 32;
}
}
else if (cnt16 && cnt)
{
LD_UB2(rp, 16, src0, src1);
LD_UB2(pp, 16, src4, src5);
ADD2(src0, src4, src1, src5, src0, src1);
ST_UB2(src0, src1, rp, 16);
rp += 32;
}
else if (cnt16 || cnt)
{
src0 = LD_UB(rp);
src4 = LD_UB(pp);
pp += 16;
src0 += src4;
ST_UB(src0, rp);
rp += 16;
}
}
}
void png_read_filter_row_sub4_msa(png_row_infop row_info, png_bytep row,
png_const_bytep prev_row)
{
460
461
size_t count;
size_t istop = row_info->rowbytes;
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
png_bytep src = row;
png_bytep nxt = row + 4;
int32_t inp0;
v16u8 src0, src1, src2, src3, src4;
v16u8 dst0, dst1;
v16u8 zero = { 0 };
istop -= 4;
inp0 = LW(src);
src += 4;
src0 = (v16u8) __msa_insert_w((v4i32) zero, 0, inp0);
for (count = 0; count < istop; count += 16)
{
src1 = LD_UB(src);
src += 16;
src2 = (v16u8) __msa_sldi_b((v16i8) zero, (v16i8) src1, 4);
src3 = (v16u8) __msa_sldi_b((v16i8) zero, (v16i8) src1, 8);
src4 = (v16u8) __msa_sldi_b((v16i8) zero, (v16i8) src1, 12);
src1 += src0;
src2 += src1;
src3 += src2;
src4 += src3;
src0 = src4;
ILVEV_W2_UB(src1, src2, src3, src4, dst0, dst1);
dst0 = (v16u8) __msa_pckev_d((v2i64) dst1, (v2i64) dst0);
ST_UB(dst0, nxt);
nxt += 16;
}
}
void png_read_filter_row_sub3_msa(png_row_infop row_info, png_bytep row,
png_const_bytep prev_row)
{
499
500
size_t count;
size_t istop = row_info->rowbytes;
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
png_bytep src = row;
png_bytep nxt = row + 3;
int64_t out0;
int32_t inp0, out1;
v16u8 src0, src1, src2, src3, src4, dst0, dst1;
v16u8 zero = { 0 };
v16i8 mask0 = { 0, 1, 2, 16, 17, 18, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 };
v16i8 mask1 = { 0, 1, 2, 3, 4, 5, 16, 17, 18, 19, 20, 21, 0, 0, 0, 0 };
istop -= 3;
inp0 = LW(src);
src += 3;
src0 = (v16u8) __msa_insert_w((v4i32) zero, 0, inp0);
for (count = 0; count < istop; count += 12)
{
src1 = LD_UB(src);
src += 12;
src2 = (v16u8) __msa_sldi_b((v16i8) zero, (v16i8) src1, 3);
src3 = (v16u8) __msa_sldi_b((v16i8) zero, (v16i8) src1, 6);
src4 = (v16u8) __msa_sldi_b((v16i8) zero, (v16i8) src1, 9);
src1 += src0;
src2 += src1;
src3 += src2;
src4 += src3;
src0 = src4;
VSHF_B2_UB(src1, src2, src3, src4, mask0, mask0, dst0, dst1);
dst0 = (v16u8) __msa_vshf_b(mask1, (v16i8) dst1, (v16i8) dst0);
out0 = __msa_copy_s_d((v2i64) dst0, 0);
out1 = __msa_copy_s_w((v4i32) dst0, 2);
SD(out0, nxt);
nxt += 8;
SW(out1, nxt);
nxt += 4;
}
}
void png_read_filter_row_avg4_msa(png_row_infop row_info, png_bytep row,
png_const_bytep prev_row)
{
545
546
547
png_bytep src = row;
png_bytep nxt = row;
png_const_bytep pp = prev_row;
548
size_t istop = row_info->rowbytes - 4;
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
int32_t inp0, inp1, out0;
v16u8 src0, src1, src2, src3, src4, src5, src6, src7, src8, src9, dst0, dst1;
v16u8 zero = { 0 };
inp0 = LW(pp);
pp += 4;
inp1 = LW(src);
src += 4;
src0 = (v16u8) __msa_insert_w((v4i32) zero, 0, inp0);
src1 = (v16u8) __msa_insert_w((v4i32) zero, 0, inp1);
src0 = (v16u8) MSA_SRLI_B(src0, 1);
src1 += src0;
out0 = __msa_copy_s_w((v4i32) src1, 0);
SW(out0, nxt);
nxt += 4;
for (i = 0; i < istop; i += 16)
{
src2 = LD_UB(pp);
pp += 16;
src6 = LD_UB(src);
src += 16;
SLDI_B2_0_UB(src2, src6, src3, src7, 4);
SLDI_B2_0_UB(src2, src6, src4, src8, 8);
SLDI_B2_0_UB(src2, src6, src5, src9, 12);
src2 = __msa_ave_u_b(src2, src1);
src6 += src2;
src3 = __msa_ave_u_b(src3, src6);
src7 += src3;
src4 = __msa_ave_u_b(src4, src7);
src8 += src4;
src5 = __msa_ave_u_b(src5, src8);
src9 += src5;
src1 = src9;
ILVEV_W2_UB(src6, src7, src8, src9, dst0, dst1);
dst0 = (v16u8) __msa_pckev_d((v2i64) dst1, (v2i64) dst0);
ST_UB(dst0, nxt);
nxt += 16;
}
}
void png_read_filter_row_avg3_msa(png_row_infop row_info, png_bytep row,
png_const_bytep prev_row)
{
596
597
598
png_bytep src = row;
png_bytep nxt = row;
png_const_bytep pp = prev_row;
599
size_t istop = row_info->rowbytes - 3;
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
int64_t out0;
int32_t inp0, inp1, out1;
int16_t out2;
v16u8 src0, src1, src2, src3, src4, src5, src6, src7, src8, src9, dst0, dst1;
v16u8 zero = { 0 };
v16i8 mask0 = { 0, 1, 2, 16, 17, 18, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 };
v16i8 mask1 = { 0, 1, 2, 3, 4, 5, 16, 17, 18, 19, 20, 21, 0, 0, 0, 0 };
inp0 = LW(pp);
pp += 3;
inp1 = LW(src);
src += 3;
src0 = (v16u8) __msa_insert_w((v4i32) zero, 0, inp0);
src1 = (v16u8) __msa_insert_w((v4i32) zero, 0, inp1);
src0 = (v16u8) MSA_SRLI_B(src0, 1);
src1 += src0;
out2 = __msa_copy_s_h((v8i16) src1, 0);
SH(out2, nxt);
nxt += 2;
nxt[0] = src1[2];
nxt++;
for (i = 0; i < istop; i += 12)
{
src2 = LD_UB(pp);
pp += 12;
src6 = LD_UB(src);
src += 12;
SLDI_B2_0_UB(src2, src6, src3, src7, 3);
SLDI_B2_0_UB(src2, src6, src4, src8, 6);
SLDI_B2_0_UB(src2, src6, src5, src9, 9);
src2 = __msa_ave_u_b(src2, src1);
src6 += src2;
src3 = __msa_ave_u_b(src3, src6);
src7 += src3;
src4 = __msa_ave_u_b(src4, src7);
src8 += src4;
src5 = __msa_ave_u_b(src5, src8);
src9 += src5;
src1 = src9;
VSHF_B2_UB(src6, src7, src8, src9, mask0, mask0, dst0, dst1);
dst0 = (v16u8) __msa_vshf_b(mask1, (v16i8) dst1, (v16i8) dst0);
out0 = __msa_copy_s_d((v2i64) dst0, 0);
out1 = __msa_copy_s_w((v4i32) dst0, 2);
SD(out0, nxt);
nxt += 8;
SW(out1, nxt);
nxt += 4;
}
}
void png_read_filter_row_paeth4_msa(png_row_infop row_info,
png_bytep row,
png_const_bytep prev_row)
{
int32_t count, rp_end;
png_bytep nxt;
png_const_bytep prev_nxt;
int32_t inp0, inp1, res0;
v16u8 src0, src1, src2, src3, src4, src5, src6, src7, src8, src9;
v16u8 src10, src11, src12, src13, dst0, dst1;
v8i16 vec0, vec1, vec2;
v16u8 zero = { 0 };
nxt = row;
prev_nxt = prev_row;
inp0 = LW(nxt);
inp1 = LW(prev_nxt);
prev_nxt += 4;
src0 = (v16u8) __msa_insert_w((v4i32) zero, 0, inp0);
src1 = (v16u8) __msa_insert_w((v4i32) zero, 0, inp1);
src1 += src0;
res0 = __msa_copy_s_w((v4i32) src1, 0);
SW(res0, nxt);
nxt += 4;
/* Remainder */
rp_end = row_info->rowbytes - 4;
for (count = 0; count < rp_end; count += 16)
{
src2 = LD_UB(prev_nxt);
prev_nxt += 16;
src6 = LD_UB(prev_row);
prev_row += 16;
src10 = LD_UB(nxt);
SLDI_B3_0_UB(src2, src6, src10, src3, src7, src11, 4);
SLDI_B3_0_UB(src2, src6, src10, src4, src8, src12, 8);
SLDI_B3_0_UB(src2, src6, src10, src5, src9, src13, 12);
ILVR_B2_SH(src2, src6, src1, src6, vec0, vec1);
HSUB_UB2_SH(vec0, vec1, vec0, vec1);
vec2 = vec0 + vec1;
ADD_ABS_H3_SH(vec0, vec1, vec2, vec0, vec1, vec2);
CMP_AND_SELECT(vec0, vec1, vec2, src1, src2, src6, src10);
ILVR_B2_SH(src3, src7, src10, src7, vec0, vec1);
HSUB_UB2_SH(vec0, vec1, vec0, vec1);
vec2 = vec0 + vec1;
ADD_ABS_H3_SH(vec0, vec1, vec2, vec0, vec1, vec2);
CMP_AND_SELECT(vec0, vec1, vec2, src10, src3, src7, src11);
ILVR_B2_SH(src4, src8, src11, src8, vec0, vec1);
HSUB_UB2_SH(vec0, vec1, vec0, vec1);
vec2 = vec0 + vec1;
ADD_ABS_H3_SH(vec0, vec1, vec2, vec0, vec1, vec2);
CMP_AND_SELECT(vec0, vec1, vec2, src11, src4, src8, src12);
ILVR_B2_SH(src5, src9, src12, src9, vec0, vec1);
HSUB_UB2_SH(vec0, vec1, vec0, vec1);
vec2 = vec0 + vec1;
ADD_ABS_H3_SH(vec0, vec1, vec2, vec0, vec1, vec2);
CMP_AND_SELECT(vec0, vec1, vec2, src12, src5, src9, src13);
src1 = src13;
ILVEV_W2_UB(src10, src11, src12, src1, dst0, dst1);
dst0 = (v16u8) __msa_pckev_d((v2i64) dst1, (v2i64) dst0);
ST_UB(dst0, nxt);
nxt += 16;
}
}
void png_read_filter_row_paeth3_msa(png_row_infop row_info,
png_bytep row,
png_const_bytep prev_row)
{
int32_t count, rp_end;
png_bytep nxt;
png_const_bytep prev_nxt;
int64_t out0;
int32_t inp0, inp1, out1;
int16_t out2;
v16u8 src0, src1, src2, src3, src4, src5, src6, src7, src8, src9, dst0, dst1;
v16u8 src10, src11, src12, src13;
v8i16 vec0, vec1, vec2;
v16u8 zero = { 0 };
v16i8 mask0 = { 0, 1, 2, 16, 17, 18, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 };
v16i8 mask1 = { 0, 1, 2, 3, 4, 5, 16, 17, 18, 19, 20, 21, 0, 0, 0, 0 };
nxt = row;
prev_nxt = prev_row;
inp0 = LW(nxt);
inp1 = LW(prev_nxt);
prev_nxt += 3;
src0 = (v16u8) __msa_insert_w((v4i32) zero, 0, inp0);
src1 = (v16u8) __msa_insert_w((v4i32) zero, 0, inp1);
src1 += src0;
out2 = __msa_copy_s_h((v8i16) src1, 0);
SH(out2, nxt);
nxt += 2;
nxt[0] = src1[2];
nxt++;
/* Remainder */
rp_end = row_info->rowbytes - 3;
for (count = 0; count < rp_end; count += 12)
{
src2 = LD_UB(prev_nxt);
prev_nxt += 12;
src6 = LD_UB(prev_row);
prev_row += 12;
src10 = LD_UB(nxt);
SLDI_B3_0_UB(src2, src6, src10, src3, src7, src11, 3);
SLDI_B3_0_UB(src2, src6, src10, src4, src8, src12, 6);
SLDI_B3_0_UB(src2, src6, src10, src5, src9, src13, 9);
ILVR_B2_SH(src2, src6, src1, src6, vec0, vec1);
HSUB_UB2_SH(vec0, vec1, vec0, vec1);
vec2 = vec0 + vec1;
ADD_ABS_H3_SH(vec0, vec1, vec2, vec0, vec1, vec2);
CMP_AND_SELECT(vec0, vec1, vec2, src1, src2, src6, src10);
ILVR_B2_SH(src3, src7, src10, src7, vec0, vec1);
HSUB_UB2_SH(vec0, vec1, vec0, vec1);
vec2 = vec0 + vec1;
ADD_ABS_H3_SH(vec0, vec1, vec2, vec0, vec1, vec2);
CMP_AND_SELECT(vec0, vec1, vec2, src10, src3, src7, src11);
ILVR_B2_SH(src4, src8, src11, src8, vec0, vec1);
HSUB_UB2_SH(vec0, vec1, vec0, vec1);
vec2 = vec0 + vec1;
ADD_ABS_H3_SH(vec0, vec1, vec2, vec0, vec1, vec2);
CMP_AND_SELECT(vec0, vec1, vec2, src11, src4, src8, src12);
ILVR_B2_SH(src5, src9, src12, src9, vec0, vec1);
HSUB_UB2_SH(vec0, vec1, vec0, vec1);
vec2 = vec0 + vec1;
ADD_ABS_H3_SH(vec0, vec1, vec2, vec0, vec1, vec2);
CMP_AND_SELECT(vec0, vec1, vec2, src12, src5, src9, src13);
src1 = src13;
VSHF_B2_UB(src10, src11, src12, src13, mask0, mask0, dst0, dst1);
dst0 = (v16u8) __msa_vshf_b(mask1, (v16i8) dst1, (v16i8) dst0);
out0 = __msa_copy_s_d((v2i64) dst0, 0);
out1 = __msa_copy_s_w((v4i32) dst0, 2);
SD(out0, nxt);
nxt += 8;
SW(out1, nxt);
nxt += 4;
}
}
#endif /* PNG_MIPS_MSA_OPT > 0 */
#endif /* PNG_MIPS_MSA_IMPLEMENTATION == 1 (intrinsics) */
#endif /* READ */