112 |
unsigned int cpu; |
unsigned int cpu; |
113 |
} CPU; |
} CPU; |
114 |
|
|
115 |
CPU cpu_list[] = |
CPU cpu_list[] = { |
116 |
{ { "PLAINC", 0 } |
{ "PLAINC ", 0 }, |
117 |
#ifdef ARCH_IS_IA32 |
#ifdef ARCH_IS_IA32 |
118 |
, { "MMX ", XVID_CPU_MMX } |
{ "MMX ", XVID_CPU_MMX }, |
119 |
, { "MMXEXT", XVID_CPU_MMXEXT | XVID_CPU_MMX } |
{ "MMXEXT ", XVID_CPU_MMXEXT | XVID_CPU_MMX }, |
120 |
, { "SSE2 ", XVID_CPU_SSE2 | XVID_CPU_MMX } |
{ "SSE2 ", XVID_CPU_SSE2 | XVID_CPU_MMX }, |
121 |
, { "3DNOW ", XVID_CPU_3DNOW } |
{ "3DNOW ", XVID_CPU_3DNOW }, |
122 |
, { "3DNOWE", XVID_CPU_3DNOW | XVID_CPU_3DNOWEXT } |
{ "3DNOWE ", XVID_CPU_3DNOW | XVID_CPU_3DNOWEXT }, |
123 |
#endif |
#endif |
124 |
#ifdef ARCH_IS_PPC |
#ifdef ARCH_IS_PPC |
125 |
, { "ALTIVEC", XVID_CPU_ALTIVEC } |
{ "ALTIVEC", XVID_CPU_ALTIVEC }, |
126 |
#endif |
#endif |
127 |
#ifdef ARCH_IS_X86_64 |
#ifdef ARCH_IS_X86_64 |
128 |
, { "X86_64", XVID_CPU_ASM} |
{ "X86_64 ", XVID_CPU_ASM}, |
129 |
#endif |
#endif |
130 |
//, { "IA64 ", XVID_CPU_IA64 } |
#ifdef ARCH_IS_IA64 |
131 |
//, { "TSC ", XVID_CPU_TSC } |
// { "IA64 ", XVID_CPU_IA64 }, |
|
, { 0, 0 } }; |
|
|
|
|
|
CPU cpu_short_list[] = |
|
|
{ { "PLAINC", 0 } |
|
|
#ifdef ARCH_IS_IA32 |
|
|
, { "MMX ", XVID_CPU_MMX } |
|
|
//, { "MMXEXT", XVID_CPU_MMXEXT | XVID_CPU_MMX } |
|
|
#endif |
|
|
//, { "IA64 ", XVID_CPU_IA64 } |
|
|
, { 0, 0 } }; |
|
|
|
|
|
CPU cpu_short_list2[] = |
|
|
{ { "PLAINC", 0 } |
|
|
#ifdef ARCH_IS_IA32 |
|
|
, { "MMX ", XVID_CPU_MMX } |
|
|
, { "SSE2 ", XVID_CPU_SSE2 | XVID_CPU_MMX } |
|
132 |
#endif |
#endif |
133 |
, { 0, 0 } }; |
// { "TSC ", XVID_CPU_TSC }, |
134 |
|
{ 0, 0 } |
135 |
|
}; |
136 |
|
|
137 |
|
|
138 |
int init_cpu(CPU *cpu) |
int init_cpu(CPU *cpu) |
260 |
return crc; |
return crc; |
261 |
} |
} |
262 |
|
|
263 |
|
void byte_swap(uint8_t *mem, int len, int element_size) { |
264 |
|
#ifdef ARCH_IS_BIG_ENDIAN |
265 |
|
int i; |
266 |
|
|
267 |
|
if(element_size == 1) { |
268 |
|
/* No need to swap */ |
269 |
|
} else if(element_size == 2) { |
270 |
|
uint8_t temp[2]; |
271 |
|
|
272 |
|
for(i=0; i < (len/2); i++ ) { |
273 |
|
temp[0] = mem[0]; |
274 |
|
temp[1] = mem[1]; |
275 |
|
mem[0] = temp[1]; |
276 |
|
mem[1] = temp[0]; |
277 |
|
|
278 |
|
mem += 2; |
279 |
|
} |
280 |
|
} else if(element_size == 4) { |
281 |
|
uint8_t temp[4]; |
282 |
|
|
283 |
|
for(i=0; i < (len/4); i++ ) { |
284 |
|
temp[0] = mem[0]; |
285 |
|
temp[1] = mem[1]; |
286 |
|
temp[2] = mem[2]; |
287 |
|
temp[3] = mem[3]; |
288 |
|
mem[0] = temp[3]; |
289 |
|
mem[1] = temp[2]; |
290 |
|
mem[2] = temp[1]; |
291 |
|
mem[3] = temp[0]; |
292 |
|
|
293 |
|
mem += 4; |
294 |
|
} |
295 |
|
} else { |
296 |
|
printf("ERROR: byte_swap unsupported element_size(%u)\n", element_size); |
297 |
|
} |
298 |
|
#endif |
299 |
|
} |
300 |
|
|
301 |
/********************************************************************* |
/********************************************************************* |
302 |
* test DCT |
* test DCT |
303 |
*********************************************************************/ |
*********************************************************************/ |
518 |
(iCrc!=8107)?"| ERROR": "" ); |
(iCrc!=8107)?"| ERROR": "" ); |
519 |
#endif |
#endif |
520 |
|
|
521 |
|
/* New functions for field prediction by CK 1.10.2005 */ |
522 |
|
#pragma NEW8X4 |
523 |
|
TEST_MB(interpolate8x4_halfpel_h, 0); |
524 |
|
printf("%s - interpfield-h -round0 %.3f usec crc32=0x%08x %s\n", |
525 |
|
cpu->name, t, iCrc, |
526 |
|
(iCrc!=0x9538d6df)?"| ERROR": "" ); |
527 |
|
|
528 |
|
TEST_MB(interpolate8x4_halfpel_h, 1); |
529 |
|
printf("%s - round1 %.3f usec crc32=0x%08x %s\n", |
530 |
|
cpu->name, t, iCrc, |
531 |
|
(iCrc!=0xde5f1db4)?"| ERROR": "" ); |
532 |
|
|
533 |
|
|
534 |
|
TEST_MB(interpolate8x4_halfpel_v, 0); |
535 |
|
printf("%s - interpfield- v-round0 %.3f usec crc32=0x%08x %s\n", |
536 |
|
cpu->name, t, iCrc, |
537 |
|
(iCrc!=0xea5a69ef)?"| ERROR": "" ); |
538 |
|
|
539 |
|
TEST_MB(interpolate8x4_halfpel_v, 1); |
540 |
|
printf("%s - round1 %.3f usec crc32=0x%08x %s\n", |
541 |
|
cpu->name, t, iCrc, |
542 |
|
(iCrc!=0x4f10ec0f)?"| ERROR": "" ); |
543 |
|
|
544 |
|
|
545 |
|
TEST_MB(interpolate8x4_halfpel_hv, 0); |
546 |
|
printf("%s - interpfield-hv-round0 %.3f usec crc32=0x%08x %s\n", |
547 |
|
cpu->name, t, iCrc, |
548 |
|
(iCrc!=0xf97ee367)?"| ERROR": "" ); |
549 |
|
|
550 |
|
TEST_MB(interpolate8x4_halfpel_hv, 1); |
551 |
|
printf("%s - round1 %.3f usec crc32=0x%08x %s\n", |
552 |
|
cpu->name, t, iCrc, |
553 |
|
(iCrc!=0xb6a9f581)?"| ERROR": "" ); |
554 |
|
/* End of 8x4 functions */ |
555 |
|
|
556 |
printf( " --- \n" ); |
printf( " --- \n" ); |
557 |
} |
} |
558 |
} |
} |
559 |
|
|
560 |
|
#undef ENTER |
561 |
|
#undef LEAVE |
562 |
|
#undef TEST_MB |
563 |
|
#undef TEST_MB2 |
564 |
|
|
565 |
/********************************************************************* |
/********************************************************************* |
566 |
* test transfer |
* test transfer |
567 |
*********************************************************************/ |
*********************************************************************/ |
591 |
} \ |
} \ |
592 |
emms(); \ |
emms(); \ |
593 |
t = (gettime_usec()-t -overhead) / nb_tests;\ |
t = (gettime_usec()-t -overhead) / nb_tests;\ |
594 |
|
byte_swap((uint8_t*)(DST), 8*32*sizeof((DST)[0]), sizeof((DST)[0])); \ |
595 |
s = calc_crc((uint8_t*)(DST), 8*32*sizeof((DST)[0]), CRC32_INITIAL) |
s = calc_crc((uint8_t*)(DST), 8*32*sizeof((DST)[0]), CRC32_INITIAL) |
596 |
|
|
597 |
#define TEST_TRANSFER(FUNC, DST, SRC) \ |
#define TEST_TRANSFER(FUNC, DST, SRC) \ |
618 |
} \ |
} \ |
619 |
emms(); \ |
emms(); \ |
620 |
t = (gettime_usec()-t -overhead) / nb_tests;\ |
t = (gettime_usec()-t -overhead) / nb_tests;\ |
621 |
|
byte_swap((uint8_t*)(DST), 8*32*sizeof((DST)[0]), sizeof((DST)[0])); \ |
622 |
s = calc_crc((uint8_t*)(DST), 8*32*sizeof((DST)[0]), CRC32_INITIAL) |
s = calc_crc((uint8_t*)(DST), 8*32*sizeof((DST)[0]), CRC32_INITIAL) |
623 |
|
|
624 |
#define TEST_TRANSFER2(FUNC, DST, SRC, R1) \ |
#define TEST_TRANSFER2(FUNC, DST, SRC, R1) \ |
665 |
cpu->name, t, s, |
cpu->name, t, s, |
666 |
(s!=0xee7ccbb4)?"| ERROR": ""); |
(s!=0xee7ccbb4)?"| ERROR": ""); |
667 |
|
|
668 |
|
/* New functions for field prediction by CK 1.10.2005 */ |
669 |
|
#pragma NEW8X4 |
670 |
|
TEST_TRANSFER(transfer8x4_copy, Dst8, Src8); |
671 |
|
printf("%s - 8to4 %.3f usec crc32=0x%08x %s\n", |
672 |
|
cpu->name, t, s, |
673 |
|
(s!=0xbb9c3db5)?"| ERROR": ""); |
674 |
|
/* End of new functions */ |
675 |
|
|
676 |
TEST_TRANSFER(transfer8x8_copy, Dst8, Src8); |
TEST_TRANSFER(transfer8x8_copy, Dst8, Src8); |
677 |
printf("%s - 8to8 %.3f usec crc32=0x%08x %s\n", |
printf("%s - 8to8 %.3f usec crc32=0x%08x %s\n", |
678 |
cpu->name, t, s, |
cpu->name, t, s, |
716 |
for(q=1; q<=max_Q; ++q) { \ |
for(q=1; q<=max_Q; ++q) { \ |
717 |
for(tst=0; tst<nb_tests; ++tst) \ |
for(tst=0; tst<nb_tests; ++tst) \ |
718 |
(FUNC)((DST), (SRC), q, mpeg_quant_matrices); \ |
(FUNC)((DST), (SRC), q, mpeg_quant_matrices); \ |
719 |
s = calc_crc((uint8_t*)(DST), 64*sizeof(int16_t), s); \ |
byte_swap((uint8_t*)(DST), 64*sizeof((DST)[0]), sizeof((DST)[0])); \ |
720 |
|
s = calc_crc((uint8_t*)(DST), 64*sizeof((DST)[0]), s); \ |
721 |
} \ |
} \ |
722 |
emms(); \ |
emms(); \ |
723 |
} \ |
} \ |
732 |
for(q=1; q<=max_Q; ++q) { \ |
for(q=1; q<=max_Q; ++q) { \ |
733 |
for(tst=0; tst<nb_tests; ++tst) \ |
for(tst=0; tst<nb_tests; ++tst) \ |
734 |
(FUNC)((DST), (SRC), q, q, mpeg_quant_matrices); \ |
(FUNC)((DST), (SRC), q, q, mpeg_quant_matrices); \ |
735 |
s = calc_crc((uint8_t*)(DST), 64*sizeof(int16_t), s); \ |
byte_swap((uint8_t*)(DST), 64*sizeof((DST)[0]), sizeof((DST)[0])); \ |
736 |
|
s = calc_crc((uint8_t*)(DST), 64*sizeof((DST)[0]), s); \ |
737 |
} \ |
} \ |
738 |
emms(); \ |
emms(); \ |
739 |
} \ |
} \ |
953 |
* test non-zero AC counting |
* test non-zero AC counting |
954 |
*********************************************************************/ |
*********************************************************************/ |
955 |
|
|
956 |
#define TEST_CBP(FUNC, SRC) \ |
#define TEST_CBP(FUNC, SRC, NB) \ |
957 |
t = gettime_usec(); \ |
t = gettime_usec(); \ |
958 |
emms(); \ |
emms(); \ |
959 |
for(tst=0; tst<nb_tests; ++tst) { \ |
for(tst=0; tst<NB; ++tst) { \ |
960 |
cbp = (FUNC)((SRC)); \ |
cbp = (FUNC)((SRC)); \ |
961 |
} \ |
} \ |
962 |
emms(); \ |
emms(); \ |
965 |
void test_cbp() |
void test_cbp() |
966 |
{ |
{ |
967 |
const int nb_tests = 10000*speed_ref; |
const int nb_tests = 10000*speed_ref; |
968 |
int i; |
int i, n, m; |
969 |
CPU *cpu; |
CPU *cpu; |
970 |
DECLARE_ALIGNED_MATRIX(Src1, 6, 64, int16_t, 16); |
DECLARE_ALIGNED_MATRIX(Src1, 6, 64, int16_t, 16); |
971 |
DECLARE_ALIGNED_MATRIX(Src2, 6, 64, int16_t, 16); |
DECLARE_ALIGNED_MATRIX(Src2, 6, 64, int16_t, 16); |
972 |
DECLARE_ALIGNED_MATRIX(Src3, 6, 64, int16_t, 16); |
DECLARE_ALIGNED_MATRIX(Src3, 6, 64, int16_t, 16); |
973 |
DECLARE_ALIGNED_MATRIX(Src4, 6, 64, int16_t, 16); |
DECLARE_ALIGNED_MATRIX(Src4, 6, 64, int16_t, 16); |
974 |
|
DECLARE_ALIGNED_MATRIX(Src5, 6, 64, int16_t, 16); |
975 |
|
|
976 |
printf( "\n ===== test cbp =====\n" ); |
printf( "\n ===== test cbp =====\n" ); |
977 |
|
|
980 |
Src2[i] = (i<3*64); /* half-full */ |
Src2[i] = (i<3*64); /* half-full */ |
981 |
Src3[i] = ((i+32)>3*64); |
Src3[i] = ((i+32)>3*64); |
982 |
Src4[i] = (i==(3*64+2) || i==(5*64+9)); |
Src4[i] = (i==(3*64+2) || i==(5*64+9)); |
983 |
|
Src5[i] = ieee_rand(0,1) ? -1 : 1; /* +/- test */ |
984 |
} |
} |
985 |
|
|
986 |
for(cpu = cpu_list; cpu->name!=0; ++cpu) |
for(cpu = cpu_list; cpu->name!=0; ++cpu) |
991 |
if (!init_cpu(cpu)) |
if (!init_cpu(cpu)) |
992 |
continue; |
continue; |
993 |
|
|
994 |
TEST_CBP(calc_cbp, Src1); |
TEST_CBP(calc_cbp, Src1, nb_tests); |
995 |
printf("%s - calc_cbp#1 %.3f usec cbp=0x%02x %s\n", |
printf("%s - calc_cbp#1 %.3f usec cbp=0x%02x %s\n", |
996 |
cpu->name, t, cbp, (cbp!=0x15)?"| ERROR": ""); |
cpu->name, t, cbp, (cbp!=0x15)?"| ERROR": ""); |
997 |
TEST_CBP(calc_cbp, Src2); |
TEST_CBP(calc_cbp, Src2, nb_tests); |
998 |
printf("%s - calc_cbp#2 %.3f usec cbp=0x%02x %s\n", |
printf("%s - calc_cbp#2 %.3f usec cbp=0x%02x %s\n", |
999 |
cpu->name, t, cbp, (cbp!=0x38)?"| ERROR": ""); |
cpu->name, t, cbp, (cbp!=0x38)?"| ERROR": ""); |
1000 |
TEST_CBP(calc_cbp, Src3); |
TEST_CBP(calc_cbp, Src3, nb_tests); |
1001 |
printf("%s - calc_cbp#3 %.3f usec cbp=0x%02x %s\n", |
printf("%s - calc_cbp#3 %.3f usec cbp=0x%02x %s\n", |
1002 |
cpu->name, t, cbp, (cbp!=0x0f)?"| ERROR": "" ); |
cpu->name, t, cbp, (cbp!=0x0f)?"| ERROR": "" ); |
1003 |
TEST_CBP(calc_cbp, Src4); |
TEST_CBP(calc_cbp, Src4, nb_tests); |
1004 |
printf("%s - calc_cbp#4 %.3f usec cbp=0x%02x %s\n", |
printf("%s - calc_cbp#4 %.3f usec cbp=0x%02x %s\n", |
1005 |
cpu->name, t, cbp, (cbp!=0x05)?"| ERROR": "" ); |
cpu->name, t, cbp, (cbp!=0x05)?"| ERROR": "" ); |
1006 |
|
TEST_CBP(calc_cbp, Src5, nb_tests); |
1007 |
|
printf("%s - calc_cbp#4 %.3f usec cbp=0x%02x %s\n", |
1008 |
|
cpu->name, t, cbp, (cbp!=0x3f)?"| ERROR": "" ); |
1009 |
printf( " --- \n" ); |
printf( " --- \n" ); |
1010 |
} |
} |
1011 |
|
|
1012 |
|
for(cpu = cpu_list; cpu->name!=0; ++cpu) /* bench suggested by Carlo (carlo dot bramix at libero dot it) */ |
1013 |
|
{ |
1014 |
|
double t; |
1015 |
|
int tst, cbp, err; |
1016 |
|
|
1017 |
|
if (!init_cpu(cpu)) |
1018 |
|
continue; |
1019 |
|
|
1020 |
|
err = 0; |
1021 |
|
for(n=0; n<6; ++n) |
1022 |
|
{ |
1023 |
|
for(m=0; m<64; ++m) |
1024 |
|
{ |
1025 |
|
for(i=0; i<6*64; ++i) |
1026 |
|
Src1[i] = (i== (m + n*64)); |
1027 |
|
|
1028 |
|
TEST_CBP(calc_cbp, Src1, 1); |
1029 |
|
if (cbp!= (((m!=0)<<(5-n)))) |
1030 |
|
{ |
1031 |
|
printf( "%s - calc_cbp#5: ERROR at pos %d / %d!\n", cpu->name, n, m); |
1032 |
|
err = 1; |
1033 |
|
break; |
1034 |
|
} |
1035 |
|
} |
1036 |
|
} |
1037 |
|
if (!err) |
1038 |
|
printf( " %s - calc_cbp#5 : OK\n", cpu->name ); |
1039 |
|
|
1040 |
|
} |
1041 |
} |
} |
1042 |
|
|
1043 |
/********************************************************************* |
/********************************************************************* |
1500 |
xframe.bitstream = buf + pos; |
xframe.bitstream = buf + pos; |
1501 |
xframe.length = buf_size - pos; |
xframe.length = buf_size - pos; |
1502 |
xframe.output.plane[0] = (uint8_t*)(((size_t)yuv_out + 15) & ~15); |
xframe.output.plane[0] = (uint8_t*)(((size_t)yuv_out + 15) & ~15); |
1503 |
xframe.output.plane[1] = xframe.output.plane[0] + bps*height; |
xframe.output.plane[1] = (uint8_t*)xframe.output.plane[0] + bps*height; |
1504 |
xframe.output.plane[2] = xframe.output.plane[1] + bps/2; |
xframe.output.plane[2] = (uint8_t*)xframe.output.plane[1] + bps/2; |
1505 |
xframe.output.stride[0] = bps; |
xframe.output.stride[0] = bps; |
1506 |
xframe.output.stride[1] = bps; |
xframe.output.stride[1] = bps; |
1507 |
xframe.output.stride[2] = bps; |
xframe.output.stride[2] = bps; |
1519 |
nb++; |
nb++; |
1520 |
|
|
1521 |
for(y=0; y<height/2; ++y) { |
for(y=0; y<height/2; ++y) { |
1522 |
chksum = calc_crc(xframe.output.plane[0] + (2*y+0)*bps, width, chksum); |
chksum = calc_crc((uint8_t*)xframe.output.plane[0] + (2*y+0)*bps, width, chksum); |
1523 |
chksum = calc_crc(xframe.output.plane[0] + (2*y+1)*bps, width, chksum); |
chksum = calc_crc((uint8_t*)xframe.output.plane[0] + (2*y+1)*bps, width, chksum); |
1524 |
chksum = calc_crc(xframe.output.plane[1] + y*bps, width/2, chksum); |
chksum = calc_crc((uint8_t*)xframe.output.plane[1] + y*bps, width/2, chksum); |
1525 |
chksum = calc_crc(xframe.output.plane[2] + y*bps, width/2, chksum); |
chksum = calc_crc((uint8_t*)xframe.output.plane[2] + y*bps, width/2, chksum); |
1526 |
} |
} |
1527 |
if (pos==buf_size) |
if (pos==buf_size) |
1528 |
break; |
break; |
1705 |
} |
} |
1706 |
#endif |
#endif |
1707 |
} |
} |
1708 |
|
|
1709 |
|
/********************************************************************* |
1710 |
|
* test some YUV func |
1711 |
|
*********************************************************************/ |
1712 |
|
|
1713 |
|
#define ENTER \ |
1714 |
|
for(i=0; i<(int)sizeof(Dst0); ++i) Dst0[0][i] = 0; \ |
1715 |
|
t = gettime_usec(); \ |
1716 |
|
emms(); |
1717 |
|
|
1718 |
|
#define LEAVE \ |
1719 |
|
emms(); \ |
1720 |
|
t = (gettime_usec() - t) / nb_tests; \ |
1721 |
|
iCrc = calc_crc((uint8_t*)Dst0, sizeof(Dst0), CRC32_INITIAL) |
1722 |
|
|
1723 |
|
#define TEST_YUYV(FUNC, S) \ |
1724 |
|
ENTER \ |
1725 |
|
for(tst=0; tst<nb_tests; ++tst) (FUNC)(Dst0[0], S*WIDTH, Src0[0], Src0[1], Src0[2], WIDTH, WIDTH/2, WIDTH, HEIGHT, 0); \ |
1726 |
|
LEAVE |
1727 |
|
|
1728 |
|
#define WIDTH 128 |
1729 |
|
#define HEIGHT 32 |
1730 |
|
void test_yuv() |
1731 |
|
{ |
1732 |
|
const int nb_tests = 200*speed_ref; |
1733 |
|
CPU *cpu; |
1734 |
|
uint8_t Src0[3][WIDTH*HEIGHT]; |
1735 |
|
uint8_t Dst0[4][WIDTH*HEIGHT]; |
1736 |
|
int i, j; |
1737 |
|
double t; |
1738 |
|
int tst, iCrc; |
1739 |
|
|
1740 |
|
colorspace_init(); |
1741 |
|
ieee_reseed(1); |
1742 |
|
for(i=0; i<(int)sizeof(Src0); ++i) Src0[0][i] = ieee_rand(0,255); |
1743 |
|
for(i=0; i<(int)sizeof(Dst0); ++i) Dst0[0][i] = 0x5a; |
1744 |
|
|
1745 |
|
printf( "\n === test YUV ===\n" ); |
1746 |
|
|
1747 |
|
init_cpu(&cpu_list[0]); |
1748 |
|
TEST_YUYV(yv12_to_yuyv_c, 4); |
1749 |
|
printf(" yv12_to_yuyv_c %.3f usec crc32=0x%08x %s\n", |
1750 |
|
t, iCrc, (iCrc!=0xeb1a0b0a)?"| ERROR": "" ); |
1751 |
|
TEST_YUYV(yv12_to_uyvy_c, 4); |
1752 |
|
printf(" yv12_to_uyvy_c %.3f usec crc32=0x%08x %s\n", |
1753 |
|
t, iCrc, (iCrc!=0x6e82f55b)?"| ERROR": "" ); |
1754 |
|
|
1755 |
|
#ifdef ARCH_IS_IA32 |
1756 |
|
init_cpu(&cpu_list[1]); |
1757 |
|
TEST_YUYV(yv12_to_yuyv_mmx, 4); |
1758 |
|
printf(" yv12_to_yuyv_mmx %.3f usec crc32=0x%08x %s\n", |
1759 |
|
t, iCrc, (iCrc!=0xeb1a0b0a)?"| ERROR": "" ); |
1760 |
|
|
1761 |
|
TEST_YUYV(yv12_to_uyvy_mmx, 4); |
1762 |
|
printf(" yv12_to_uyvy_mmx %.3f usec crc32=0x%08x %s\n", |
1763 |
|
t, iCrc, (iCrc!=0x6e82f55b)?"| ERROR": "" ); |
1764 |
|
#endif |
1765 |
|
|
1766 |
|
#ifdef ARCH_IS_PPC |
1767 |
|
init_cpu(&cpu_list[1]); |
1768 |
|
TEST_YUYV(yv12_to_yuyv_altivec_c, 4); |
1769 |
|
printf(" yv12_to_yuyv_altivec_c %.3f usec crc32=0x%08x %s\n", |
1770 |
|
t, iCrc, (iCrc!=0xeb1a0b0a)?"| ERROR": "" ); |
1771 |
|
|
1772 |
|
TEST_YUYV(yv12_to_uyvy_altivec_c, 4); |
1773 |
|
printf(" yv12_to_uyvy_altivec_c %.3f usec crc32=0x%08x %s\n", |
1774 |
|
t, iCrc, (iCrc!=0x6e82f55b)?"| ERROR": "" ); |
1775 |
|
#endif |
1776 |
|
printf( " --- \n" ); |
1777 |
|
} |
1778 |
|
|
1779 |
|
#define TEST_YV2(FUNC, WITH_UV, WITH_FLIP) \ |
1780 |
|
ENTER \ |
1781 |
|
for(tst=0; tst<nb_tests; ++tst) (FUNC)(Dst0[0], Dst0[1], Dst0[2], WIDTH, WIDTH, \ |
1782 |
|
Src0[0], (WITH_UV) ? Src0[1] : 0, (WITH_UV) ? Src0[2] : 0, WIDTH, WIDTH, \ |
1783 |
|
WIDTH-2, HEIGHT-2, WITH_FLIP); \ |
1784 |
|
LEAVE |
1785 |
|
|
1786 |
|
#define PRINT_NxN(DATA,W,H,STR) { \ |
1787 |
|
int i,j; \ |
1788 |
|
for(j=0; j<(H); ++j) { \ |
1789 |
|
for(i=0; i<(W); ++i) printf( "0x%.2x ", (DATA)[i+j*(STR)] );\ |
1790 |
|
printf("\n"); \ |
1791 |
|
} \ |
1792 |
|
printf("---\n"); \ |
1793 |
|
} |
1794 |
|
|
1795 |
|
static const int yv12_CRCs[2][2] = { |
1796 |
|
{0x5cab7cf0,0xdab46541} |
1797 |
|
, {0xe8bae865,0x1faf77b7} |
1798 |
|
}; |
1799 |
|
|
1800 |
|
void test_yuv2() |
1801 |
|
{ |
1802 |
|
const int nb_tests = 800*speed_ref; |
1803 |
|
CPU *cpu; |
1804 |
|
uint8_t Src0[3][WIDTH*HEIGHT]; |
1805 |
|
uint8_t Dst0[3][WIDTH*HEIGHT]; |
1806 |
|
int with_uv, with_flip; |
1807 |
|
int i, j; |
1808 |
|
double t; |
1809 |
|
int tst, iCrc; |
1810 |
|
|
1811 |
|
colorspace_init(); |
1812 |
|
ieee_reseed(1); |
1813 |
|
for(i=0; i<(int)sizeof(Src0); ++i) Src0[0][i] = ieee_rand(0,255); |
1814 |
|
|
1815 |
|
printf( "\n === test YV2 ===\n" ); |
1816 |
|
for(with_flip=0; with_flip<=1; ++with_flip) { |
1817 |
|
for(with_uv=0; with_uv<=1; ++with_uv) { |
1818 |
|
init_cpu(&cpu_list[0]); |
1819 |
|
TEST_YV2(yv12_to_yv12_c, with_uv, with_flip); |
1820 |
|
printf(" yv12_to_yv12_c %.3f usec \tcrc32=0x%08x %s\n", |
1821 |
|
t, iCrc, (iCrc!=yv12_CRCs[with_flip][with_uv])?"| ERROR": "" ); |
1822 |
|
/* if (!with_uv) PRINT_NxN(Dst0[1], WIDTH/2, HEIGHT/2, WIDTH ); */ |
1823 |
|
|
1824 |
|
#ifdef ARCH_IS_IA32 |
1825 |
|
init_cpu(&cpu_list[1]); |
1826 |
|
TEST_YV2(yv12_to_yv12_mmx, with_uv, with_flip); |
1827 |
|
printf(" yv12_to_yv12_mmx %.3f usec \tcrc32=0x%08x %s\n", |
1828 |
|
t, iCrc, (iCrc!=yv12_CRCs[with_flip][with_uv])?"| ERROR": "" ); |
1829 |
|
/* if (!with_uv) PRINT_NxN(Dst0[1], WIDTH/2, HEIGHT/2, WIDTH ); */ |
1830 |
|
|
1831 |
|
TEST_YV2(yv12_to_yv12_xmm, with_uv, with_flip); |
1832 |
|
printf(" yv12_to_yv12_xmm %.3f usec \tcrc32=0x%08x %s\n", |
1833 |
|
t, iCrc, (iCrc!=yv12_CRCs[with_flip][with_uv])?"| ERROR": "" ); |
1834 |
|
#endif |
1835 |
|
} |
1836 |
|
|
1837 |
|
printf( " --- \n" ); |
1838 |
|
} |
1839 |
|
printf( " ===== \n" ); |
1840 |
|
} |
1841 |
|
|
1842 |
|
#undef WIDTH |
1843 |
|
#undef HEIGHT |
1844 |
|
#undef ENTER |
1845 |
|
#undef LEAVE |
1846 |
|
|
1847 |
/*********************************************************************/ |
/*********************************************************************/ |
1848 |
|
|
1849 |
static uint32_t __inline log2bin_v1(uint32_t value) |
static uint32_t __inline log2bin_v1(uint32_t value) |
1966 |
} |
} |
1967 |
|
|
1968 |
/********************************************************************* |
/********************************************************************* |
1969 |
|
* test compiler |
1970 |
|
*********************************************************************/ |
1971 |
|
|
1972 |
|
void test_compiler() { |
1973 |
|
int nb_err = 0; |
1974 |
|
int32_t v; |
1975 |
|
if (sizeof(uint16_t)<2) { |
1976 |
|
printf( "ERROR: sizeof(uint16_t)<2 !!\n" ); |
1977 |
|
nb_err++; |
1978 |
|
} |
1979 |
|
if (sizeof(int16_t)<2) { |
1980 |
|
printf( "ERROR: sizeof(int16_t)<2 !!\n" ); |
1981 |
|
nb_err++; |
1982 |
|
} |
1983 |
|
if (sizeof(uint8_t)!=1) { |
1984 |
|
printf( "ERROR: sizeof(uint8_t)!=1 !!\n" ); |
1985 |
|
nb_err++; |
1986 |
|
} |
1987 |
|
if (sizeof(int8_t)!=1) { |
1988 |
|
printf( "ERROR: sizeof(int8_t)!=1 !!\n" ); |
1989 |
|
nb_err++; |
1990 |
|
} |
1991 |
|
if (sizeof(uint32_t)<4) { |
1992 |
|
printf( "ERROR: sizeof(uint32_t)<4 !!\n" ); |
1993 |
|
nb_err++; |
1994 |
|
} |
1995 |
|
if (sizeof(int32_t)<4) { |
1996 |
|
printf( "ERROR: sizeof(int32_t)<4 !!\n" ); |
1997 |
|
nb_err++; |
1998 |
|
} |
1999 |
|
/* yes, i know, this test is silly. But better be safe than sorry. :) */ |
2000 |
|
for(v=1000; v>=0; v--) { |
2001 |
|
if ( (v>>2) != v/4) |
2002 |
|
nb_err++; |
2003 |
|
} |
2004 |
|
for(v=-1000; v!=-1; v++) { |
2005 |
|
if ( (v>>2) != (v/4)-!!(v%4)) |
2006 |
|
nb_err++; |
2007 |
|
} |
2008 |
|
if (nb_err!=0) { |
2009 |
|
printf( "ERROR! please post your platform/compiler specs to xvid-devel@xvid.org !\n" ); |
2010 |
|
} |
2011 |
|
} |
2012 |
|
/********************************************************************* |
2013 |
|
* test SSIM functions |
2014 |
|
*********************************************************************/ |
2015 |
|
|
2016 |
|
typedef int (*lumfunc)(uint8_t* ptr, int stride); |
2017 |
|
typedef void (*csfunc)(uint8_t* ptro, uint8_t* ptrc, int stride, int lumo, int lumc, int* pdevo, int* pdevc, int* pcorr); |
2018 |
|
|
2019 |
|
extern int lum_8x8_c(uint8_t* ptr, int stride); |
2020 |
|
extern int lum_8x8_mmx(uint8_t* ptr, int stride); |
2021 |
|
extern int lum_2x8_c(uint8_t* ptr, int stride); |
2022 |
|
extern void iconsim_c(uint8_t* ptro, uint8_t* ptrc, int stride, int lumo, int lumc, int* pdevo, int* pdevc, int* pcorr); |
2023 |
|
extern void consim_mmx(uint8_t* ptro, uint8_t* ptrc, int stride, int lumo, int lumc, int* pdevo, int* pdevc, int* pcorr); |
2024 |
|
extern void consim_sse2(uint8_t* ptro, uint8_t* ptrc, int stride, int lumo, int lumc, int* pdevo, int* pdevc, int* pcorr); |
2025 |
|
|
2026 |
|
void test_SSIM() |
2027 |
|
{ |
2028 |
|
const int nb_tests = 3000*speed_ref; |
2029 |
|
int tst; |
2030 |
|
CPU *cpu; |
2031 |
|
int i; |
2032 |
|
int devs[3]; |
2033 |
|
long lumo, lumc; |
2034 |
|
DECLARE_ALIGNED_MATRIX(Ref1, 16, 16, uint8_t, 16); |
2035 |
|
DECLARE_ALIGNED_MATRIX(Ref2, 16, 16, uint8_t, 16); |
2036 |
|
lumfunc lum8x8; |
2037 |
|
lumfunc lum2x8; |
2038 |
|
csfunc csim; |
2039 |
|
|
2040 |
|
ieee_reseed(1); |
2041 |
|
printf( "\n ====== test SSIM ======\n" ); |
2042 |
|
for(i=0; i<16*16;++i) { |
2043 |
|
long v1, v2; |
2044 |
|
v1 = ieee_rand(-256, 511); |
2045 |
|
v2 = ieee_rand(-256, 511); |
2046 |
|
Ref1[i] = (v1<0) ? 0 : (v1>255) ? 255 : v1; |
2047 |
|
Ref2[i] = (v2<0) ? 0 : (v2>255) ? 255 : v2; |
2048 |
|
} |
2049 |
|
lumc = ieee_rand(0, 255); |
2050 |
|
lumo = ieee_rand(0, 255); |
2051 |
|
|
2052 |
|
for(cpu = cpu_list; cpu->name!=0; ++cpu) |
2053 |
|
{ |
2054 |
|
double t; |
2055 |
|
int m; |
2056 |
|
if (!init_cpu(cpu)) |
2057 |
|
continue; |
2058 |
|
lum8x8 = lum_8x8_c; |
2059 |
|
lum2x8 = lum_2x8_c; |
2060 |
|
csim = iconsim_c; |
2061 |
|
if (cpu->cpu & XVID_CPU_MMX){ |
2062 |
|
lum8x8 = lum_8x8_mmx; |
2063 |
|
csim = consim_mmx; |
2064 |
|
} |
2065 |
|
if (cpu->cpu & XVID_CPU_MMX){ |
2066 |
|
csim = consim_sse2; |
2067 |
|
} |
2068 |
|
|
2069 |
|
t = gettime_usec(); |
2070 |
|
emms(); |
2071 |
|
for(tst=0; tst<nb_tests; ++tst) m = lum8x8(Ref1, 16); |
2072 |
|
emms(); |
2073 |
|
t = (gettime_usec() - t) / nb_tests; |
2074 |
|
printf("%s - ssim-lum8x8 %.3f usec m=%d %s\n", |
2075 |
|
cpu->name, t, m, |
2076 |
|
(m!=8230)?"| ERROR": "" ); |
2077 |
|
|
2078 |
|
t = gettime_usec(); |
2079 |
|
emms(); |
2080 |
|
for(tst=0; tst<nb_tests; ++tst) m = lum2x8(Ref1+8, 16); |
2081 |
|
emms(); |
2082 |
|
t = (gettime_usec() - t) / nb_tests; |
2083 |
|
printf("%s - ssim-lum2x8 %.3f usec m=%d %s\n", |
2084 |
|
cpu->name, t, m, |
2085 |
|
(m!=681)?"| ERROR": "" ); |
2086 |
|
|
2087 |
|
t = gettime_usec(); |
2088 |
|
emms(); |
2089 |
|
for(tst=0; tst<nb_tests; ++tst) csim(Ref1, Ref2, 16, lumo, lumc, devs+0, devs+1, devs+2); |
2090 |
|
emms(); |
2091 |
|
t = (gettime_usec() - t) / nb_tests; |
2092 |
|
printf("%s - ssim-consim %.3f usec devs=[0x%x 0x%x 0x%x] %s\n", |
2093 |
|
cpu->name, t, devs[0], devs[1], devs[2], |
2094 |
|
(devs[0]!=0xeba80 || devs[1]!=0x1053e7 || devs[2]!=0x51215)?"| ERROR": "" ); |
2095 |
|
|
2096 |
|
printf( " --- \n" ); |
2097 |
|
} |
2098 |
|
} |
2099 |
|
|
2100 |
|
/********************************************************************* |
2101 |
* main |
* main |
2102 |
*********************************************************************/ |
*********************************************************************/ |
2103 |
|
|
2165 |
if (what==0 || what==10) test_sse(); |
if (what==0 || what==10) test_sse(); |
2166 |
if (what==0 || what==11) test_log2bin(); |
if (what==0 || what==11) test_log2bin(); |
2167 |
if (what==0 || what==12) test_gcd(); |
if (what==0 || what==12) test_gcd(); |
2168 |
|
if (what==0 || what==13) test_compiler(); |
2169 |
|
if (what==0 || what==14) test_yuv(); |
2170 |
|
if (what==0 || what==15) test_SSIM(); |
2171 |
|
if (what==0 || what==16) test_yuv2(); |
2172 |
|
|
2173 |
|
|
2174 |
if (what==7) { |
if (what==7) { |