19 |
* along with this program; if not, write to the Free Software |
* along with this program; if not, write to the Free Software |
20 |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
21 |
* |
* |
22 |
* $Id: xvid_bench.c,v 1.19 2005-05-23 09:29:43 Skal Exp $ |
* $Id: xvid_bench.c,v 1.36 2008-11-14 15:43:27 Isibaar Exp $ |
23 |
* |
* |
24 |
****************************************************************************/ |
****************************************************************************/ |
25 |
|
|
58 |
#include "utils/timer.h" |
#include "utils/timer.h" |
59 |
#include "quant/quant_matrix.c" |
#include "quant/quant_matrix.c" |
60 |
#include "bitstream/cbp.h" |
#include "bitstream/cbp.h" |
61 |
|
#include "bitstream/bitstream.h" |
62 |
|
|
63 |
#include <math.h> |
#include <math.h> |
64 |
|
|
113 |
unsigned int cpu; |
unsigned int cpu; |
114 |
} CPU; |
} CPU; |
115 |
|
|
116 |
CPU cpu_list[] = |
CPU cpu_list[] = { |
117 |
{ { "PLAINC", 0 } |
{ "PLAINC ", 0 }, |
118 |
#ifdef ARCH_IS_IA32 |
#ifdef ARCH_IS_IA32 |
119 |
, { "MMX ", XVID_CPU_MMX } |
{ "MMX ", XVID_CPU_MMX }, |
120 |
, { "MMXEXT", XVID_CPU_MMXEXT | XVID_CPU_MMX } |
{ "MMXEXT ", XVID_CPU_MMXEXT | XVID_CPU_MMX }, |
121 |
, { "SSE2 ", XVID_CPU_SSE2 | XVID_CPU_MMX } |
{ "SSE2 ", XVID_CPU_SSE2 | XVID_CPU_MMX }, |
122 |
, { "3DNOW ", XVID_CPU_3DNOW } |
{ "SSE3 ", XVID_CPU_SSE3 | XVID_CPU_SSE2 | XVID_CPU_MMX }, |
123 |
, { "3DNOWE", XVID_CPU_3DNOW | XVID_CPU_3DNOWEXT } |
{ "SSE41 ", XVID_CPU_SSE41| XVID_CPU_SSE3 | XVID_CPU_SSE2 | XVID_CPU_MMX }, |
124 |
|
{ "3DNOW ", XVID_CPU_3DNOW }, |
125 |
|
{ "3DNOWE ", XVID_CPU_3DNOW | XVID_CPU_3DNOWEXT }, |
126 |
#endif |
#endif |
127 |
#ifdef ARCH_IS_PPC |
#ifdef ARCH_IS_PPC |
128 |
, { "ALTIVEC", XVID_CPU_ALTIVEC } |
{ "ALTIVEC", XVID_CPU_ALTIVEC }, |
129 |
#endif |
#endif |
130 |
#ifdef ARCH_IS_X86_64 |
#ifdef ARCH_IS_X86_64 |
131 |
, { "X86_64", XVID_CPU_ASM} |
{ "X86_64 ", XVID_CPU_ASM}, |
132 |
#endif |
#endif |
133 |
//, { "IA64 ", XVID_CPU_IA64 } |
#ifdef ARCH_IS_IA64 |
134 |
//, { "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 } |
|
135 |
#endif |
#endif |
136 |
, { 0, 0 } }; |
// { "TSC ", XVID_CPU_TSC }, |
137 |
|
{ 0, 0 } |
138 |
|
}; |
139 |
|
|
140 |
|
|
141 |
int init_cpu(CPU *cpu) |
int init_cpu(CPU *cpu) |
263 |
return crc; |
return crc; |
264 |
} |
} |
265 |
|
|
266 |
|
void byte_swap(uint8_t *mem, int len, int element_size) { |
267 |
|
#ifdef ARCH_IS_BIG_ENDIAN |
268 |
|
int i; |
269 |
|
|
270 |
|
if(element_size == 1) { |
271 |
|
/* No need to swap */ |
272 |
|
} else if(element_size == 2) { |
273 |
|
uint8_t temp[2]; |
274 |
|
|
275 |
|
for(i=0; i < (len/2); i++ ) { |
276 |
|
temp[0] = mem[0]; |
277 |
|
temp[1] = mem[1]; |
278 |
|
mem[0] = temp[1]; |
279 |
|
mem[1] = temp[0]; |
280 |
|
|
281 |
|
mem += 2; |
282 |
|
} |
283 |
|
} else if(element_size == 4) { |
284 |
|
uint8_t temp[4]; |
285 |
|
|
286 |
|
for(i=0; i < (len/4); i++ ) { |
287 |
|
temp[0] = mem[0]; |
288 |
|
temp[1] = mem[1]; |
289 |
|
temp[2] = mem[2]; |
290 |
|
temp[3] = mem[3]; |
291 |
|
mem[0] = temp[3]; |
292 |
|
mem[1] = temp[2]; |
293 |
|
mem[2] = temp[1]; |
294 |
|
mem[3] = temp[0]; |
295 |
|
|
296 |
|
mem += 4; |
297 |
|
} |
298 |
|
} else { |
299 |
|
printf("ERROR: byte_swap unsupported element_size(%u)\n", element_size); |
300 |
|
} |
301 |
|
#endif |
302 |
|
} |
303 |
|
|
304 |
/********************************************************************* |
/********************************************************************* |
305 |
* test DCT |
* test DCT |
306 |
*********************************************************************/ |
*********************************************************************/ |
521 |
(iCrc!=8107)?"| ERROR": "" ); |
(iCrc!=8107)?"| ERROR": "" ); |
522 |
#endif |
#endif |
523 |
|
|
524 |
|
/* New functions for field prediction by CK 1.10.2005 */ |
525 |
|
#pragma NEW8X4 |
526 |
|
TEST_MB(interpolate8x4_halfpel_h, 0); |
527 |
|
printf("%s - interpfield-h -round0 %.3f usec crc32=0x%08x %s\n", |
528 |
|
cpu->name, t, iCrc, |
529 |
|
(iCrc!=0x9538d6df)?"| ERROR": "" ); |
530 |
|
|
531 |
|
TEST_MB(interpolate8x4_halfpel_h, 1); |
532 |
|
printf("%s - round1 %.3f usec crc32=0x%08x %s\n", |
533 |
|
cpu->name, t, iCrc, |
534 |
|
(iCrc!=0xde5f1db4)?"| ERROR": "" ); |
535 |
|
|
536 |
|
|
537 |
|
TEST_MB(interpolate8x4_halfpel_v, 0); |
538 |
|
printf("%s - interpfield- v-round0 %.3f usec crc32=0x%08x %s\n", |
539 |
|
cpu->name, t, iCrc, |
540 |
|
(iCrc!=0xea5a69ef)?"| ERROR": "" ); |
541 |
|
|
542 |
|
TEST_MB(interpolate8x4_halfpel_v, 1); |
543 |
|
printf("%s - round1 %.3f usec crc32=0x%08x %s\n", |
544 |
|
cpu->name, t, iCrc, |
545 |
|
(iCrc!=0x4f10ec0f)?"| ERROR": "" ); |
546 |
|
|
547 |
|
|
548 |
|
TEST_MB(interpolate8x4_halfpel_hv, 0); |
549 |
|
printf("%s - interpfield-hv-round0 %.3f usec crc32=0x%08x %s\n", |
550 |
|
cpu->name, t, iCrc, |
551 |
|
(iCrc!=0xf97ee367)?"| ERROR": "" ); |
552 |
|
|
553 |
|
TEST_MB(interpolate8x4_halfpel_hv, 1); |
554 |
|
printf("%s - round1 %.3f usec crc32=0x%08x %s\n", |
555 |
|
cpu->name, t, iCrc, |
556 |
|
(iCrc!=0xb6a9f581)?"| ERROR": "" ); |
557 |
|
/* End of 8x4 functions */ |
558 |
|
|
559 |
printf( " --- \n" ); |
printf( " --- \n" ); |
560 |
} |
} |
561 |
} |
} |
562 |
|
|
563 |
|
#undef ENTER |
564 |
|
#undef LEAVE |
565 |
|
#undef TEST_MB |
566 |
|
#undef TEST_MB2 |
567 |
|
|
568 |
/********************************************************************* |
/********************************************************************* |
569 |
* test transfer |
* test transfer |
570 |
*********************************************************************/ |
*********************************************************************/ |
594 |
} \ |
} \ |
595 |
emms(); \ |
emms(); \ |
596 |
t = (gettime_usec()-t -overhead) / nb_tests;\ |
t = (gettime_usec()-t -overhead) / nb_tests;\ |
597 |
s = calc_crc((uint8_t*)(DST), sizeof((DST)), CRC32_INITIAL) |
byte_swap((uint8_t*)(DST), 8*32*sizeof((DST)[0]), sizeof((DST)[0])); \ |
598 |
|
s = calc_crc((uint8_t*)(DST), 8*32*sizeof((DST)[0]), CRC32_INITIAL) |
599 |
|
|
600 |
#define TEST_TRANSFER(FUNC, DST, SRC) \ |
#define TEST_TRANSFER(FUNC, DST, SRC) \ |
601 |
TEST_TRANSFER_BEGIN(DST); \ |
TEST_TRANSFER_BEGIN(DST); \ |
621 |
} \ |
} \ |
622 |
emms(); \ |
emms(); \ |
623 |
t = (gettime_usec()-t -overhead) / nb_tests;\ |
t = (gettime_usec()-t -overhead) / nb_tests;\ |
624 |
s = calc_crc((uint8_t*)(DST), sizeof((DST)), CRC32_INITIAL) |
byte_swap((uint8_t*)(DST), 8*32*sizeof((DST)[0]), sizeof((DST)[0])); \ |
625 |
|
s = calc_crc((uint8_t*)(DST), 8*32*sizeof((DST)[0]), CRC32_INITIAL) |
626 |
|
|
627 |
#define TEST_TRANSFER2(FUNC, DST, SRC, R1) \ |
#define TEST_TRANSFER2(FUNC, DST, SRC, R1) \ |
628 |
TEST_TRANSFER2_BEGIN(DST,SRC); \ |
TEST_TRANSFER2_BEGIN(DST,SRC); \ |
639 |
const int nb_tests = 4000*speed_ref; |
const int nb_tests = 4000*speed_ref; |
640 |
int i; |
int i; |
641 |
CPU *cpu; |
CPU *cpu; |
642 |
uint8_t Src8[8*32], Dst8[8*32], Ref1[8*32], Ref2[8*32]; |
// uint8_t Src8[8*32], Dst8[8*32], Ref1[8*32], Ref2[8*32]; |
643 |
int16_t Src16[8*32], Dst16[8*32]; |
// int16_t Src16[8*32], Dst16[8*32]; |
644 |
|
DECLARE_ALIGNED_MATRIX(Src8, 8, 32, uint8_t, CACHE_LINE); |
645 |
|
DECLARE_ALIGNED_MATRIX(Dst8, 8, 32, uint8_t, CACHE_LINE); |
646 |
|
DECLARE_ALIGNED_MATRIX(Ref1, 8, 32, uint8_t, CACHE_LINE); |
647 |
|
DECLARE_ALIGNED_MATRIX(Ref2, 8, 32, uint8_t, CACHE_LINE); |
648 |
|
DECLARE_ALIGNED_MATRIX(Src16, 8, 32, uint16_t, CACHE_LINE); |
649 |
|
DECLARE_ALIGNED_MATRIX(Dst16, 8, 32, uint16_t, CACHE_LINE); |
650 |
|
|
651 |
printf( "\n === test transfer ===\n" ); |
printf( "\n === test transfer ===\n" ); |
652 |
|
|
668 |
cpu->name, t, s, |
cpu->name, t, s, |
669 |
(s!=0xee7ccbb4)?"| ERROR": ""); |
(s!=0xee7ccbb4)?"| ERROR": ""); |
670 |
|
|
671 |
|
/* New functions for field prediction by CK 1.10.2005 */ |
672 |
|
#pragma NEW8X4 |
673 |
|
TEST_TRANSFER(transfer8x4_copy, Dst8, Src8); |
674 |
|
printf("%s - 8to4 %.3f usec crc32=0x%08x %s\n", |
675 |
|
cpu->name, t, s, |
676 |
|
(s!=0xbb9c3db5)?"| ERROR": ""); |
677 |
|
/* End of new functions */ |
678 |
|
|
679 |
TEST_TRANSFER(transfer8x8_copy, Dst8, Src8); |
TEST_TRANSFER(transfer8x8_copy, Dst8, Src8); |
680 |
printf("%s - 8to8 %.3f usec crc32=0x%08x %s\n", |
printf("%s - 8to8 %.3f usec crc32=0x%08x %s\n", |
681 |
cpu->name, t, s, |
cpu->name, t, s, |
689 |
TEST_TRANSFER2(transfer_8to16sub, Dst16, Src8, Ref1); |
TEST_TRANSFER2(transfer_8to16sub, Dst16, Src8, Ref1); |
690 |
{ |
{ |
691 |
int s1, s2; |
int s1, s2; |
692 |
s1 = calc_crc((uint8_t*)Dst16, sizeof(Dst16), CRC32_INITIAL); |
s1 = calc_crc((uint8_t*)Dst16, 8*32*sizeof(Dst16[0]), CRC32_INITIAL); |
693 |
s2 = calc_crc((uint8_t*)Src8, sizeof(Src8), CRC32_INITIAL); |
s2 = calc_crc((uint8_t*)Src8, 8*32*sizeof(Src8[0]), CRC32_INITIAL); |
694 |
printf("%s - 8to16sub %.3f usec crc32(1)=0x%08x crc32(2)=0x%08x %s %s\n", |
printf("%s - 8to16sub %.3f usec crc32(1)=0x%08x crc32(2)=0x%08x %s %s\n", |
695 |
cpu->name, t, s1, s2, |
cpu->name, t, s1, s2, |
696 |
(s1!=0xa1e07163)?"| ERROR1": "", |
(s1!=0xa1e07163)?"| ERROR1": "", |
719 |
for(q=1; q<=max_Q; ++q) { \ |
for(q=1; q<=max_Q; ++q) { \ |
720 |
for(tst=0; tst<nb_tests; ++tst) \ |
for(tst=0; tst<nb_tests; ++tst) \ |
721 |
(FUNC)((DST), (SRC), q, mpeg_quant_matrices); \ |
(FUNC)((DST), (SRC), q, mpeg_quant_matrices); \ |
722 |
s = calc_crc((uint8_t*)(DST), 64*sizeof(int16_t), s); \ |
byte_swap((uint8_t*)(DST), 64*sizeof((DST)[0]), sizeof((DST)[0])); \ |
723 |
|
s = calc_crc((uint8_t*)(DST), 64*sizeof((DST)[0]), s); \ |
724 |
} \ |
} \ |
725 |
emms(); \ |
emms(); \ |
726 |
} \ |
} \ |
735 |
for(q=1; q<=max_Q; ++q) { \ |
for(q=1; q<=max_Q; ++q) { \ |
736 |
for(tst=0; tst<nb_tests; ++tst) \ |
for(tst=0; tst<nb_tests; ++tst) \ |
737 |
(FUNC)((DST), (SRC), q, q, mpeg_quant_matrices); \ |
(FUNC)((DST), (SRC), q, q, mpeg_quant_matrices); \ |
738 |
s = calc_crc((uint8_t*)(DST), 64*sizeof(int16_t), s); \ |
byte_swap((uint8_t*)(DST), 64*sizeof((DST)[0]), sizeof((DST)[0])); \ |
739 |
|
s = calc_crc((uint8_t*)(DST), 64*sizeof((DST)[0]), s); \ |
740 |
} \ |
} \ |
741 |
emms(); \ |
emms(); \ |
742 |
} \ |
} \ |
956 |
* test non-zero AC counting |
* test non-zero AC counting |
957 |
*********************************************************************/ |
*********************************************************************/ |
958 |
|
|
959 |
#define TEST_CBP(FUNC, SRC) \ |
#define TEST_CBP(FUNC, SRC, NB) \ |
960 |
t = gettime_usec(); \ |
t = gettime_usec(); \ |
961 |
emms(); \ |
emms(); \ |
962 |
for(tst=0; tst<nb_tests; ++tst) { \ |
for(tst=0; tst<NB; ++tst) { \ |
963 |
cbp = (FUNC)((SRC)); \ |
cbp = (FUNC)((SRC)); \ |
964 |
} \ |
} \ |
965 |
emms(); \ |
emms(); \ |
968 |
void test_cbp() |
void test_cbp() |
969 |
{ |
{ |
970 |
const int nb_tests = 10000*speed_ref; |
const int nb_tests = 10000*speed_ref; |
971 |
int i; |
int i, n, m; |
972 |
CPU *cpu; |
CPU *cpu; |
973 |
DECLARE_ALIGNED_MATRIX(Src1, 6, 64, int16_t, 16); |
DECLARE_ALIGNED_MATRIX(Src1, 6, 64, int16_t, 16); |
974 |
DECLARE_ALIGNED_MATRIX(Src2, 6, 64, int16_t, 16); |
DECLARE_ALIGNED_MATRIX(Src2, 6, 64, int16_t, 16); |
975 |
DECLARE_ALIGNED_MATRIX(Src3, 6, 64, int16_t, 16); |
DECLARE_ALIGNED_MATRIX(Src3, 6, 64, int16_t, 16); |
976 |
DECLARE_ALIGNED_MATRIX(Src4, 6, 64, int16_t, 16); |
DECLARE_ALIGNED_MATRIX(Src4, 6, 64, int16_t, 16); |
977 |
|
DECLARE_ALIGNED_MATRIX(Src5, 6, 64, int16_t, 16); |
978 |
|
|
979 |
printf( "\n ===== test cbp =====\n" ); |
printf( "\n ===== test cbp =====\n" ); |
980 |
|
|
983 |
Src2[i] = (i<3*64); /* half-full */ |
Src2[i] = (i<3*64); /* half-full */ |
984 |
Src3[i] = ((i+32)>3*64); |
Src3[i] = ((i+32)>3*64); |
985 |
Src4[i] = (i==(3*64+2) || i==(5*64+9)); |
Src4[i] = (i==(3*64+2) || i==(5*64+9)); |
986 |
|
Src5[i] = ieee_rand(0,1) ? -1 : 1; /* +/- test */ |
987 |
} |
} |
988 |
|
|
989 |
for(cpu = cpu_list; cpu->name!=0; ++cpu) |
for(cpu = cpu_list; cpu->name!=0; ++cpu) |
994 |
if (!init_cpu(cpu)) |
if (!init_cpu(cpu)) |
995 |
continue; |
continue; |
996 |
|
|
997 |
TEST_CBP(calc_cbp, Src1); |
TEST_CBP(calc_cbp, Src1, nb_tests); |
998 |
printf("%s - calc_cbp#1 %.3f usec cbp=0x%02x %s\n", |
printf("%s - calc_cbp#1 %.3f usec cbp=0x%02x %s\n", |
999 |
cpu->name, t, cbp, (cbp!=0x15)?"| ERROR": ""); |
cpu->name, t, cbp, (cbp!=0x15)?"| ERROR": ""); |
1000 |
TEST_CBP(calc_cbp, Src2); |
TEST_CBP(calc_cbp, Src2, nb_tests); |
1001 |
printf("%s - calc_cbp#2 %.3f usec cbp=0x%02x %s\n", |
printf("%s - calc_cbp#2 %.3f usec cbp=0x%02x %s\n", |
1002 |
cpu->name, t, cbp, (cbp!=0x38)?"| ERROR": ""); |
cpu->name, t, cbp, (cbp!=0x38)?"| ERROR": ""); |
1003 |
TEST_CBP(calc_cbp, Src3); |
TEST_CBP(calc_cbp, Src3, nb_tests); |
1004 |
printf("%s - calc_cbp#3 %.3f usec cbp=0x%02x %s\n", |
printf("%s - calc_cbp#3 %.3f usec cbp=0x%02x %s\n", |
1005 |
cpu->name, t, cbp, (cbp!=0x0f)?"| ERROR": "" ); |
cpu->name, t, cbp, (cbp!=0x0f)?"| ERROR": "" ); |
1006 |
TEST_CBP(calc_cbp, Src4); |
TEST_CBP(calc_cbp, Src4, nb_tests); |
1007 |
printf("%s - calc_cbp#4 %.3f usec cbp=0x%02x %s\n", |
printf("%s - calc_cbp#4 %.3f usec cbp=0x%02x %s\n", |
1008 |
cpu->name, t, cbp, (cbp!=0x05)?"| ERROR": "" ); |
cpu->name, t, cbp, (cbp!=0x05)?"| ERROR": "" ); |
1009 |
|
TEST_CBP(calc_cbp, Src5, nb_tests); |
1010 |
|
printf("%s - calc_cbp#4 %.3f usec cbp=0x%02x %s\n", |
1011 |
|
cpu->name, t, cbp, (cbp!=0x3f)?"| ERROR": "" ); |
1012 |
printf( " --- \n" ); |
printf( " --- \n" ); |
1013 |
} |
} |
1014 |
|
|
1015 |
|
for(cpu = cpu_list; cpu->name!=0; ++cpu) /* bench suggested by Carlo (carlo dot bramix at libero dot it) */ |
1016 |
|
{ |
1017 |
|
double t; |
1018 |
|
int tst, cbp, err; |
1019 |
|
|
1020 |
|
if (!init_cpu(cpu)) |
1021 |
|
continue; |
1022 |
|
|
1023 |
|
err = 0; |
1024 |
|
for(n=0; n<6; ++n) |
1025 |
|
{ |
1026 |
|
for(m=0; m<64; ++m) |
1027 |
|
{ |
1028 |
|
for(i=0; i<6*64; ++i) |
1029 |
|
Src1[i] = (i== (m + n*64)); |
1030 |
|
|
1031 |
|
TEST_CBP(calc_cbp, Src1, 1); |
1032 |
|
if (cbp!= (((m!=0)<<(5-n)))) |
1033 |
|
{ |
1034 |
|
printf( "%s - calc_cbp#5: ERROR at pos %d / %d!\n", cpu->name, n, m); |
1035 |
|
err = 1; |
1036 |
|
break; |
1037 |
|
} |
1038 |
|
} |
1039 |
|
} |
1040 |
|
if (!err) |
1041 |
|
printf( " %s - calc_cbp#5 : OK\n", cpu->name ); |
1042 |
|
|
1043 |
|
} |
1044 |
} |
} |
1045 |
|
|
1046 |
/********************************************************************* |
/********************************************************************* |
1503 |
xframe.bitstream = buf + pos; |
xframe.bitstream = buf + pos; |
1504 |
xframe.length = buf_size - pos; |
xframe.length = buf_size - pos; |
1505 |
xframe.output.plane[0] = (uint8_t*)(((size_t)yuv_out + 15) & ~15); |
xframe.output.plane[0] = (uint8_t*)(((size_t)yuv_out + 15) & ~15); |
1506 |
xframe.output.plane[1] = xframe.output.plane[0] + bps*height; |
xframe.output.plane[1] = (uint8_t*)xframe.output.plane[0] + bps*height; |
1507 |
xframe.output.plane[2] = xframe.output.plane[1] + bps/2; |
xframe.output.plane[2] = (uint8_t*)xframe.output.plane[1] + bps/2; |
1508 |
xframe.output.stride[0] = bps; |
xframe.output.stride[0] = bps; |
1509 |
xframe.output.stride[1] = bps; |
xframe.output.stride[1] = bps; |
1510 |
xframe.output.stride[2] = bps; |
xframe.output.stride[2] = bps; |
1522 |
nb++; |
nb++; |
1523 |
|
|
1524 |
for(y=0; y<height/2; ++y) { |
for(y=0; y<height/2; ++y) { |
1525 |
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); |
1526 |
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); |
1527 |
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); |
1528 |
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); |
1529 |
} |
} |
1530 |
if (pos==buf_size) |
if (pos==buf_size) |
1531 |
break; |
break; |
1708 |
} |
} |
1709 |
#endif |
#endif |
1710 |
} |
} |
1711 |
|
|
1712 |
|
/********************************************************************* |
1713 |
|
* test some YUV func |
1714 |
|
*********************************************************************/ |
1715 |
|
|
1716 |
|
#define ENTER \ |
1717 |
|
for(i=0; i<(int)sizeof(Dst0); ++i) Dst0[0][i] = 0; \ |
1718 |
|
t = gettime_usec(); \ |
1719 |
|
emms(); |
1720 |
|
|
1721 |
|
#define LEAVE \ |
1722 |
|
emms(); \ |
1723 |
|
t = (gettime_usec() - t) / nb_tests; \ |
1724 |
|
iCrc = calc_crc((uint8_t*)Dst0, sizeof(Dst0), CRC32_INITIAL) |
1725 |
|
|
1726 |
|
#define TEST_YUYV(FUNC, S) \ |
1727 |
|
ENTER \ |
1728 |
|
for(tst=0; tst<nb_tests; ++tst) (FUNC)(Dst0[0], S*WIDTH, Src0[0], Src0[1], Src0[2], WIDTH, WIDTH/2, WIDTH, HEIGHT, 0); \ |
1729 |
|
LEAVE |
1730 |
|
|
1731 |
|
#define WIDTH 128 |
1732 |
|
#define HEIGHT 32 |
1733 |
|
void test_yuv() |
1734 |
|
{ |
1735 |
|
const int nb_tests = 200*speed_ref; |
1736 |
|
CPU *cpu; |
1737 |
|
uint8_t Src0[3][WIDTH*HEIGHT]; |
1738 |
|
uint8_t Dst0[4][WIDTH*HEIGHT]; |
1739 |
|
int i, j; |
1740 |
|
double t; |
1741 |
|
int tst, iCrc; |
1742 |
|
|
1743 |
|
colorspace_init(); |
1744 |
|
ieee_reseed(1); |
1745 |
|
for(i=0; i<(int)sizeof(Src0); ++i) Src0[0][i] = ieee_rand(0,255); |
1746 |
|
for(i=0; i<(int)sizeof(Dst0); ++i) Dst0[0][i] = 0x5a; |
1747 |
|
|
1748 |
|
printf( "\n === test YUV ===\n" ); |
1749 |
|
|
1750 |
|
init_cpu(&cpu_list[0]); |
1751 |
|
TEST_YUYV(yv12_to_yuyv_c, 4); |
1752 |
|
printf(" yv12_to_yuyv_c %.3f usec crc32=0x%08x %s\n", |
1753 |
|
t, iCrc, (iCrc!=0xeb1a0b0a)?"| ERROR": "" ); |
1754 |
|
TEST_YUYV(yv12_to_uyvy_c, 4); |
1755 |
|
printf(" yv12_to_uyvy_c %.3f usec crc32=0x%08x %s\n", |
1756 |
|
t, iCrc, (iCrc!=0x6e82f55b)?"| ERROR": "" ); |
1757 |
|
|
1758 |
|
#ifdef ARCH_IS_IA32 |
1759 |
|
init_cpu(&cpu_list[1]); |
1760 |
|
TEST_YUYV(yv12_to_yuyv_mmx, 4); |
1761 |
|
printf(" yv12_to_yuyv_mmx %.3f usec crc32=0x%08x %s\n", |
1762 |
|
t, iCrc, (iCrc!=0xeb1a0b0a)?"| ERROR": "" ); |
1763 |
|
|
1764 |
|
TEST_YUYV(yv12_to_uyvy_mmx, 4); |
1765 |
|
printf(" yv12_to_uyvy_mmx %.3f usec crc32=0x%08x %s\n", |
1766 |
|
t, iCrc, (iCrc!=0x6e82f55b)?"| ERROR": "" ); |
1767 |
|
#endif |
1768 |
|
|
1769 |
|
#ifdef ARCH_IS_PPC |
1770 |
|
init_cpu(&cpu_list[1]); |
1771 |
|
TEST_YUYV(yv12_to_yuyv_altivec_c, 4); |
1772 |
|
printf(" yv12_to_yuyv_altivec_c %.3f usec crc32=0x%08x %s\n", |
1773 |
|
t, iCrc, (iCrc!=0xeb1a0b0a)?"| ERROR": "" ); |
1774 |
|
|
1775 |
|
TEST_YUYV(yv12_to_uyvy_altivec_c, 4); |
1776 |
|
printf(" yv12_to_uyvy_altivec_c %.3f usec crc32=0x%08x %s\n", |
1777 |
|
t, iCrc, (iCrc!=0x6e82f55b)?"| ERROR": "" ); |
1778 |
|
#endif |
1779 |
|
printf( " --- \n" ); |
1780 |
|
} |
1781 |
|
|
1782 |
|
#define TEST_YV2(FUNC, WITH_UV, WITH_FLIP) \ |
1783 |
|
ENTER \ |
1784 |
|
for(tst=0; tst<nb_tests; ++tst) (FUNC)(Dst0[0], Dst0[1], Dst0[2], WIDTH, WIDTH, \ |
1785 |
|
Src0[0], (WITH_UV) ? Src0[1] : 0, (WITH_UV) ? Src0[2] : 0, WIDTH, WIDTH, \ |
1786 |
|
WIDTH-2, HEIGHT-2, WITH_FLIP); \ |
1787 |
|
LEAVE |
1788 |
|
|
1789 |
|
#define PRINT_NxN(DATA,W,H,STR) { \ |
1790 |
|
int i,j; \ |
1791 |
|
for(j=0; j<(H); ++j) { \ |
1792 |
|
for(i=0; i<(W); ++i) printf( "0x%.2x ", (DATA)[i+j*(STR)] );\ |
1793 |
|
printf("\n"); \ |
1794 |
|
} \ |
1795 |
|
printf("---\n"); \ |
1796 |
|
} |
1797 |
|
|
1798 |
|
static const int yv12_CRCs[2][2] = { |
1799 |
|
{0x5cab7cf0,0xdab46541} |
1800 |
|
, {0xe8bae865,0x1faf77b7} |
1801 |
|
}; |
1802 |
|
|
1803 |
|
void test_yuv2() |
1804 |
|
{ |
1805 |
|
const int nb_tests = 800*speed_ref; |
1806 |
|
CPU *cpu; |
1807 |
|
uint8_t Src0[3][WIDTH*HEIGHT]; |
1808 |
|
uint8_t Dst0[3][WIDTH*HEIGHT]; |
1809 |
|
int with_uv, with_flip; |
1810 |
|
int i, j; |
1811 |
|
double t; |
1812 |
|
int tst, iCrc; |
1813 |
|
|
1814 |
|
colorspace_init(); |
1815 |
|
ieee_reseed(1); |
1816 |
|
for(i=0; i<(int)sizeof(Src0); ++i) Src0[0][i] = ieee_rand(0,255); |
1817 |
|
|
1818 |
|
printf( "\n === test YV2 ===\n" ); |
1819 |
|
for(with_flip=0; with_flip<=1; ++with_flip) { |
1820 |
|
for(with_uv=0; with_uv<=1; ++with_uv) { |
1821 |
|
init_cpu(&cpu_list[0]); |
1822 |
|
TEST_YV2(yv12_to_yv12_c, with_uv, with_flip); |
1823 |
|
printf(" yv12_to_yv12_c %.3f usec \tcrc32=0x%08x %s\n", |
1824 |
|
t, iCrc, (iCrc!=yv12_CRCs[with_flip][with_uv])?"| ERROR": "" ); |
1825 |
|
/* if (!with_uv) PRINT_NxN(Dst0[1], WIDTH/2, HEIGHT/2, WIDTH ); */ |
1826 |
|
|
1827 |
|
#ifdef ARCH_IS_IA32 |
1828 |
|
init_cpu(&cpu_list[1]); |
1829 |
|
TEST_YV2(yv12_to_yv12_mmx, with_uv, with_flip); |
1830 |
|
printf(" yv12_to_yv12_mmx %.3f usec \tcrc32=0x%08x %s\n", |
1831 |
|
t, iCrc, (iCrc!=yv12_CRCs[with_flip][with_uv])?"| ERROR": "" ); |
1832 |
|
/* if (!with_uv) PRINT_NxN(Dst0[1], WIDTH/2, HEIGHT/2, WIDTH ); */ |
1833 |
|
|
1834 |
|
TEST_YV2(yv12_to_yv12_xmm, with_uv, with_flip); |
1835 |
|
printf(" yv12_to_yv12_xmm %.3f usec \tcrc32=0x%08x %s\n", |
1836 |
|
t, iCrc, (iCrc!=yv12_CRCs[with_flip][with_uv])?"| ERROR": "" ); |
1837 |
|
#endif |
1838 |
|
} |
1839 |
|
|
1840 |
|
printf( " --- \n" ); |
1841 |
|
} |
1842 |
|
printf( " ===== \n" ); |
1843 |
|
} |
1844 |
|
|
1845 |
|
#undef WIDTH |
1846 |
|
#undef HEIGHT |
1847 |
|
#undef ENTER |
1848 |
|
#undef LEAVE |
1849 |
|
|
1850 |
/*********************************************************************/ |
/*********************************************************************/ |
1851 |
|
|
1852 |
static uint32_t __inline log2bin_v1(uint32_t value) |
static uint32_t __inline log2bin_v1(uint32_t value) |
1969 |
} |
} |
1970 |
|
|
1971 |
/********************************************************************* |
/********************************************************************* |
1972 |
|
* test compiler |
1973 |
|
*********************************************************************/ |
1974 |
|
|
1975 |
|
void test_compiler() { |
1976 |
|
int nb_err = 0; |
1977 |
|
int32_t v; |
1978 |
|
if (sizeof(uint16_t)<2) { |
1979 |
|
printf( "ERROR: sizeof(uint16_t)<2 !!\n" ); |
1980 |
|
nb_err++; |
1981 |
|
} |
1982 |
|
if (sizeof(int16_t)<2) { |
1983 |
|
printf( "ERROR: sizeof(int16_t)<2 !!\n" ); |
1984 |
|
nb_err++; |
1985 |
|
} |
1986 |
|
if (sizeof(uint8_t)!=1) { |
1987 |
|
printf( "ERROR: sizeof(uint8_t)!=1 !!\n" ); |
1988 |
|
nb_err++; |
1989 |
|
} |
1990 |
|
if (sizeof(int8_t)!=1) { |
1991 |
|
printf( "ERROR: sizeof(int8_t)!=1 !!\n" ); |
1992 |
|
nb_err++; |
1993 |
|
} |
1994 |
|
if (sizeof(uint32_t)<4) { |
1995 |
|
printf( "ERROR: sizeof(uint32_t)<4 !!\n" ); |
1996 |
|
nb_err++; |
1997 |
|
} |
1998 |
|
if (sizeof(int32_t)<4) { |
1999 |
|
printf( "ERROR: sizeof(int32_t)<4 !!\n" ); |
2000 |
|
nb_err++; |
2001 |
|
} |
2002 |
|
/* yes, i know, this test is silly. But better be safe than sorry. :) */ |
2003 |
|
for(v=1000; v>=0; v--) { |
2004 |
|
if ( (v>>2) != v/4) |
2005 |
|
nb_err++; |
2006 |
|
} |
2007 |
|
for(v=-1000; v!=-1; v++) { |
2008 |
|
if ( (v>>2) != (v/4)-!!(v%4)) |
2009 |
|
nb_err++; |
2010 |
|
} |
2011 |
|
if (nb_err!=0) { |
2012 |
|
printf( "ERROR! please post your platform/compiler specs to xvid-devel@xvid.org !\n" ); |
2013 |
|
} |
2014 |
|
} |
2015 |
|
|
2016 |
|
/********************************************************************* |
2017 |
|
* test SSIM functions |
2018 |
|
*********************************************************************/ |
2019 |
|
|
2020 |
|
typedef int (*lumfunc)(uint8_t* ptr, int stride); |
2021 |
|
typedef void (*csfunc)(uint8_t* ptro, uint8_t* ptrc, int stride, int lumo, int lumc, int* pdevo, int* pdevc, int* pcorr); |
2022 |
|
|
2023 |
|
extern int lum_8x8_c(uint8_t* ptr, int stride); |
2024 |
|
extern int lum_8x8_mmx(uint8_t* ptr, int stride); |
2025 |
|
extern int lum_2x8_c(uint8_t* ptr, int stride); |
2026 |
|
extern void consim_c(uint8_t* ptro, uint8_t* ptrc, int stride, int lumo, int lumc, int* pdevo, int* pdevc, int* pcorr); |
2027 |
|
extern void consim_mmx(uint8_t* ptro, uint8_t* ptrc, int stride, int lumo, int lumc, int* pdevo, int* pdevc, int* pcorr); |
2028 |
|
extern void consim_sse2(uint8_t* ptro, uint8_t* ptrc, int stride, int lumo, int lumc, int* pdevo, int* pdevc, int* pcorr); |
2029 |
|
|
2030 |
|
void test_SSIM() |
2031 |
|
{ |
2032 |
|
const int nb_tests = 3000*speed_ref; |
2033 |
|
int tst; |
2034 |
|
CPU *cpu; |
2035 |
|
int i; |
2036 |
|
int devs[3]; |
2037 |
|
long lumo, lumc; |
2038 |
|
DECLARE_ALIGNED_MATRIX(Ref1, 16, 16, uint8_t, 16); |
2039 |
|
DECLARE_ALIGNED_MATRIX(Ref2, 16, 16, uint8_t, 16); |
2040 |
|
lumfunc lum8x8; |
2041 |
|
lumfunc lum2x8; |
2042 |
|
csfunc csim; |
2043 |
|
|
2044 |
|
ieee_reseed(1); |
2045 |
|
printf( "\n ====== test SSIM ======\n" ); |
2046 |
|
for(i=0; i<16*16;++i) { |
2047 |
|
long v1, v2; |
2048 |
|
v1 = ieee_rand(-256, 511); |
2049 |
|
v2 = ieee_rand(-256, 511); |
2050 |
|
Ref1[i] = (v1<0) ? 0 : (v1>255) ? 255 : v1; |
2051 |
|
Ref2[i] = (v2<0) ? 0 : (v2>255) ? 255 : v2; |
2052 |
|
} |
2053 |
|
lumc = ieee_rand(0, 255); |
2054 |
|
lumo = ieee_rand(0, 255); |
2055 |
|
|
2056 |
|
for(cpu = cpu_list; cpu->name!=0; ++cpu) |
2057 |
|
{ |
2058 |
|
double t; |
2059 |
|
int m; |
2060 |
|
if (!init_cpu(cpu)) |
2061 |
|
continue; |
2062 |
|
lum8x8 = lum_8x8_c; |
2063 |
|
lum2x8 = lum_2x8_c; |
2064 |
|
csim = consim_c; |
2065 |
|
#ifdef ARCH_IS_IA32 |
2066 |
|
if (cpu->cpu & XVID_CPU_MMX){ |
2067 |
|
lum8x8 = lum_8x8_mmx; |
2068 |
|
csim = consim_mmx; |
2069 |
|
} |
2070 |
|
if (cpu->cpu & XVID_CPU_MMX){ |
2071 |
|
csim = consim_sse2; |
2072 |
|
} |
2073 |
|
#endif |
2074 |
|
t = gettime_usec(); |
2075 |
|
emms(); |
2076 |
|
for(tst=0; tst<nb_tests; ++tst) m = lum8x8(Ref1, 16); |
2077 |
|
emms(); |
2078 |
|
t = (gettime_usec() - t) / nb_tests; |
2079 |
|
printf("%s - ssim-lum8x8 %.3f usec m=%d %s\n", |
2080 |
|
cpu->name, t, m, |
2081 |
|
(m!=8230)?"| ERROR": "" ); |
2082 |
|
|
2083 |
|
t = gettime_usec(); |
2084 |
|
emms(); |
2085 |
|
for(tst=0; tst<nb_tests; ++tst) m = lum2x8(Ref1+8, 16); |
2086 |
|
emms(); |
2087 |
|
t = (gettime_usec() - t) / nb_tests; |
2088 |
|
printf("%s - ssim-lum2x8 %.3f usec m=%d %s\n", |
2089 |
|
cpu->name, t, m, |
2090 |
|
(m!=681)?"| ERROR": "" ); |
2091 |
|
|
2092 |
|
t = gettime_usec(); |
2093 |
|
emms(); |
2094 |
|
for(tst=0; tst<nb_tests; ++tst) csim(Ref1, Ref2, 16, lumo, lumc, devs+0, devs+1, devs+2); |
2095 |
|
emms(); |
2096 |
|
t = (gettime_usec() - t) / nb_tests; |
2097 |
|
printf("%s - ssim-consim %.3f usec devs=[0x%x 0x%x 0x%x] %s\n", |
2098 |
|
cpu->name, t, devs[0], devs[1], devs[2], |
2099 |
|
(devs[0]!=0x1bdf0f || devs[1]!=0x137258 || devs[2]!=0xcdb13)?"| ERROR": "" ); |
2100 |
|
printf( " --- \n" ); |
2101 |
|
} |
2102 |
|
} |
2103 |
|
|
2104 |
|
/********************************************************************* |
2105 |
|
* test bitstream functions |
2106 |
|
*********************************************************************/ |
2107 |
|
|
2108 |
|
#define BIT_BUF_SIZE 2000 |
2109 |
|
|
2110 |
|
static void test_bits() |
2111 |
|
{ |
2112 |
|
const int nb_tests = 50*speed_ref; |
2113 |
|
int tst; |
2114 |
|
uint32_t Crc; |
2115 |
|
uint8_t Buf[BIT_BUF_SIZE]; |
2116 |
|
uint32_t Extracted[BIT_BUF_SIZE*8]; /* worst case: bits read 1 by 1 */ |
2117 |
|
int Lens[BIT_BUF_SIZE*8]; |
2118 |
|
double t1; |
2119 |
|
|
2120 |
|
|
2121 |
|
printf( "\n === test bitstream ===\n" ); |
2122 |
|
ieee_reseed(1); |
2123 |
|
Crc = 0; |
2124 |
|
|
2125 |
|
t1 = gettime_usec(); |
2126 |
|
for(tst=0; tst<nb_tests; ++tst) { |
2127 |
|
Bitstream bs; |
2128 |
|
int m, m2, l, l2; |
2129 |
|
|
2130 |
|
for(l=0; l<BIT_BUF_SIZE; ++l) |
2131 |
|
Buf[l] = (uint8_t)ieee_rand(0,255); |
2132 |
|
|
2133 |
|
l = BIT_BUF_SIZE - ieee_rand(1,BIT_BUF_SIZE/10); |
2134 |
|
BitstreamInit(&bs, (void*)(Buf+BIT_BUF_SIZE-l), l); |
2135 |
|
|
2136 |
|
|
2137 |
|
BitstreamReset(&bs); |
2138 |
|
for(l2=l*8, m=0; l2>0; m++) { |
2139 |
|
const int b = ieee_rand(1,32); |
2140 |
|
Lens[m] = b; |
2141 |
|
l2 -= b; |
2142 |
|
if (l2<0) break; |
2143 |
|
Extracted[m] = BitstreamShowBits(&bs, b); |
2144 |
|
BitstreamSkip(&bs, b); |
2145 |
|
// printf( "<= %d: %d 0x%x\n", m, b, Extracted[m]); |
2146 |
|
} |
2147 |
|
|
2148 |
|
BitstreamReset(&bs); |
2149 |
|
for(m2=0; m2<m; ++m2) { |
2150 |
|
const int b = Lens[m2]; |
2151 |
|
const uint32_t v = BitstreamGetBits(&bs, b); |
2152 |
|
Crc |= (v!=Extracted[m2]); |
2153 |
|
// printf( "=> %d: %d 0x%x %c\n", m2, b, v, " *"[Crc]); |
2154 |
|
} |
2155 |
|
} |
2156 |
|
t1 = (gettime_usec() - t1) / nb_tests; |
2157 |
|
printf(" test_bits %.3f usec %s\n", t1, (Crc!=0)?"| ERROR": "" ); |
2158 |
|
} |
2159 |
|
|
2160 |
|
/********************************************************************* |
2161 |
* main |
* main |
2162 |
*********************************************************************/ |
*********************************************************************/ |
2163 |
|
|
2181 |
else if (!strcmp(argv[c], "-c")) cpu_mask = 0 /* PLAIN_C */ | XVID_CPU_FORCE; |
else if (!strcmp(argv[c], "-c")) cpu_mask = 0 /* PLAIN_C */ | XVID_CPU_FORCE; |
2182 |
else if (!strcmp(argv[c], "-mmx")) cpu_mask = XVID_CPU_MMX | XVID_CPU_FORCE; |
else if (!strcmp(argv[c], "-mmx")) cpu_mask = XVID_CPU_MMX | XVID_CPU_FORCE; |
2183 |
else if (!strcmp(argv[c], "-mmxext")) cpu_mask = XVID_CPU_MMXEXT | XVID_CPU_MMX | XVID_CPU_FORCE; |
else if (!strcmp(argv[c], "-mmxext")) cpu_mask = XVID_CPU_MMXEXT | XVID_CPU_MMX | XVID_CPU_FORCE; |
2184 |
else if (!strcmp(argv[c], "-sse2")) cpu_mask = XVID_CPU_SSE2 | XVID_CPU_MMX | XVID_CPU_FORCE; |
else if (!strcmp(argv[c], "-sse2")) cpu_mask = XVID_CPU_SSE2 | XVID_CPU_MMXEXT | XVID_CPU_MMX | XVID_CPU_FORCE; |
2185 |
|
else if (!strcmp(argv[c], "-sse3")) cpu_mask = XVID_CPU_SSE3 | XVID_CPU_SSE2 | XVID_CPU_MMXEXT | XVID_CPU_MMX | XVID_CPU_FORCE; |
2186 |
|
else if (!strcmp(argv[c], "-sse4")) cpu_mask = XVID_CPU_SSE41 | XVID_CPU_SSE3 | XVID_CPU_SSE2 | XVID_CPU_MMXEXT | XVID_CPU_MMX | XVID_CPU_FORCE; |
2187 |
else if (!strcmp(argv[c], "-3dnow")) cpu_mask = XVID_CPU_3DNOW | XVID_CPU_FORCE; |
else if (!strcmp(argv[c], "-3dnow")) cpu_mask = XVID_CPU_3DNOW | XVID_CPU_FORCE; |
2188 |
else if (!strcmp(argv[c], "-3dnowe")) cpu_mask = XVID_CPU_3DNOW | XVID_CPU_3DNOWEXT | XVID_CPU_FORCE; |
else if (!strcmp(argv[c], "-3dnowe")) cpu_mask = XVID_CPU_3DNOW | XVID_CPU_3DNOWEXT | XVID_CPU_FORCE; |
2189 |
else if (!strcmp(argv[c], "-altivec")) cpu_mask = XVID_CPU_ALTIVEC | XVID_CPU_FORCE; |
else if (!strcmp(argv[c], "-altivec")) cpu_mask = XVID_CPU_ALTIVEC | XVID_CPU_FORCE; |
2227 |
if (what==0 || what==10) test_sse(); |
if (what==0 || what==10) test_sse(); |
2228 |
if (what==0 || what==11) test_log2bin(); |
if (what==0 || what==11) test_log2bin(); |
2229 |
if (what==0 || what==12) test_gcd(); |
if (what==0 || what==12) test_gcd(); |
2230 |
|
if (what==0 || what==13) test_compiler(); |
2231 |
|
if (what==0 || what==14) test_yuv(); |
2232 |
|
if (what==0 || what==15) test_SSIM(); |
2233 |
|
if (what==0 || what==16) test_yuv2(); |
2234 |
|
if (what==0 || what==17) test_bits(); |
2235 |
|
|
2236 |
if (what==7) { |
if (what==7) { |
2237 |
test_IEEE1180_compliance(-256, 255, 1); |
test_IEEE1180_compliance(-256, 255, 1); |