1 |
/************************************************************************** |
/***************************************************************************** |
2 |
* |
* |
3 |
* XVID MPEG-4 VIDEO CODEC - Unit tests and benches |
* XVID MPEG-4 VIDEO CODEC |
4 |
|
* - Unit tests and benches - |
5 |
|
* |
6 |
|
* Copyright(C) 2002 Pascal Massimino <skal@planet-d.net> |
7 |
* |
* |
8 |
* This program is free software; you can redistribute it and/or modify |
* This program is free software; you can redistribute it and/or modify |
9 |
* it under the terms of the GNU General Public License as published by |
* it under the terms of the GNU General Public License as published by |
17 |
* |
* |
18 |
* You should have received a copy of the GNU General Public License |
* You should have received a copy of the GNU General Public License |
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., 675 Mass Ave, Cambridge, MA 02139, USA. |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
21 |
|
* |
22 |
|
* $Id$ |
23 |
* |
* |
24 |
*************************************************************************/ |
****************************************************************************/ |
25 |
|
|
26 |
/************************************************************************ |
/***************************************************************************** |
27 |
* |
* |
28 |
* 'Reference' output is at the end of file. |
* 'Reference' output is at the end of file. |
29 |
* Don't take the checksums and crc too seriouly, they aren't |
* Don't take the checksums and crc too seriouly, they aren't |
32 |
* compiles with something like: |
* compiles with something like: |
33 |
* gcc -o xvid_bench xvid_bench.c -I../src/ -lxvidcore -lm |
* gcc -o xvid_bench xvid_bench.c -I../src/ -lxvidcore -lm |
34 |
* |
* |
35 |
* History: |
****************************************************************************/ |
|
* |
|
|
* 06.06.2002 initial coding -Skal- |
|
|
* |
|
|
*************************************************************************/ |
|
36 |
|
|
37 |
#include <stdio.h> |
#include <stdio.h> |
38 |
#include <stdlib.h> |
#include <stdlib.h> |
39 |
#include <sys/time.h> // for gettimeofday |
#include <string.h> /* for memset */ |
|
#include <string.h> // for memset |
|
40 |
#include <assert.h> |
#include <assert.h> |
41 |
|
|
42 |
|
#ifndef WIN32 |
43 |
|
#include <sys/time.h> /* for gettimeofday */ |
44 |
|
#else |
45 |
|
#include <time.h> |
46 |
|
#endif |
47 |
|
|
48 |
|
|
49 |
#include "xvid.h" |
#include "xvid.h" |
50 |
|
|
51 |
// inner guts |
// inner guts |
54 |
#include "image/colorspace.h" |
#include "image/colorspace.h" |
55 |
#include "image/interpolate8x8.h" |
#include "image/interpolate8x8.h" |
56 |
#include "utils/mem_transfer.h" |
#include "utils/mem_transfer.h" |
57 |
#include "quant/quant_h263.h" |
#include "quant/quant.h" |
|
#include "quant/quant_mpeg4.h" |
|
58 |
#include "motion/sad.h" |
#include "motion/sad.h" |
59 |
#include "utils/emms.h" |
#include "utils/emms.h" |
60 |
#include "utils/timer.h" |
#include "utils/timer.h" |
62 |
#include "bitstream/cbp.h" |
#include "bitstream/cbp.h" |
63 |
|
|
64 |
#include <math.h> |
#include <math.h> |
65 |
const int speed_ref = 100; // on slow machines, decrease this value |
|
66 |
|
#ifndef M_PI |
67 |
|
#define M_PI 3.14159265358979323846 |
68 |
|
#endif |
69 |
|
|
70 |
|
const int speed_ref = 100; /* on slow machines, decrease this value */ |
71 |
|
|
72 |
/********************************************************************* |
/********************************************************************* |
73 |
* misc |
* misc |
76 |
/* returns time in micro-s*/ |
/* returns time in micro-s*/ |
77 |
double gettime_usec() |
double gettime_usec() |
78 |
{ |
{ |
79 |
|
#ifndef WIN32 |
80 |
struct timeval tv; |
struct timeval tv; |
81 |
gettimeofday(&tv, 0); |
gettimeofday(&tv, 0); |
82 |
return tv.tv_sec*1.0e6f + tv.tv_usec; |
return tv.tv_sec*1.0e6 + tv.tv_usec; |
83 |
|
#else |
84 |
|
clock_t clk; |
85 |
|
clk = clock(); |
86 |
|
return clk * 1000000 / CLOCKS_PER_SEC; |
87 |
|
#endif |
88 |
} |
} |
89 |
|
|
90 |
/* returns squared deviates (mean(v*v)-mean(v)^2) of a 8x8 block */ |
/* returns squared deviates (mean(v*v)-mean(v)^2) of a 8x8 block */ |
114 |
|
|
115 |
CPU cpu_list[] = |
CPU cpu_list[] = |
116 |
{ { "PLAINC", 0 } |
{ { "PLAINC", 0 } |
117 |
|
#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_3DNOWEXT } |
, { "3DNOWE", XVID_CPU_3DNOWEXT } |
123 |
|
#endif |
124 |
|
//, { "IA64 ", XVID_CPU_IA64 } |
125 |
//, { "TSC ", XVID_CPU_TSC } |
//, { "TSC ", XVID_CPU_TSC } |
126 |
, { 0, 0 } } |
, { 0, 0 } }; |
127 |
|
|
128 |
, cpu_short_list[] = |
CPU cpu_short_list[] = |
129 |
{ { "PLAINC", 0 } |
{ { "PLAINC", 0 } |
130 |
|
#ifdef ARCH_IS_IA32 |
131 |
, { "MMX ", XVID_CPU_MMX } |
, { "MMX ", XVID_CPU_MMX } |
132 |
//, { "MMXEXT", XVID_CPU_MMXEXT | XVID_CPU_MMX } |
//, { "MMXEXT", XVID_CPU_MMXEXT | XVID_CPU_MMX } |
133 |
, { 0, 0 } } |
#endif |
134 |
|
//, { "IA64 ", XVID_CPU_IA64 } |
135 |
|
, { 0, 0 } }; |
136 |
|
|
137 |
, cpu_short_list2[] = |
CPU cpu_short_list2[] = |
138 |
{ { "PLAINC", 0 } |
{ { "PLAINC", 0 } |
139 |
|
#ifdef ARCH_IS_IA32 |
140 |
, { "MMX ", XVID_CPU_MMX } |
, { "MMX ", XVID_CPU_MMX } |
141 |
, { "SSE2 ", XVID_CPU_SSE2 | XVID_CPU_MMX } |
, { "SSE2 ", XVID_CPU_SSE2 | XVID_CPU_MMX } |
142 |
|
#endif |
143 |
, { 0, 0 } }; |
, { 0, 0 } }; |
144 |
|
|
145 |
|
|
146 |
int init_cpu(CPU *cpu) |
int init_cpu(CPU *cpu) |
147 |
{ |
{ |
148 |
int xerr, cpu_type; |
int xerr, cpu_type; |
149 |
XVID_INIT_PARAM xinit; |
xvid_gbl_init_t xinit; |
150 |
|
|
151 |
cpu_type = check_cpu_features() & cpu->cpu; |
memset(&xinit, 0, sizeof(xinit)); |
152 |
xinit.cpu_flags = cpu_type | XVID_CPU_FORCE; |
xinit.cpu_flags = cpu->cpu | XVID_CPU_FORCE; |
153 |
// xinit.cpu_flags = XVID_CPU_MMX | XVID_CPU_FORCE; |
xinit.version = XVID_VERSION; |
154 |
xerr = xvid_init(NULL, 0, &xinit, NULL); |
xerr = xvid_global(NULL, 0, &xinit, NULL); |
155 |
if (cpu->cpu>0 && (cpu_type==0 || xerr!=XVID_ERR_OK)) { |
if (cpu->cpu>0 && xerr==XVID_ERR_FAIL) { |
156 |
printf( "%s - skipped...\n", cpu->name ); |
printf( "%s - skipped...\n", cpu->name ); |
157 |
return 0; |
return 0; |
158 |
} |
} |
308 |
const int nb_tests = 2000*speed_ref; |
const int nb_tests = 2000*speed_ref; |
309 |
CPU *cpu; |
CPU *cpu; |
310 |
const uint8_t Src0[16*9] = { |
const uint8_t Src0[16*9] = { |
311 |
// try to have every possible combinaison of rounding... |
/* try to have every possible combinaison of rounding... */ |
312 |
0, 0, 1, 0, 2, 0, 3, 0, 4 ,0,0,0, 0,0,0,0 |
0, 0, 1, 0, 2, 0, 3, 0, 4 ,0,0,0, 0,0,0,0 |
313 |
, 0, 1, 1, 1, 2, 1, 3, 1, 3 ,0,0,0, 0,0,0,0 |
, 0, 1, 1, 1, 2, 1, 3, 1, 3 ,0,0,0, 0,0,0,0 |
314 |
, 0, 2, 1, 2, 2, 2, 3, 2, 2 ,0,0,0, 0,0,0,0 |
, 0, 2, 1, 2, 2, 2, 3, 2, 2 ,0,0,0, 0,0,0,0 |
358 |
if (iCrc!=8103) printf( "*** CRC ERROR! ***\n" ); |
if (iCrc!=8103) printf( "*** CRC ERROR! ***\n" ); |
359 |
|
|
360 |
|
|
361 |
// this is a new function, as of 06.06.2002 |
/* this is a new function, as of 06.06.2002 */ |
362 |
#if 0 |
#if 0 |
363 |
TEST_MB2(interpolate8x8_avrg); |
TEST_MB2(interpolate8x8_avrg); |
364 |
printf( "%s - interpolate8x8_c %.3f usec iCrc=%d\n", cpu->name, t, iCrc ); |
printf( "%s - interpolate8x8_c %.3f usec iCrc=%d\n", cpu->name, t, iCrc ); |
446 |
|
|
447 |
printf( "\n === test transfer ===\n" ); |
printf( "\n === test transfer ===\n" ); |
448 |
|
|
449 |
for(cpu = cpu_short_list; cpu->name!=0; ++cpu) |
for(cpu = cpu_list; cpu->name!=0; ++cpu) |
450 |
{ |
{ |
451 |
double t, overhead; |
double t, overhead; |
452 |
int tst, s; |
int tst, s; |
534 |
|
|
535 |
printf( "\n ===== test quant =====\n" ); |
printf( "\n ===== test quant =====\n" ); |
536 |
|
|
537 |
// we deliberately enfringe the norm's specified range [-127,127], |
/* we deliberately enfringe the norm's specified range [-127,127], */ |
538 |
// to test the robustness of the iquant module |
/* to test the robustness of the iquant module */ |
539 |
for(i=0; i<64; ++i) { |
for(i=0; i<64; ++i) { |
540 |
Src[i] = 1 + (i-32) * (i&6); |
Src[i] = 1 + (i-32) * (i&6); |
541 |
Dst[i] = 0; |
Dst[i] = 0; |
542 |
} |
} |
543 |
|
|
544 |
for(cpu = cpu_short_list; cpu->name!=0; ++cpu) |
for(cpu = cpu_list; cpu->name!=0; ++cpu) |
545 |
{ |
{ |
546 |
double t, overhead; |
double t, overhead; |
547 |
int tst, q; |
int tst, q; |
560 |
overhead += gettime_usec(); |
overhead += gettime_usec(); |
561 |
|
|
562 |
#if 1 |
#if 1 |
563 |
TEST_QUANT2(quant4_intra, Dst, Src); |
TEST_QUANT2(quant_mpeg_intra, Dst, Src); |
564 |
printf( "%s - quant4_intra %.3f usec crc=%d\n", cpu->name, t, s ); |
printf( "%s - quant_mpeg_intra %.3f usec crc=%d\n", cpu->name, t, s ); |
565 |
if (s!=29809) printf( "*** CRC ERROR! ***\n" ); |
if (s!=29809) printf( "*** CRC ERROR! ***\n" ); |
566 |
|
|
567 |
TEST_QUANT(quant4_inter, Dst, Src); |
TEST_QUANT(quant_mpeg_inter, Dst, Src); |
568 |
printf( "%s - quant4_inter %.3f usec crc=%d\n", cpu->name, t, s ); |
printf( "%s - quant_mpeg_inter %.3f usec crc=%d\n", cpu->name, t, s ); |
569 |
if (s!=12574) printf( "*** CRC ERROR! ***\n" ); |
if (s!=12574) printf( "*** CRC ERROR! ***\n" ); |
570 |
#endif |
#endif |
571 |
#if 1 |
#if 1 |
572 |
TEST_QUANT2(dequant4_intra, Dst, Src); |
TEST_QUANT2(dequant_mpeg_intra, Dst, Src); |
573 |
printf( "%s - dequant4_intra %.3f usec crc=%d\n", cpu->name, t, s ); |
printf( "%s - dequant_mpeg_intra %.3f usec crc=%d\n", cpu->name, t, s ); |
574 |
if (s!=24052) printf( "*** CRC ERROR! ***\n" ); |
if (s!=24052) printf( "*** CRC ERROR! ***\n" ); |
575 |
|
|
576 |
TEST_QUANT(dequant4_inter, Dst, Src); |
TEST_QUANT(dequant_mpeg_inter, Dst, Src); |
577 |
printf( "%s - dequant4_inter %.3f usec crc=%d\n", cpu->name, t, s ); |
printf( "%s - dequant_mpeg_inter %.3f usec crc=%d\n", cpu->name, t, s ); |
578 |
if (s!=63847) printf( "*** CRC ERROR! ***\n" ); |
if (s!=63847) printf( "*** CRC ERROR! ***\n" ); |
579 |
#endif |
#endif |
580 |
#if 1 |
#if 1 |
581 |
TEST_QUANT2(quant_intra, Dst, Src); |
TEST_QUANT2(quant_h263_intra, Dst, Src); |
582 |
printf( "%s - quant_intra %.3f usec crc=%d\n", cpu->name, t, s ); |
printf( "%s - quant_h263_intra %.3f usec crc=%d\n", cpu->name, t, s ); |
583 |
if (s!=25662) printf( "*** CRC ERROR! ***\n" ); |
if (s!=25662) printf( "*** CRC ERROR! ***\n" ); |
584 |
|
|
585 |
TEST_QUANT(quant_inter, Dst, Src); |
TEST_QUANT(quant_h263_inter, Dst, Src); |
586 |
printf( "%s - quant_inter %.3f usec crc=%d\n", cpu->name, t, s ); |
printf( "%s - quant_h263_inter %.3f usec crc=%d\n", cpu->name, t, s ); |
587 |
if (s!=23972) printf( "*** CRC ERROR! ***\n" ); |
if (s!=23972) printf( "*** CRC ERROR! ***\n" ); |
588 |
#endif |
#endif |
589 |
#if 1 |
#if 1 |
590 |
TEST_QUANT2(dequant_intra, Dst, Src); |
TEST_QUANT2(dequant_h263_intra, Dst, Src); |
591 |
printf( "%s - dequant_intra %.3f usec crc=%d\n", cpu->name, t, s ); |
printf( "%s - dequant_h263_intra %.3f usec crc=%d\n", cpu->name, t, s ); |
592 |
if (s!=49900) printf( "*** CRC ERROR! ***\n" ); |
if (s!=49900) printf( "*** CRC ERROR! ***\n" ); |
593 |
|
|
594 |
TEST_QUANT(dequant_inter, Dst, Src); |
TEST_QUANT(dequant_h263_inter, Dst, Src); |
595 |
printf( "%s - dequant_inter %.3f usec crc=%d\n", cpu->name, t, s ); |
printf( "%s - dequant_h263_inter %.3f usec crc=%d\n", cpu->name, t, s ); |
596 |
if (s!=48899) printf( "*** CRC ERROR! ***\n" ); |
if (s!=48899) printf( "*** CRC ERROR! ***\n" ); |
597 |
#endif |
#endif |
598 |
printf( " --- \n" ); |
printf( " --- \n" ); |
622 |
printf( "\n ===== test cbp =====\n" ); |
printf( "\n ===== test cbp =====\n" ); |
623 |
|
|
624 |
for(i=0; i<6*64; ++i) { |
for(i=0; i<6*64; ++i) { |
625 |
Src1[i] = (i*i*3/8192)&(i/64)&1; // 'random' |
Src1[i] = (i*i*3/8192)&(i/64)&1; /* 'random' */ |
626 |
Src2[i] = (i<3*64); // half-full |
Src2[i] = (i<3*64); /* half-full */ |
627 |
Src3[i] = ((i+32)>3*64); |
Src3[i] = ((i+32)>3*64); |
628 |
Src4[i] = (i==(3*64+2) || i==(5*64+9)); |
Src4[i] = (i==(3*64+2) || i==(5*64+9)); |
629 |
} |
} |
630 |
|
|
631 |
for(cpu = cpu_short_list2; cpu->name!=0; ++cpu) |
for(cpu = cpu_list; cpu->name!=0; ++cpu) |
632 |
{ |
{ |
633 |
double t; |
double t; |
634 |
int tst, cbp; |
int tst, cbp; |
778 |
} |
} |
779 |
} |
} |
780 |
|
|
781 |
////////////////////////////////////////////////////////// |
///* ////////////////////////////////////////////////////// */ |
782 |
/* Pseudo-random generator specified by IEEE 1180 */ |
/* Pseudo-random generator specified by IEEE 1180 */ |
783 |
|
|
784 |
static long ieee_seed = 1; |
static long ieee_seed = 1; |
865 |
static const double ILimits[5] = { 1., 0.06, 0.02, 0.015, 0.0015 }; |
static const double ILimits[5] = { 1., 0.06, 0.02, 0.015, 0.0015 }; |
866 |
int Loops = 10000; |
int Loops = 10000; |
867 |
int i, m, n; |
int i, m, n; |
868 |
short Blk0[64]; // reference |
short Blk0[64]; /* reference */ |
869 |
short Blk[64], iBlk[64]; |
short Blk[64], iBlk[64]; |
870 |
short Ref_FDCT[64]; |
short Ref_FDCT[64]; |
871 |
short Ref_IDCT[64]; |
short Ref_IDCT[64]; |
872 |
|
|
873 |
STATS_8x8 FStats; // forward dct stats |
STATS_8x8 FStats; /* forward dct stats */ |
874 |
STATS_8x8 IStats; // inverse dct stats |
STATS_8x8 IStats; /* inverse dct stats */ |
875 |
|
|
876 |
CPU *cpu; |
CPU *cpu; |
877 |
|
|
894 |
for(i=0; i<64; ++i) |
for(i=0; i<64; ++i) |
895 |
Blk0[i] = (short)ieee_rand(Min,Max) * Sign; |
Blk0[i] = (short)ieee_rand(Min,Max) * Sign; |
896 |
|
|
897 |
// hmm, I'm not quite sure this is exactly |
/* hmm, I'm not quite sure this is exactly */ |
898 |
// the tests described in the norm. check... |
/* the tests described in the norm. check... */ |
899 |
|
|
900 |
memcpy(Ref_FDCT, Blk0, 64*sizeof(short)); |
memcpy(Ref_FDCT, Blk0, 64*sizeof(short)); |
901 |
ref_fdct(Ref_FDCT); |
ref_fdct(Ref_FDCT); |
922 |
|
|
923 |
printf( "\n -- FDCT report --\n" ); |
printf( "\n -- FDCT report --\n" ); |
924 |
// print_stats(&FStats); |
// print_stats(&FStats); |
925 |
report_stats(&FStats, 0); // so far I know, IEEE1180 says nothing for fdct |
report_stats(&FStats, 0); /* so far I know, IEEE1180 says nothing for fdct */ |
926 |
|
|
927 |
for(i=0; i<64; i++) Blk[i] = 0; |
for(i=0; i<64; i++) Blk[i] = 0; |
928 |
emms(); fdct(Blk); emms(); |
emms(); fdct(Blk); emms(); |
944 |
|
|
945 |
void test_dct_saturation(int Min, int Max) |
void test_dct_saturation(int Min, int Max) |
946 |
{ |
{ |
947 |
// test behaviour on input range fringe |
/* test behaviour on input range fringe */ |
948 |
|
|
949 |
int i, n, p; |
int i, n, p; |
950 |
CPU *cpu; |
CPU *cpu; |
951 |
// const short IDCT_MAX = 2047; // 12bits input |
// const short IDCT_MAX = 2047; /* 12bits input */ |
952 |
// const short IDCT_MIN = -2048; |
// const short IDCT_MIN = -2048; |
953 |
// const short IDCT_OUT = 256; // 9bits ouput |
// const short IDCT_OUT = 256; /* 9bits ouput */ |
954 |
const int Partitions = 4; |
const int Partitions = 4; |
955 |
const int Loops = 10000 / Partitions; |
const int Loops = 10000 / Partitions; |
956 |
|
|
967 |
printf( "\n===== IEEE test for %s Min=%d Max=%d =====\n", |
printf( "\n===== IEEE test for %s Min=%d Max=%d =====\n", |
968 |
cpu->name, Min, Max ); |
cpu->name, Min, Max ); |
969 |
|
|
970 |
// FDCT tests // |
/* FDCT tests // */ |
971 |
|
|
972 |
init_stats(&Stats); |
init_stats(&Stats); |
973 |
|
|
974 |
// test each computation channels separately |
/* test each computation channels separately */ |
975 |
for(i=0; i<64; i++) Blk[i] = Blk0[i] = ((i/8)==(i%8)) ? Max : 0; |
for(i=0; i<64; i++) Blk[i] = Blk0[i] = ((i/8)==(i%8)) ? Max : 0; |
976 |
ref_fdct(Blk0); |
ref_fdct(Blk0); |
977 |
emms(); fdct(Blk); emms(); |
emms(); fdct(Blk); emms(); |
982 |
emms(); fdct(Blk); emms(); |
emms(); fdct(Blk); emms(); |
983 |
store_stats(&Stats, Blk, Blk0); |
store_stats(&Stats, Blk, Blk0); |
984 |
|
|
985 |
// randomly saturated inputs |
/* randomly saturated inputs */ |
986 |
for(p=0; p<Partitions; ++p) |
for(p=0; p<Partitions; ++p) |
987 |
{ |
{ |
988 |
for(n=0; n<Loops; ++n) |
for(n=0; n<Loops; ++n) |
998 |
report_stats(&Stats, 0); |
report_stats(&Stats, 0); |
999 |
|
|
1000 |
|
|
1001 |
// IDCT tests // |
/* IDCT tests // */ |
1002 |
#if 0 |
#if 0 |
1003 |
// no finished yet |
/* no finished yet */ |
1004 |
|
|
1005 |
init_stats(&Stats); |
init_stats(&Stats); |
1006 |
|
|
1007 |
// test each computation channel separately |
/* test each computation channel separately */ |
1008 |
for(i=0; i<64; i++) Blk[i] = Blk0[i] = ((i/8)==(i%8)) ? IDCT_MAX : 0; |
for(i=0; i<64; i++) Blk[i] = Blk0[i] = ((i/8)==(i%8)) ? IDCT_MAX : 0; |
1009 |
ref_idct(Blk0); |
ref_idct(Blk0); |
1010 |
emms(); idct(Blk); emms(); |
emms(); idct(Blk); emms(); |
1017 |
for(i=0; i<64; i++) { CLAMP(Blk0[i], IDCT_OUT); CLAMP(Blk[i], IDCT_OUT); } |
for(i=0; i<64; i++) { CLAMP(Blk0[i], IDCT_OUT); CLAMP(Blk[i], IDCT_OUT); } |
1018 |
store_stats(&Stats, Blk, Blk0); |
store_stats(&Stats, Blk, Blk0); |
1019 |
|
|
1020 |
// randomly saturated inputs |
/* randomly saturated inputs */ |
1021 |
for(p=0; p<Partitions; ++p) |
for(p=0; p<Partitions; ++p) |
1022 |
{ |
{ |
1023 |
for(n=0; n<Loops; ++n) |
for(n=0; n<Loops; ++n) |
1047 |
FILE *f = 0; |
FILE *f = 0; |
1048 |
void *dechandle = 0; |
void *dechandle = 0; |
1049 |
int xerr; |
int xerr; |
1050 |
XVID_INIT_PARAM xinit; |
xvid_gbl_init_t xinit; |
1051 |
XVID_DEC_PARAM xparam; |
xvid_dec_create_t xparam; |
1052 |
XVID_DEC_FRAME xframe; |
xvid_dec_frame_t xframe; |
1053 |
double t = 0.; |
double t = 0.; |
1054 |
int nb = 0; |
int nb = 0; |
1055 |
uint8_t *buf = 0; |
uint8_t *buf = 0; |
1057 |
int buf_size, pos; |
int buf_size, pos; |
1058 |
uint32_t chksum = 0; |
uint32_t chksum = 0; |
1059 |
|
|
1060 |
|
memset(&xinit, 0, sizeof(xinit)); |
1061 |
xinit.cpu_flags = XVID_CPU_MMX | XVID_CPU_FORCE; |
xinit.cpu_flags = XVID_CPU_MMX | XVID_CPU_FORCE; |
1062 |
xvid_init(NULL, 0, &xinit, NULL); |
xinit.version = XVID_VERSION; |
1063 |
printf( "API version: %d, core build:%d\n", xinit.api_version, xinit.core_build); |
xvid_global(NULL, 0, &xinit, NULL); |
|
|
|
1064 |
|
|
1065 |
|
memset(&xparam, 0, sizeof(xparam)); |
1066 |
xparam.width = width; |
xparam.width = width; |
1067 |
xparam.height = height; |
xparam.height = height; |
1068 |
|
xparam.version = XVID_VERSION; |
1069 |
xerr = xvid_decore(NULL, XVID_DEC_CREATE, &xparam, NULL); |
xerr = xvid_decore(NULL, XVID_DEC_CREATE, &xparam, NULL); |
1070 |
if (xerr!=XVID_ERR_OK) { |
if (xerr==XVID_ERR_FAIL) { |
1071 |
printf("can't init decoder (err=%d)\n", xerr); |
printf("can't init decoder (err=%d)\n", xerr); |
1072 |
return; |
return; |
1073 |
} |
} |
1088 |
} |
} |
1089 |
else printf( "Input size: %d\n", buf_size); |
else printf( "Input size: %d\n", buf_size); |
1090 |
|
|
1091 |
buf = malloc(buf_size); // should be enuf' |
buf = malloc(buf_size); /* should be enuf' */ |
1092 |
rgb_out = calloc(4, width*height); // <-room for _RGB24 |
rgb_out = calloc(4, width*height); /* <-room for _RGB24 */ |
1093 |
if (buf==0 || rgb_out==0) { |
if (buf==0 || rgb_out==0) { |
1094 |
printf( "malloc failed!\n" ); |
printf( "malloc failed!\n" ); |
1095 |
goto End; |
goto End; |
1104 |
pos = 0; |
pos = 0; |
1105 |
t = -gettime_usec(); |
t = -gettime_usec(); |
1106 |
while(1) { |
while(1) { |
1107 |
|
memset(&xframe, 0, sizeof(xframe)); |
1108 |
|
xframe.version = XVID_VERSION; |
1109 |
xframe.bitstream = buf + pos; |
xframe.bitstream = buf + pos; |
1110 |
xframe.length = buf_size - pos; |
xframe.length = buf_size - pos; |
1111 |
xframe.image = rgb_out; |
xframe.output.plane[0] = rgb_out; |
1112 |
xframe.stride = width; |
xframe.output.stride[0] = width; |
1113 |
xframe.colorspace = XVID_CSP_RGB24; |
xframe.output.csp = XVID_CSP_BGR; |
1114 |
xerr = xvid_decore(dechandle, XVID_DEC_DECODE, &xframe, 0); |
xerr = xvid_decore(dechandle, XVID_DEC_DECODE, &xframe, 0); |
1115 |
nb++; |
nb++; |
1116 |
pos += xframe.length; |
pos += xframe.length; |
1121 |
} |
} |
1122 |
if (pos==buf_size) |
if (pos==buf_size) |
1123 |
break; |
break; |
1124 |
if (xerr!=XVID_ERR_OK) { |
if (xerr==XVID_ERR_FAIL) { |
1125 |
printf("decoding failed for frame #%d (err=%d)!\n", nb, xerr); |
printf("decoding failed for frame #%d (err=%d)!\n", nb, xerr); |
1126 |
break; |
break; |
1127 |
} |
} |
1137 |
if (buf!=0) free(buf); |
if (buf!=0) free(buf); |
1138 |
if (dechandle!=0) { |
if (dechandle!=0) { |
1139 |
xerr= xvid_decore(dechandle, XVID_DEC_DESTROY, NULL, NULL); |
xerr= xvid_decore(dechandle, XVID_DEC_DESTROY, NULL, NULL); |
1140 |
if (xerr!=XVID_ERR_OK) |
if (xerr==XVID_ERR_FAIL) |
1141 |
printf("destroy-decoder failed (err=%d)!\n", xerr); |
printf("destroy-decoder failed (err=%d)!\n", xerr); |
1142 |
} |
} |
1143 |
if (f!=0) fclose(f); |
if (f!=0) fclose(f); |
1153 |
|
|
1154 |
printf( "\n ===== (de)quant4_intra saturation bug? =====\n" ); |
printf( "\n ===== (de)quant4_intra saturation bug? =====\n" ); |
1155 |
|
|
1156 |
for(cpu = cpu_short_list; cpu->name!=0; ++cpu) |
for(cpu = cpu_list; cpu->name!=0; ++cpu) |
1157 |
{ |
{ |
1158 |
int i; |
int i; |
1159 |
int16_t Src[8*8], Dst[8*8]; |
int16_t Src[8*8], Dst[8*8]; |
1163 |
|
|
1164 |
for(i=0; i<64; ++i) Src[i] = i-32; |
for(i=0; i<64; ++i) Src[i] = i-32; |
1165 |
set_intra_matrix( get_default_intra_matrix() ); |
set_intra_matrix( get_default_intra_matrix() ); |
1166 |
dequant4_intra(Dst, Src, 31, 5); |
dequant_mpeg_intra(Dst, Src, 31, 5); |
1167 |
printf( "dequant4_intra with CPU=%s: ", cpu->name); |
printf( "dequant_mpeg_intra with CPU=%s: ", cpu->name); |
1168 |
printf( " Out[]= " ); |
printf( " Out[]= " ); |
1169 |
for(i=0; i<64; ++i) printf( "[%d]", Dst[i]); |
for(i=0; i<64; ++i) printf( "[%d]", Dst[i]); |
1170 |
printf( "\n" ); |
printf( "\n" ); |
1172 |
|
|
1173 |
printf( "\n ===== (de)quant4_inter saturation bug? =====\n" ); |
printf( "\n ===== (de)quant4_inter saturation bug? =====\n" ); |
1174 |
|
|
1175 |
for(cpu = cpu_short_list; cpu->name!=0; ++cpu) |
for(cpu = cpu_list; cpu->name!=0; ++cpu) |
1176 |
{ |
{ |
1177 |
int i; |
int i; |
1178 |
int16_t Src[8*8], Dst[8*8]; |
int16_t Src[8*8], Dst[8*8]; |
1182 |
|
|
1183 |
for(i=0; i<64; ++i) Src[i] = i-32; |
for(i=0; i<64; ++i) Src[i] = i-32; |
1184 |
set_inter_matrix( get_default_inter_matrix() ); |
set_inter_matrix( get_default_inter_matrix() ); |
1185 |
dequant4_inter(Dst, Src, 31); |
dequant_mpeg_inter(Dst, Src, 31); |
1186 |
printf( "dequant4_inter with CPU=%s: ", cpu->name); |
printf( "dequant_mpeg_inter with CPU=%s: ", cpu->name); |
1187 |
printf( " Out[]= " ); |
printf( " Out[]= " ); |
1188 |
for(i=0; i<64; ++i) printf( "[%d]", Dst[i]); |
for(i=0; i<64; ++i) printf( "[%d]", Dst[i]); |
1189 |
printf( "\n" ); |
printf( "\n" ); |
1197 |
|
|
1198 |
printf( "\n ===== fdct/idct precision diffs =====\n" ); |
printf( "\n ===== fdct/idct precision diffs =====\n" ); |
1199 |
|
|
1200 |
for(cpu = cpu_short_list; cpu->name!=0; ++cpu) |
for(cpu = cpu_list; cpu->name!=0; ++cpu) |
1201 |
{ |
{ |
1202 |
int i; |
int i; |
1203 |
|
|
1250 |
|
|
1251 |
for(q=1; q<=max_Q; ++q) { |
for(q=1; q<=max_Q; ++q) { |
1252 |
emms(); |
emms(); |
1253 |
quant4_inter( Dst, Src, q ); |
quant_mpeg_inter( Dst, Src, q ); |
1254 |
emms(); |
emms(); |
1255 |
for(s=0, i=0; i<64; ++i) s+=((uint16_t)Dst[i])^i; |
for(s=0, i=0; i<64; ++i) s+=((uint16_t)Dst[i])^i; |
1256 |
Crcs_Inter[n][q] = s; |
Crcs_Inter[n][q] = s; |
1280 |
|
|
1281 |
for(q=1; q<=max_Q; ++q) { |
for(q=1; q<=max_Q; ++q) { |
1282 |
emms(); |
emms(); |
1283 |
quant4_intra( Dst, Src, q, q); |
quant_mpeg_intra( Dst, Src, q, q); |
1284 |
emms(); |
emms(); |
1285 |
for(s=0, i=0; i<64; ++i) s+=((uint16_t)Dst[i])^i; |
for(s=0, i=0; i<64; ++i) s+=((uint16_t)Dst[i])^i; |
1286 |
Crcs_Intra[n][q] = s; |
Crcs_Intra[n][q] = s; |