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: xvid_bench.c,v 1.9.2.7 2003-11-03 19:58:16 edgomez Exp $ |
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 <string.h> // for memset |
#include <string.h> /* for memset */ |
40 |
#include <assert.h> |
#include <assert.h> |
41 |
|
|
42 |
#ifndef WIN32 |
#ifndef WIN32 |
43 |
#include <sys/time.h> // for gettimeofday |
#include <sys/time.h> /* for gettimeofday */ |
44 |
#else |
#else |
45 |
#include <time.h> |
#include <time.h> |
46 |
#endif |
#endif |
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" |
67 |
#define M_PI 3.14159265358979323846 |
#define M_PI 3.14159265358979323846 |
68 |
#endif |
#endif |
69 |
|
|
70 |
const int speed_ref = 100; // on slow machines, decrease this value |
const int speed_ref = 100; /* on slow machines, decrease this value */ |
71 |
|
|
72 |
/********************************************************************* |
/********************************************************************* |
73 |
* misc |
* misc |
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 |
, { "IA64 ", XVID_CPU_IA64 } |
#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 |
, { "IA64 ", XVID_CPU_IA64 } |
#endif |
134 |
, { 0, 0 } } |
//, { "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; |
xvid_gbl_info_t xinfo; |
|
XVID_INIT_PARAM xinit; |
|
149 |
|
|
150 |
cpu_type = check_cpu_features() & cpu->cpu; |
/* Get the available CPU flags */ |
151 |
xinit.cpu_flags = cpu_type | XVID_CPU_FORCE; |
memset(&xinfo, 0, sizeof(xinfo)); |
152 |
// xinit.cpu_flags = XVID_CPU_MMX | XVID_CPU_FORCE; |
xinfo.version = XVID_VERSION; |
153 |
xerr = xvid_init(NULL, 0, &xinit, NULL); |
xvid_global(NULL, XVID_GBL_INFO, &xinfo, NULL); |
154 |
if (cpu->cpu>0 && (cpu_type==0 || xerr!=XVID_ERR_OK)) { |
|
155 |
printf( "%s - skipped...\n", cpu->name ); |
/* Are we trying to test a subset of the host CPU features */ |
156 |
|
if ((xinfo.cpu_flags & cpu->cpu) == cpu->cpu) { |
157 |
|
int xerr; |
158 |
|
xvid_gbl_init_t xinit; |
159 |
|
memset(&xinit, 0, sizeof(xinit)); |
160 |
|
xinit.cpu_flags = cpu->cpu | XVID_CPU_FORCE; |
161 |
|
xinit.version = XVID_VERSION; |
162 |
|
xerr = xvid_global(NULL, XVID_GBL_INIT, &xinit, NULL); |
163 |
|
if (xerr==XVID_ERR_FAIL) { |
164 |
|
/* libxvidcore failed to init */ |
165 |
return 0; |
return 0; |
166 |
} |
} |
167 |
|
} else { |
168 |
|
/* The host CPU doesn't support some required feature for this test */ |
169 |
|
return(0); |
170 |
|
} |
171 |
return 1; |
return 1; |
172 |
} |
} |
173 |
|
|
183 |
int tst; |
int tst; |
184 |
CPU *cpu; |
CPU *cpu; |
185 |
int i; |
int i; |
186 |
short iDst0[8*8], iDst[8*8], fDst[8*8]; |
DECLARE_ALIGNED_MATRIX(iDst0, 8, 8, short, 16); |
187 |
|
DECLARE_ALIGNED_MATRIX(iDst, 8, 8, short, 16); |
188 |
|
DECLARE_ALIGNED_MATRIX(fDst, 8, 8, short, 16); |
189 |
double overhead; |
double overhead; |
190 |
|
|
191 |
printf( "\n ===== test fdct/idct =====\n" ); |
printf( "\n ===== test fdct/idct =====\n" ); |
322 |
const int nb_tests = 2000*speed_ref; |
const int nb_tests = 2000*speed_ref; |
323 |
CPU *cpu; |
CPU *cpu; |
324 |
const uint8_t Src0[16*9] = { |
const uint8_t Src0[16*9] = { |
325 |
// try to have every possible combinaison of rounding... |
/* try to have every possible combinaison of rounding... */ |
326 |
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 |
327 |
, 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 |
328 |
, 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 |
372 |
if (iCrc!=8103) printf( "*** CRC ERROR! ***\n" ); |
if (iCrc!=8103) printf( "*** CRC ERROR! ***\n" ); |
373 |
|
|
374 |
|
|
375 |
// this is a new function, as of 06.06.2002 |
/* this is a new function, as of 06.06.2002 */ |
376 |
#if 0 |
#if 0 |
377 |
TEST_MB2(interpolate8x8_avrg); |
TEST_MB2(interpolate8x8_avrg); |
378 |
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 ); |
460 |
|
|
461 |
printf( "\n === test transfer ===\n" ); |
printf( "\n === test transfer ===\n" ); |
462 |
|
|
463 |
for(cpu = cpu_short_list; cpu->name!=0; ++cpu) |
for(cpu = cpu_list; cpu->name!=0; ++cpu) |
464 |
{ |
{ |
465 |
double t, overhead; |
double t, overhead; |
466 |
int tst, s; |
int tst, s; |
548 |
|
|
549 |
printf( "\n ===== test quant =====\n" ); |
printf( "\n ===== test quant =====\n" ); |
550 |
|
|
551 |
// we deliberately enfringe the norm's specified range [-127,127], |
/* we deliberately enfringe the norm's specified range [-127,127], */ |
552 |
// to test the robustness of the iquant module |
/* to test the robustness of the iquant module */ |
553 |
for(i=0; i<64; ++i) { |
for(i=0; i<64; ++i) { |
554 |
Src[i] = 1 + (i-32) * (i&6); |
Src[i] = 1 + (i-32) * (i&6); |
555 |
Dst[i] = 0; |
Dst[i] = 0; |
556 |
} |
} |
557 |
|
|
558 |
for(cpu = cpu_short_list; cpu->name!=0; ++cpu) |
for(cpu = cpu_list; cpu->name!=0; ++cpu) |
559 |
{ |
{ |
560 |
double t, overhead; |
double t, overhead; |
561 |
int tst, q; |
int tst, q; |
574 |
overhead += gettime_usec(); |
overhead += gettime_usec(); |
575 |
|
|
576 |
#if 1 |
#if 1 |
577 |
TEST_QUANT2(quant4_intra, Dst, Src); |
TEST_QUANT2(quant_mpeg_intra, Dst, Src); |
578 |
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 ); |
579 |
if (s!=29809) printf( "*** CRC ERROR! ***\n" ); |
if (s!=29809) printf( "*** CRC ERROR! ***\n" ); |
580 |
|
|
581 |
TEST_QUANT(quant4_inter, Dst, Src); |
TEST_QUANT(quant_mpeg_inter, Dst, Src); |
582 |
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 ); |
583 |
if (s!=12574) printf( "*** CRC ERROR! ***\n" ); |
if (s!=12574) printf( "*** CRC ERROR! ***\n" ); |
584 |
#endif |
#endif |
585 |
#if 1 |
#if 1 |
586 |
TEST_QUANT2(dequant4_intra, Dst, Src); |
TEST_QUANT2(dequant_mpeg_intra, Dst, Src); |
587 |
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 ); |
588 |
if (s!=24052) printf( "*** CRC ERROR! ***\n" ); |
if (s!=24052) printf( "*** CRC ERROR! ***\n" ); |
589 |
|
|
590 |
TEST_QUANT(dequant4_inter, Dst, Src); |
TEST_QUANT(dequant_mpeg_inter, Dst, Src); |
591 |
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 ); |
592 |
if (s!=63847) printf( "*** CRC ERROR! ***\n" ); |
if (s!=63847) printf( "*** CRC ERROR! ***\n" ); |
593 |
#endif |
#endif |
594 |
#if 1 |
#if 1 |
595 |
TEST_QUANT2(quant_intra, Dst, Src); |
TEST_QUANT2(quant_h263_intra, Dst, Src); |
596 |
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 ); |
597 |
if (s!=25662) printf( "*** CRC ERROR! ***\n" ); |
if (s!=25662) printf( "*** CRC ERROR! ***\n" ); |
598 |
|
|
599 |
TEST_QUANT(quant_inter, Dst, Src); |
TEST_QUANT(quant_h263_inter, Dst, Src); |
600 |
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 ); |
601 |
if (s!=23972) printf( "*** CRC ERROR! ***\n" ); |
if (s!=23972) printf( "*** CRC ERROR! ***\n" ); |
602 |
#endif |
#endif |
603 |
#if 1 |
#if 1 |
604 |
TEST_QUANT2(dequant_intra, Dst, Src); |
TEST_QUANT2(dequant_h263_intra, Dst, Src); |
605 |
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 ); |
606 |
if (s!=49900) printf( "*** CRC ERROR! ***\n" ); |
if (s!=49900) printf( "*** CRC ERROR! ***\n" ); |
607 |
|
|
608 |
TEST_QUANT(dequant_inter, Dst, Src); |
TEST_QUANT(dequant_h263_inter, Dst, Src); |
609 |
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 ); |
610 |
if (s!=48899) printf( "*** CRC ERROR! ***\n" ); |
if (s!=48899) printf( "*** CRC ERROR! ***\n" ); |
611 |
#endif |
#endif |
612 |
printf( " --- \n" ); |
printf( " --- \n" ); |
636 |
printf( "\n ===== test cbp =====\n" ); |
printf( "\n ===== test cbp =====\n" ); |
637 |
|
|
638 |
for(i=0; i<6*64; ++i) { |
for(i=0; i<6*64; ++i) { |
639 |
Src1[i] = (i*i*3/8192)&(i/64)&1; // 'random' |
Src1[i] = (i*i*3/8192)&(i/64)&1; /* 'random' */ |
640 |
Src2[i] = (i<3*64); // half-full |
Src2[i] = (i<3*64); /* half-full */ |
641 |
Src3[i] = ((i+32)>3*64); |
Src3[i] = ((i+32)>3*64); |
642 |
Src4[i] = (i==(3*64+2) || i==(5*64+9)); |
Src4[i] = (i==(3*64+2) || i==(5*64+9)); |
643 |
} |
} |
644 |
|
|
645 |
for(cpu = cpu_short_list2; cpu->name!=0; ++cpu) |
for(cpu = cpu_list; cpu->name!=0; ++cpu) |
646 |
{ |
{ |
647 |
double t; |
double t; |
648 |
int tst, cbp; |
int tst, cbp; |
792 |
} |
} |
793 |
} |
} |
794 |
|
|
795 |
////////////////////////////////////////////////////////// |
///* ////////////////////////////////////////////////////// */ |
796 |
/* Pseudo-random generator specified by IEEE 1180 */ |
/* Pseudo-random generator specified by IEEE 1180 */ |
797 |
|
|
798 |
static long ieee_seed = 1; |
static long ieee_seed = 1; |
879 |
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 }; |
880 |
int Loops = 10000; |
int Loops = 10000; |
881 |
int i, m, n; |
int i, m, n; |
882 |
short Blk0[64]; // reference |
DECLARE_ALIGNED_MATRIX(Blk0, 8, 8, short, 16); /* reference */ |
883 |
short Blk[64], iBlk[64]; |
DECLARE_ALIGNED_MATRIX(Blk, 8, 8, short, 16); |
884 |
short Ref_FDCT[64]; |
DECLARE_ALIGNED_MATRIX(iBlk, 8, 8, short, 16); |
885 |
short Ref_IDCT[64]; |
DECLARE_ALIGNED_MATRIX(Ref_FDCT, 8, 8, short, 16); |
886 |
|
DECLARE_ALIGNED_MATRIX(Ref_IDCT, 8, 8, short, 16); |
887 |
|
|
888 |
STATS_8x8 FStats; // forward dct stats |
STATS_8x8 FStats; /* forward dct stats */ |
889 |
STATS_8x8 IStats; // inverse dct stats |
STATS_8x8 IStats; /* inverse dct stats */ |
890 |
|
|
891 |
CPU *cpu; |
CPU *cpu; |
892 |
|
|
909 |
for(i=0; i<64; ++i) |
for(i=0; i<64; ++i) |
910 |
Blk0[i] = (short)ieee_rand(Min,Max) * Sign; |
Blk0[i] = (short)ieee_rand(Min,Max) * Sign; |
911 |
|
|
912 |
// hmm, I'm not quite sure this is exactly |
/* hmm, I'm not quite sure this is exactly */ |
913 |
// the tests described in the norm. check... |
/* the tests described in the norm. check... */ |
914 |
|
|
915 |
memcpy(Ref_FDCT, Blk0, 64*sizeof(short)); |
memcpy(Ref_FDCT, Blk0, 64*sizeof(short)); |
916 |
ref_fdct(Ref_FDCT); |
ref_fdct(Ref_FDCT); |
937 |
|
|
938 |
printf( "\n -- FDCT report --\n" ); |
printf( "\n -- FDCT report --\n" ); |
939 |
// print_stats(&FStats); |
// print_stats(&FStats); |
940 |
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 */ |
941 |
|
|
942 |
for(i=0; i<64; i++) Blk[i] = 0; |
for(i=0; i<64; i++) Blk[i] = 0; |
943 |
emms(); fdct(Blk); emms(); |
emms(); fdct(Blk); emms(); |
959 |
|
|
960 |
void test_dct_saturation(int Min, int Max) |
void test_dct_saturation(int Min, int Max) |
961 |
{ |
{ |
962 |
// test behaviour on input range fringe |
/* test behaviour on input range fringe */ |
963 |
|
|
964 |
int i, n, p; |
int i, n, p; |
965 |
CPU *cpu; |
CPU *cpu; |
966 |
// const short IDCT_MAX = 2047; // 12bits input |
// const short IDCT_MAX = 2047; /* 12bits input */ |
967 |
// const short IDCT_MIN = -2048; |
// const short IDCT_MIN = -2048; |
968 |
// const short IDCT_OUT = 256; // 9bits ouput |
// const short IDCT_OUT = 256; /* 9bits ouput */ |
969 |
const int Partitions = 4; |
const int Partitions = 4; |
970 |
const int Loops = 10000 / Partitions; |
const int Loops = 10000 / Partitions; |
971 |
|
|
982 |
printf( "\n===== IEEE test for %s Min=%d Max=%d =====\n", |
printf( "\n===== IEEE test for %s Min=%d Max=%d =====\n", |
983 |
cpu->name, Min, Max ); |
cpu->name, Min, Max ); |
984 |
|
|
985 |
// FDCT tests // |
/* FDCT tests // */ |
986 |
|
|
987 |
init_stats(&Stats); |
init_stats(&Stats); |
988 |
|
|
989 |
// test each computation channels separately |
/* test each computation channels separately */ |
990 |
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; |
991 |
ref_fdct(Blk0); |
ref_fdct(Blk0); |
992 |
emms(); fdct(Blk); emms(); |
emms(); fdct(Blk); emms(); |
997 |
emms(); fdct(Blk); emms(); |
emms(); fdct(Blk); emms(); |
998 |
store_stats(&Stats, Blk, Blk0); |
store_stats(&Stats, Blk, Blk0); |
999 |
|
|
1000 |
// randomly saturated inputs |
/* randomly saturated inputs */ |
1001 |
for(p=0; p<Partitions; ++p) |
for(p=0; p<Partitions; ++p) |
1002 |
{ |
{ |
1003 |
for(n=0; n<Loops; ++n) |
for(n=0; n<Loops; ++n) |
1013 |
report_stats(&Stats, 0); |
report_stats(&Stats, 0); |
1014 |
|
|
1015 |
|
|
1016 |
// IDCT tests // |
/* IDCT tests // */ |
1017 |
#if 0 |
#if 0 |
1018 |
// no finished yet |
/* no finished yet */ |
1019 |
|
|
1020 |
init_stats(&Stats); |
init_stats(&Stats); |
1021 |
|
|
1022 |
// test each computation channel separately |
/* test each computation channel separately */ |
1023 |
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; |
1024 |
ref_idct(Blk0); |
ref_idct(Blk0); |
1025 |
emms(); idct(Blk); emms(); |
emms(); idct(Blk); emms(); |
1032 |
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); } |
1033 |
store_stats(&Stats, Blk, Blk0); |
store_stats(&Stats, Blk, Blk0); |
1034 |
|
|
1035 |
// randomly saturated inputs |
/* randomly saturated inputs */ |
1036 |
for(p=0; p<Partitions; ++p) |
for(p=0; p<Partitions; ++p) |
1037 |
{ |
{ |
1038 |
for(n=0; n<Loops; ++n) |
for(n=0; n<Loops; ++n) |
1062 |
FILE *f = 0; |
FILE *f = 0; |
1063 |
void *dechandle = 0; |
void *dechandle = 0; |
1064 |
int xerr; |
int xerr; |
1065 |
XVID_INIT_PARAM xinit; |
xvid_gbl_init_t xinit; |
1066 |
XVID_DEC_PARAM xparam; |
xvid_dec_create_t xparam; |
1067 |
XVID_DEC_FRAME xframe; |
xvid_dec_frame_t xframe; |
1068 |
double t = 0.; |
double t = 0.; |
1069 |
int nb = 0; |
int nb = 0; |
1070 |
uint8_t *buf = 0; |
uint8_t *buf = 0; |
1072 |
int buf_size, pos; |
int buf_size, pos; |
1073 |
uint32_t chksum = 0; |
uint32_t chksum = 0; |
1074 |
|
|
1075 |
|
memset(&xinit, 0, sizeof(xinit)); |
1076 |
xinit.cpu_flags = XVID_CPU_MMX | XVID_CPU_FORCE; |
xinit.cpu_flags = XVID_CPU_MMX | XVID_CPU_FORCE; |
1077 |
xvid_init(NULL, 0, &xinit, NULL); |
xinit.version = XVID_VERSION; |
1078 |
printf( "API version: %d, core build:%d\n", xinit.api_version, xinit.core_build); |
xvid_global(NULL, 0, &xinit, NULL); |
|
|
|
1079 |
|
|
1080 |
|
memset(&xparam, 0, sizeof(xparam)); |
1081 |
xparam.width = width; |
xparam.width = width; |
1082 |
xparam.height = height; |
xparam.height = height; |
1083 |
|
xparam.version = XVID_VERSION; |
1084 |
xerr = xvid_decore(NULL, XVID_DEC_CREATE, &xparam, NULL); |
xerr = xvid_decore(NULL, XVID_DEC_CREATE, &xparam, NULL); |
1085 |
if (xerr!=XVID_ERR_OK) { |
if (xerr==XVID_ERR_FAIL) { |
1086 |
printf("can't init decoder (err=%d)\n", xerr); |
printf("can't init decoder (err=%d)\n", xerr); |
1087 |
return; |
return; |
1088 |
} |
} |
1103 |
} |
} |
1104 |
else printf( "Input size: %d\n", buf_size); |
else printf( "Input size: %d\n", buf_size); |
1105 |
|
|
1106 |
buf = malloc(buf_size); // should be enuf' |
buf = malloc(buf_size); /* should be enuf' */ |
1107 |
rgb_out = calloc(4, width*height); // <-room for _RGB24 |
rgb_out = calloc(4, width*height); /* <-room for _RGB24 */ |
1108 |
if (buf==0 || rgb_out==0) { |
if (buf==0 || rgb_out==0) { |
1109 |
printf( "malloc failed!\n" ); |
printf( "malloc failed!\n" ); |
1110 |
goto End; |
goto End; |
1119 |
pos = 0; |
pos = 0; |
1120 |
t = -gettime_usec(); |
t = -gettime_usec(); |
1121 |
while(1) { |
while(1) { |
1122 |
|
memset(&xframe, 0, sizeof(xframe)); |
1123 |
|
xframe.version = XVID_VERSION; |
1124 |
xframe.bitstream = buf + pos; |
xframe.bitstream = buf + pos; |
1125 |
xframe.length = buf_size - pos; |
xframe.length = buf_size - pos; |
1126 |
xframe.image = rgb_out; |
xframe.output.plane[0] = rgb_out; |
1127 |
xframe.stride = width; |
xframe.output.stride[0] = width; |
1128 |
xframe.colorspace = XVID_CSP_RGB24; |
xframe.output.csp = XVID_CSP_BGR; |
1129 |
xerr = xvid_decore(dechandle, XVID_DEC_DECODE, &xframe, 0); |
xerr = xvid_decore(dechandle, XVID_DEC_DECODE, &xframe, 0); |
1130 |
nb++; |
nb++; |
1131 |
pos += xframe.length; |
pos += xframe.length; |
1136 |
} |
} |
1137 |
if (pos==buf_size) |
if (pos==buf_size) |
1138 |
break; |
break; |
1139 |
if (xerr!=XVID_ERR_OK) { |
if (xerr==XVID_ERR_FAIL) { |
1140 |
printf("decoding failed for frame #%d (err=%d)!\n", nb, xerr); |
printf("decoding failed for frame #%d (err=%d)!\n", nb, xerr); |
1141 |
break; |
break; |
1142 |
} |
} |
1152 |
if (buf!=0) free(buf); |
if (buf!=0) free(buf); |
1153 |
if (dechandle!=0) { |
if (dechandle!=0) { |
1154 |
xerr= xvid_decore(dechandle, XVID_DEC_DESTROY, NULL, NULL); |
xerr= xvid_decore(dechandle, XVID_DEC_DESTROY, NULL, NULL); |
1155 |
if (xerr!=XVID_ERR_OK) |
if (xerr==XVID_ERR_FAIL) |
1156 |
printf("destroy-decoder failed (err=%d)!\n", xerr); |
printf("destroy-decoder failed (err=%d)!\n", xerr); |
1157 |
} |
} |
1158 |
if (f!=0) fclose(f); |
if (f!=0) fclose(f); |
1168 |
|
|
1169 |
printf( "\n ===== (de)quant4_intra saturation bug? =====\n" ); |
printf( "\n ===== (de)quant4_intra saturation bug? =====\n" ); |
1170 |
|
|
1171 |
for(cpu = cpu_short_list; cpu->name!=0; ++cpu) |
for(cpu = cpu_list; cpu->name!=0; ++cpu) |
1172 |
{ |
{ |
1173 |
int i; |
int i; |
1174 |
int16_t Src[8*8], Dst[8*8]; |
int16_t Src[8*8], Dst[8*8]; |
1178 |
|
|
1179 |
for(i=0; i<64; ++i) Src[i] = i-32; |
for(i=0; i<64; ++i) Src[i] = i-32; |
1180 |
set_intra_matrix( get_default_intra_matrix() ); |
set_intra_matrix( get_default_intra_matrix() ); |
1181 |
dequant4_intra(Dst, Src, 31, 5); |
dequant_mpeg_intra(Dst, Src, 31, 5); |
1182 |
printf( "dequant4_intra with CPU=%s: ", cpu->name); |
printf( "dequant_mpeg_intra with CPU=%s: ", cpu->name); |
1183 |
printf( " Out[]= " ); |
printf( " Out[]= " ); |
1184 |
for(i=0; i<64; ++i) printf( "[%d]", Dst[i]); |
for(i=0; i<64; ++i) printf( "[%d]", Dst[i]); |
1185 |
printf( "\n" ); |
printf( "\n" ); |
1187 |
|
|
1188 |
printf( "\n ===== (de)quant4_inter saturation bug? =====\n" ); |
printf( "\n ===== (de)quant4_inter saturation bug? =====\n" ); |
1189 |
|
|
1190 |
for(cpu = cpu_short_list; cpu->name!=0; ++cpu) |
for(cpu = cpu_list; cpu->name!=0; ++cpu) |
1191 |
{ |
{ |
1192 |
int i; |
int i; |
1193 |
int16_t Src[8*8], Dst[8*8]; |
int16_t Src[8*8], Dst[8*8]; |
1197 |
|
|
1198 |
for(i=0; i<64; ++i) Src[i] = i-32; |
for(i=0; i<64; ++i) Src[i] = i-32; |
1199 |
set_inter_matrix( get_default_inter_matrix() ); |
set_inter_matrix( get_default_inter_matrix() ); |
1200 |
dequant4_inter(Dst, Src, 31); |
dequant_mpeg_inter(Dst, Src, 31); |
1201 |
printf( "dequant4_inter with CPU=%s: ", cpu->name); |
printf( "dequant_mpeg_inter with CPU=%s: ", cpu->name); |
1202 |
printf( " Out[]= " ); |
printf( " Out[]= " ); |
1203 |
for(i=0; i<64; ++i) printf( "[%d]", Dst[i]); |
for(i=0; i<64; ++i) printf( "[%d]", Dst[i]); |
1204 |
printf( "\n" ); |
printf( "\n" ); |
1212 |
|
|
1213 |
printf( "\n ===== fdct/idct precision diffs =====\n" ); |
printf( "\n ===== fdct/idct precision diffs =====\n" ); |
1214 |
|
|
1215 |
for(cpu = cpu_short_list; cpu->name!=0; ++cpu) |
for(cpu = cpu_list; cpu->name!=0; ++cpu) |
1216 |
{ |
{ |
1217 |
int i; |
int i; |
1218 |
|
|
1265 |
|
|
1266 |
for(q=1; q<=max_Q; ++q) { |
for(q=1; q<=max_Q; ++q) { |
1267 |
emms(); |
emms(); |
1268 |
quant4_inter( Dst, Src, q ); |
quant_mpeg_inter( Dst, Src, q ); |
1269 |
emms(); |
emms(); |
1270 |
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; |
1271 |
Crcs_Inter[n][q] = s; |
Crcs_Inter[n][q] = s; |
1295 |
|
|
1296 |
for(q=1; q<=max_Q; ++q) { |
for(q=1; q<=max_Q; ++q) { |
1297 |
emms(); |
emms(); |
1298 |
quant4_intra( Dst, Src, q, q); |
quant_mpeg_intra( Dst, Src, q, q); |
1299 |
emms(); |
emms(); |
1300 |
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; |
1301 |
Crcs_Intra[n][q] = s; |
Crcs_Intra[n][q] = s; |
1328 |
|
|
1329 |
if (what==7) { |
if (what==7) { |
1330 |
test_IEEE1180_compliance(-256, 255, 1); |
test_IEEE1180_compliance(-256, 255, 1); |
|
#if 0 |
|
1331 |
test_IEEE1180_compliance(-256, 255,-1); |
test_IEEE1180_compliance(-256, 255,-1); |
1332 |
test_IEEE1180_compliance( -5, 5, 1); |
test_IEEE1180_compliance( -5, 5, 1); |
1333 |
test_IEEE1180_compliance( -5, 5,-1); |
test_IEEE1180_compliance( -5, 5,-1); |
1334 |
test_IEEE1180_compliance(-300, 300, 1); |
test_IEEE1180_compliance(-300, 300, 1); |
1335 |
test_IEEE1180_compliance(-300, 300,-1); |
test_IEEE1180_compliance(-300, 300,-1); |
|
#endif |
|
1336 |
} |
} |
1337 |
if (what==8) test_dct_saturation(-256, 255); |
if (what==8) test_dct_saturation(-256, 255); |
1338 |
|
|