35 |
|
|
36 |
#include <stdio.h> |
#include <stdio.h> |
37 |
#include <stdlib.h> |
#include <stdlib.h> |
38 |
#include <sys/time.h> // for gettimeofday |
#ifdef WIN32 |
39 |
#include <string.h> // for memset |
#include <time.h> /* for clock */ |
40 |
|
#else |
41 |
|
#include <sys/time.h> /* for gettimeofday */ |
42 |
|
#endif |
43 |
|
#include <string.h> /* for memset */ |
44 |
#include <assert.h> |
#include <assert.h> |
45 |
|
|
46 |
#include "xvid.h" |
#include "xvid.h" |
47 |
|
|
48 |
// inner guts |
/* inner guts */ |
49 |
#include "dct/idct.h" |
#include "dct/idct.h" |
50 |
#include "dct/fdct.h" |
#include "dct/fdct.h" |
51 |
#include "image/colorspace.h" |
#include "image/colorspace.h" |
60 |
#include "bitstream/cbp.h" |
#include "bitstream/cbp.h" |
61 |
|
|
62 |
#include <math.h> |
#include <math.h> |
63 |
const int speed_ref = 100; // on slow machines, decrease this value |
#ifndef M_PI |
64 |
|
# define M_PI 3.14159265359 |
65 |
|
# define M_PI_2 1.5707963268 |
66 |
|
#endif |
67 |
|
const int speed_ref = 100; /* on slow machines, decrease this value */ |
68 |
|
|
69 |
/********************************************************************* |
/********************************************************************* |
70 |
* misc |
* misc |
73 |
/* returns time in micro-s*/ |
/* returns time in micro-s*/ |
74 |
double gettime_usec() |
double gettime_usec() |
75 |
{ |
{ |
76 |
|
#ifdef WIN32 |
77 |
|
return clock()*1000; |
78 |
|
#else |
79 |
struct timeval tv; |
struct timeval tv; |
80 |
gettimeofday(&tv, 0); |
gettimeofday(&tv, 0); |
81 |
return tv.tv_sec*1.0e6f + tv.tv_usec; |
return tv.tv_sec*1.0e6 + tv.tv_usec; |
82 |
|
#endif |
83 |
} |
} |
84 |
|
|
85 |
/* 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 |
, { "SSE2 ", XVID_CPU_SSE2 | XVID_CPU_MMX } |
, { "SSE2 ", XVID_CPU_SSE2 | XVID_CPU_MMX } |
115 |
, { "3DNOW ", XVID_CPU_3DNOW } |
, { "3DNOW ", XVID_CPU_3DNOW } |
116 |
, { "3DNOWE", XVID_CPU_3DNOWEXT } |
, { "3DNOWE", XVID_CPU_3DNOWEXT } |
117 |
//, { "TSC ", XVID_CPU_TSC } |
, { "IA64 ", XVID_CPU_IA64 } |
118 |
|
/*, { "TSC ", XVID_CPU_TSC } */ |
119 |
, { 0, 0 } } |
, { 0, 0 } } |
120 |
|
|
121 |
, cpu_short_list[] = |
, cpu_short_list[] = |
122 |
{ { "PLAINC", 0 } |
{ { "PLAINC", 0 } |
123 |
, { "MMX ", XVID_CPU_MMX } |
, { "MMX ", XVID_CPU_MMX } |
124 |
//, { "MMXEXT", XVID_CPU_MMXEXT | XVID_CPU_MMX } |
/*, { "MMXEXT", XVID_CPU_MMXEXT | XVID_CPU_MMX } */ |
125 |
|
, { "IA64 ", XVID_CPU_IA64 } |
126 |
, { 0, 0 } } |
, { 0, 0 } } |
127 |
|
|
128 |
, cpu_short_list2[] = |
, cpu_short_list2[] = |
139 |
|
|
140 |
cpu_type = check_cpu_features() & cpu->cpu; |
cpu_type = check_cpu_features() & cpu->cpu; |
141 |
xinit.cpu_flags = cpu_type | XVID_CPU_FORCE; |
xinit.cpu_flags = cpu_type | XVID_CPU_FORCE; |
142 |
// xinit.cpu_flags = XVID_CPU_MMX | XVID_CPU_FORCE; |
/* xinit.cpu_flags = XVID_CPU_MMX | XVID_CPU_FORCE; */ |
143 |
xerr = xvid_init(NULL, 0, &xinit, NULL); |
xerr = xvid_init(NULL, 0, &xinit, NULL); |
144 |
if (cpu->cpu>0 && (cpu_type==0 || xerr!=XVID_ERR_OK)) { |
if (cpu->cpu>0 && (cpu_type==0 || xerr!=XVID_ERR_OK)) { |
145 |
printf( "%s - skipped...\n", cpu->name ); |
printf( "%s - skipped...\n", cpu->name ); |
297 |
const int nb_tests = 2000*speed_ref; |
const int nb_tests = 2000*speed_ref; |
298 |
CPU *cpu; |
CPU *cpu; |
299 |
const uint8_t Src0[16*9] = { |
const uint8_t Src0[16*9] = { |
300 |
// try to have every possible combinaison of rounding... |
/* try to have every possible combinaison of rounding... */ |
301 |
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 |
302 |
, 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 |
303 |
, 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 |
347 |
if (iCrc!=8103) printf( "*** CRC ERROR! ***\n" ); |
if (iCrc!=8103) printf( "*** CRC ERROR! ***\n" ); |
348 |
|
|
349 |
|
|
350 |
// this is a new function, as of 06.06.2002 |
/* this is a new function, as of 06.06.2002 */ |
351 |
#if 0 |
#if 0 |
352 |
TEST_MB2(interpolate8x8_avrg); |
TEST_MB2(interpolate8x8_avrg); |
353 |
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 ); |
469 |
TEST_TRANSFER3(transfer_8to16sub2, Dst16, Src8, Ref1, Ref2); |
TEST_TRANSFER3(transfer_8to16sub2, Dst16, Src8, Ref1, Ref2); |
470 |
printf( "%s - 8to16sub2 %.3f usec crc=%d\n", cpu->name, t, s ); |
printf( "%s - 8to16sub2 %.3f usec crc=%d\n", cpu->name, t, s ); |
471 |
if (s!=20384) printf( "*** CRC ERROR! ***\n" ); |
if (s!=20384) printf( "*** CRC ERROR! ***\n" ); |
472 |
// for(i=0; i<64; ++i) printf( "[%d]", Dst16[i]); |
/* for(i=0; i<64; ++i) printf( "[%d]", Dst16[i]); */ |
473 |
// printf("\n"); |
/* printf("\n"); */ |
474 |
#endif |
#endif |
475 |
printf( " --- \n" ); |
printf( " --- \n" ); |
476 |
} |
} |
523 |
|
|
524 |
printf( "\n ===== test quant =====\n" ); |
printf( "\n ===== test quant =====\n" ); |
525 |
|
|
526 |
// we deliberately enfringe the norm's specified range [-127,127], |
/* we deliberately enfringe the norm's specified range [-127,127], */ |
527 |
// to test the robustness of the iquant module |
/* to test the robustness of the iquant module */ |
528 |
for(i=0; i<64; ++i) { |
for(i=0; i<64; ++i) { |
529 |
Src[i] = 1 + (i-32) * (i&6); |
Src[i] = 1 + (i-32) * (i&6); |
530 |
Dst[i] = 0; |
Dst[i] = 0; |
611 |
printf( "\n ===== test cbp =====\n" ); |
printf( "\n ===== test cbp =====\n" ); |
612 |
|
|
613 |
for(i=0; i<6*64; ++i) { |
for(i=0; i<6*64; ++i) { |
614 |
Src1[i] = (i*i*3/8192)&(i/64)&1; // 'random' |
Src1[i] = (i*i*3/8192)&(i/64)&1; /* 'random' */ |
615 |
Src2[i] = (i<3*64); // half-full |
Src2[i] = (i<3*64); /* half-full */ |
616 |
Src3[i] = ((i+32)>3*64); |
Src3[i] = ((i+32)>3*64); |
617 |
Src4[i] = (i==(3*64+2) || i==(5*64+9)); |
Src4[i] = (i==(3*64+2) || i==(5*64+9)); |
618 |
} |
} |
767 |
} |
} |
768 |
} |
} |
769 |
|
|
770 |
////////////////////////////////////////////////////////// |
/*//////////////////////////////////////////////////////// */ |
771 |
/* Pseudo-random generator specified by IEEE 1180 */ |
/* Pseudo-random generator specified by IEEE 1180 */ |
772 |
|
|
773 |
static long ieee_seed = 1; |
static long ieee_seed = 1; |
854 |
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 }; |
855 |
int Loops = 10000; |
int Loops = 10000; |
856 |
int i, m, n; |
int i, m, n; |
857 |
short Blk0[64]; // reference |
short Blk0[64]; /* reference */ |
858 |
short Blk[64], iBlk[64]; |
short Blk[64], iBlk[64]; |
859 |
short Ref_FDCT[64]; |
short Ref_FDCT[64]; |
860 |
short Ref_IDCT[64]; |
short Ref_IDCT[64]; |
861 |
|
|
862 |
STATS_8x8 FStats; // forward dct stats |
STATS_8x8 FStats; /* forward dct stats */ |
863 |
STATS_8x8 IStats; // inverse dct stats |
STATS_8x8 IStats; /* inverse dct stats */ |
864 |
|
|
865 |
CPU *cpu; |
CPU *cpu; |
866 |
|
|
883 |
for(i=0; i<64; ++i) |
for(i=0; i<64; ++i) |
884 |
Blk0[i] = (short)ieee_rand(Min,Max) * Sign; |
Blk0[i] = (short)ieee_rand(Min,Max) * Sign; |
885 |
|
|
886 |
// hmm, I'm not quite sure this is exactly |
/* hmm, I'm not quite sure this is exactly */ |
887 |
// the tests described in the norm. check... |
/* the tests described in the norm. check... */ |
888 |
|
|
889 |
memcpy(Ref_FDCT, Blk0, 64*sizeof(short)); |
memcpy(Ref_FDCT, Blk0, 64*sizeof(short)); |
890 |
ref_fdct(Ref_FDCT); |
ref_fdct(Ref_FDCT); |
910 |
|
|
911 |
|
|
912 |
printf( "\n -- FDCT report --\n" ); |
printf( "\n -- FDCT report --\n" ); |
913 |
// print_stats(&FStats); |
/* print_stats(&FStats); */ |
914 |
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 */ |
915 |
|
|
916 |
for(i=0; i<64; i++) Blk[i] = 0; |
for(i=0; i<64; i++) Blk[i] = 0; |
917 |
emms(); fdct(Blk); emms(); |
emms(); fdct(Blk); emms(); |
919 |
printf( "FDCT(0) == 0 ? %s\n", (m!=0) ? "NOPE!" : "yup." ); |
printf( "FDCT(0) == 0 ? %s\n", (m!=0) ? "NOPE!" : "yup." ); |
920 |
|
|
921 |
printf( "\n -- IDCT report --\n" ); |
printf( "\n -- IDCT report --\n" ); |
922 |
// print_stats(&IStats); |
/* print_stats(&IStats); */ |
923 |
report_stats(&IStats, ILimits); |
report_stats(&IStats, ILimits); |
924 |
|
|
925 |
|
|
933 |
|
|
934 |
void test_dct_saturation(int Min, int Max) |
void test_dct_saturation(int Min, int Max) |
935 |
{ |
{ |
936 |
// test behaviour on input range fringe |
/* test behaviour on input range fringe */ |
937 |
|
|
938 |
int i, n, p; |
int i, n, p; |
939 |
CPU *cpu; |
CPU *cpu; |
940 |
// const short IDCT_MAX = 2047; // 12bits input |
/* const short IDCT_MAX = 2047; // 12bits input */ |
941 |
// const short IDCT_MIN = -2048; |
/* const short IDCT_MIN = -2048; */ |
942 |
// const short IDCT_OUT = 256; // 9bits ouput |
/* const short IDCT_OUT = 256; // 9bits ouput */ |
943 |
const int Partitions = 4; |
const int Partitions = 4; |
944 |
const int Loops = 10000 / Partitions; |
const int Loops = 10000 / Partitions; |
945 |
|
|
956 |
printf( "\n===== IEEE test for %s Min=%d Max=%d =====\n", |
printf( "\n===== IEEE test for %s Min=%d Max=%d =====\n", |
957 |
cpu->name, Min, Max ); |
cpu->name, Min, Max ); |
958 |
|
|
959 |
// FDCT tests // |
/* FDCT tests // */ |
960 |
|
|
961 |
init_stats(&Stats); |
init_stats(&Stats); |
962 |
|
|
963 |
// test each computation channels separately |
/* test each computation channels separately */ |
964 |
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; |
965 |
ref_fdct(Blk0); |
ref_fdct(Blk0); |
966 |
emms(); fdct(Blk); emms(); |
emms(); fdct(Blk); emms(); |
971 |
emms(); fdct(Blk); emms(); |
emms(); fdct(Blk); emms(); |
972 |
store_stats(&Stats, Blk, Blk0); |
store_stats(&Stats, Blk, Blk0); |
973 |
|
|
974 |
// randomly saturated inputs |
/* randomly saturated inputs */ |
975 |
for(p=0; p<Partitions; ++p) |
for(p=0; p<Partitions; ++p) |
976 |
{ |
{ |
977 |
for(n=0; n<Loops; ++n) |
for(n=0; n<Loops; ++n) |
987 |
report_stats(&Stats, 0); |
report_stats(&Stats, 0); |
988 |
|
|
989 |
|
|
990 |
// IDCT tests // |
/* IDCT tests */ |
991 |
#if 0 |
#if 0 |
992 |
// no finished yet |
/* no finished yet */ |
993 |
|
|
994 |
init_stats(&Stats); |
init_stats(&Stats); |
995 |
|
|
996 |
// test each computation channel separately |
/* test each computation channel separately */ |
997 |
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; |
998 |
ref_idct(Blk0); |
ref_idct(Blk0); |
999 |
emms(); idct(Blk); emms(); |
emms(); idct(Blk); emms(); |
1006 |
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); } |
1007 |
store_stats(&Stats, Blk, Blk0); |
store_stats(&Stats, Blk, Blk0); |
1008 |
|
|
1009 |
// randomly saturated inputs |
/* randomly saturated inputs */ |
1010 |
for(p=0; p<Partitions; ++p) |
for(p=0; p<Partitions; ++p) |
1011 |
{ |
{ |
1012 |
for(n=0; n<Loops; ++n) |
for(n=0; n<Loops; ++n) |
1075 |
} |
} |
1076 |
else printf( "Input size: %d\n", buf_size); |
else printf( "Input size: %d\n", buf_size); |
1077 |
|
|
1078 |
buf = malloc(buf_size); // should be enuf' |
buf = malloc(buf_size); /* should be enuf' */ |
1079 |
rgb_out = calloc(4, width*height); // <-room for _RGB24 |
rgb_out = calloc(4, width*height); /* <-room for _RGB24 */ |
1080 |
if (buf==0 || rgb_out==0) { |
if (buf==0 || rgb_out==0) { |
1081 |
printf( "malloc failed!\n" ); |
printf( "malloc failed!\n" ); |
1082 |
goto End; |
goto End; |