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 1.30 2006-10-13 15:16:25 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 |
|
|
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 |
|
{ "SSE3 ", XVID_CPU_SSE3 | XVID_CPU_SSE2 | XVID_CPU_MMX }, |
123 |
|
{ "SSE41 ", XVID_CPU_SSE41| XVID_CPU_SSE3 | XVID_CPU_SSE2 | XVID_CPU_MMX }, |
124 |
{ "3DNOW ", XVID_CPU_3DNOW }, |
{ "3DNOW ", XVID_CPU_3DNOW }, |
125 |
{ "3DNOWE ", XVID_CPU_3DNOW | XVID_CPU_3DNOWEXT }, |
{ "3DNOWE ", XVID_CPU_3DNOW | XVID_CPU_3DNOWEXT }, |
126 |
#endif |
#endif |
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); \ |
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 |
LEAVE |
1730 |
|
|
1731 |
#define WIDTH 64 |
#define WIDTH 128 |
1732 |
#define HEIGHT 64 |
#define HEIGHT 32 |
1733 |
void test_yuv() |
void test_yuv() |
1734 |
{ |
{ |
1735 |
const int nb_tests = 200*speed_ref; |
const int nb_tests = 200*speed_ref; |
1743 |
colorspace_init(); |
colorspace_init(); |
1744 |
ieee_reseed(1); |
ieee_reseed(1); |
1745 |
for(i=0; i<(int)sizeof(Src0); ++i) Src0[0][i] = ieee_rand(0,255); |
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" ); |
printf( "\n === test YUV ===\n" ); |
1749 |
|
|
1778 |
#endif |
#endif |
1779 |
printf( " --- \n" ); |
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 |
#undef WIDTH |
1846 |
#undef HEIGHT |
#undef HEIGHT |
1847 |
#undef ENTER |
#undef ENTER |
2012 |
printf( "ERROR! please post your platform/compiler specs to xvid-devel@xvid.org !\n" ); |
printf( "ERROR! please post your platform/compiler specs to xvid-devel@xvid.org !\n" ); |
2013 |
} |
} |
2014 |
} |
} |
2015 |
|
|
2016 |
/********************************************************************* |
/********************************************************************* |
2017 |
* test SSIM functions |
* test SSIM functions |
2018 |
*********************************************************************/ |
*********************************************************************/ |
2023 |
extern int lum_8x8_c(uint8_t* ptr, int stride); |
extern int lum_8x8_c(uint8_t* ptr, int stride); |
2024 |
extern int lum_8x8_mmx(uint8_t* ptr, int stride); |
extern int lum_8x8_mmx(uint8_t* ptr, int stride); |
2025 |
extern int lum_2x8_c(uint8_t* ptr, int stride); |
extern int lum_2x8_c(uint8_t* ptr, int stride); |
2026 |
extern void iconsim_c(uint8_t* ptro, uint8_t* ptrc, int stride, int lumo, int lumc, int* pdevo, int* pdevc, int* pcorr); |
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); |
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); |
extern void consim_sse2(uint8_t* ptro, uint8_t* ptrc, int stride, int lumo, int lumc, int* pdevo, int* pdevc, int* pcorr); |
2029 |
|
|
2061 |
continue; |
continue; |
2062 |
lum8x8 = lum_8x8_c; |
lum8x8 = lum_8x8_c; |
2063 |
lum2x8 = lum_2x8_c; |
lum2x8 = lum_2x8_c; |
2064 |
csim = iconsim_c; |
csim = consim_c; |
2065 |
|
#ifdef ARCH_IS_IA32 |
2066 |
if (cpu->cpu & XVID_CPU_MMX){ |
if (cpu->cpu & XVID_CPU_MMX){ |
2067 |
lum8x8 = lum_8x8_mmx; |
lum8x8 = lum_8x8_mmx; |
2068 |
csim = consim_mmx; |
csim = consim_mmx; |
2070 |
if (cpu->cpu & XVID_CPU_MMX){ |
if (cpu->cpu & XVID_CPU_MMX){ |
2071 |
csim = consim_sse2; |
csim = consim_sse2; |
2072 |
} |
} |
2073 |
|
#endif |
2074 |
t = gettime_usec(); |
t = gettime_usec(); |
2075 |
emms(); |
emms(); |
2076 |
for(tst=0; tst<nb_tests; ++tst) m = lum8x8(Ref1, 16); |
for(tst=0; tst<nb_tests; ++tst) m = lum8x8(Ref1, 16); |
2082 |
|
|
2083 |
t = gettime_usec(); |
t = gettime_usec(); |
2084 |
emms(); |
emms(); |
2085 |
for(tst=0; tst<nb_tests; ++tst) m = lum2x8(Ref1, 16); |
for(tst=0; tst<nb_tests; ++tst) m = lum2x8(Ref1+8, 16); |
2086 |
emms(); |
emms(); |
2087 |
t = (gettime_usec() - t) / nb_tests; |
t = (gettime_usec() - t) / nb_tests; |
2088 |
printf("%s - ssim-lum2x8 %.3f usec m=%d %s\n", |
printf("%s - ssim-lum2x8 %.3f usec m=%d %s\n", |
2089 |
cpu->name, t, m, |
cpu->name, t, m, |
2090 |
(m!=-841)?"| ERROR": "" ); |
(m!=681)?"| ERROR": "" ); |
2091 |
|
|
2092 |
t = gettime_usec(); |
t = gettime_usec(); |
2093 |
emms(); |
emms(); |
2096 |
t = (gettime_usec() - t) / nb_tests; |
t = (gettime_usec() - t) / nb_tests; |
2097 |
printf("%s - ssim-consim %.3f usec devs=[0x%x 0x%x 0x%x] %s\n", |
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], |
cpu->name, t, devs[0], devs[1], devs[2], |
2099 |
(devs[0]!=0xeba80 || devs[1]!=0x1053e7 || devs[2]!=0x51215)?"| ERROR": "" ); |
(devs[0]!=0x1bdf0f || devs[1]!=0x137258 || devs[2]!=0xcdb13)?"| ERROR": "" ); |
|
|
|
2100 |
printf( " --- \n" ); |
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 |
|
|
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_MMXEXT | 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; |
2230 |
if (what==0 || what==13) test_compiler(); |
if (what==0 || what==13) test_compiler(); |
2231 |
if (what==0 || what==14) test_yuv(); |
if (what==0 || what==14) test_yuv(); |
2232 |
if (what==0 || what==15) test_SSIM(); |
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); |