26 |
/***************************************************************************** |
/***************************************************************************** |
27 |
* |
* |
28 |
* 'Reference' output is at the end of file. |
* 'Reference' output is at the end of file. |
|
* Don't take the checksums and crc too seriouly, they aren't |
|
|
* bullet-proof (should plug some .md5 here)... |
|
29 |
* |
* |
30 |
* compiles with something like: |
* compiles with something like: |
31 |
* gcc -o xvid_bench xvid_bench.c -I../src/ -lxvidcore -lm |
* gcc -o xvid_bench xvid_bench.c -I../src/ -lxvidcore -lm |
65 |
#define M_PI 3.14159265358979323846 |
#define M_PI 3.14159265358979323846 |
66 |
#endif |
#endif |
67 |
|
|
68 |
const int speed_ref = 100; /* on slow machines, decrease this value */ |
int speed_ref = 100; /* on slow machines, decrease this value */ |
69 |
|
int verbose = 0; |
70 |
|
unsigned int cpu_mask; |
71 |
|
|
72 |
/********************************************************************* |
/********************************************************************* |
73 |
* misc |
* misc |
83 |
#else |
#else |
84 |
clock_t clk; |
clock_t clk; |
85 |
clk = clock(); |
clk = clock(); |
86 |
return clk * 1000000 / CLOCKS_PER_SEC; |
return clk * 1000. / CLOCKS_PER_SEC; /* clock() returns time in Milliseconds */ |
87 |
#endif |
#endif |
88 |
} |
} |
89 |
|
|
112 |
unsigned int cpu; |
unsigned int cpu; |
113 |
} CPU; |
} CPU; |
114 |
|
|
115 |
CPU cpu_list[] = |
CPU cpu_list[] = { |
116 |
{ { "PLAINC", 0 } |
{ "PLAINC ", 0 }, |
117 |
#ifdef ARCH_IS_IA32 |
#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_3DNOW | XVID_CPU_3DNOWEXT }, |
123 |
#endif |
#endif |
124 |
//, { "IA64 ", XVID_CPU_IA64 } |
#ifdef ARCH_IS_PPC |
125 |
//, { "TSC ", XVID_CPU_TSC } |
{ "ALTIVEC", XVID_CPU_ALTIVEC }, |
|
, { 0, 0 } }; |
|
|
|
|
|
CPU cpu_short_list[] = |
|
|
{ { "PLAINC", 0 } |
|
|
#ifdef ARCH_IS_IA32 |
|
|
, { "MMX ", XVID_CPU_MMX } |
|
|
//, { "MMXEXT", XVID_CPU_MMXEXT | XVID_CPU_MMX } |
|
126 |
#endif |
#endif |
127 |
//, { "IA64 ", XVID_CPU_IA64 } |
#ifdef ARCH_IS_X86_64 |
128 |
, { 0, 0 } }; |
{ "X86_64 ", XVID_CPU_ASM}, |
129 |
|
#endif |
130 |
CPU cpu_short_list2[] = |
#ifdef ARCH_IS_IA64 |
131 |
{ { "PLAINC", 0 } |
// { "IA64 ", XVID_CPU_IA64 }, |
|
#ifdef ARCH_IS_IA32 |
|
|
, { "MMX ", XVID_CPU_MMX } |
|
|
, { "SSE2 ", XVID_CPU_SSE2 | XVID_CPU_MMX } |
|
132 |
#endif |
#endif |
133 |
, { 0, 0 } }; |
// { "TSC ", XVID_CPU_TSC }, |
134 |
|
{ 0, 0 } |
135 |
|
}; |
136 |
|
|
137 |
|
|
138 |
int init_cpu(CPU *cpu) |
int init_cpu(CPU *cpu) |
163 |
return 1; |
return 1; |
164 |
} |
} |
165 |
|
|
166 |
|
#define CRC32_REMAINDER 0xCBF43926 |
167 |
|
#define CRC32_INITIAL 0xffffffff |
168 |
|
|
169 |
|
#define DO1(c, crc) ((crc) = crc32tab[((unsigned int)((crc)>>24) ^ (*c++)) & 0xff] ^ ((crc) << 8)) |
170 |
|
#define DO2(c, crc) DO1(c, crc); DO1(c, crc); |
171 |
|
#define DO4(c, crc) DO2(c, crc); DO2(c, crc); |
172 |
|
#define DO8(c, crc) DO4(c, crc); DO4(c, crc); |
173 |
|
|
174 |
|
/****************************************************************************** |
175 |
|
* Precomputed AAL5 CRC32 lookup table |
176 |
|
******************************************************************************/ |
177 |
|
|
178 |
|
static unsigned long crc32tab[256] = { |
179 |
|
|
180 |
|
0x00000000L, 0x04C11DB7L, 0x09823B6EL, 0x0D4326D9L, |
181 |
|
0x130476DCL, 0x17C56B6BL, 0x1A864DB2L, 0x1E475005L, |
182 |
|
0x2608EDB8L, 0x22C9F00FL, 0x2F8AD6D6L, 0x2B4BCB61L, |
183 |
|
0x350C9B64L, 0x31CD86D3L, 0x3C8EA00AL, 0x384FBDBDL, |
184 |
|
0x4C11DB70L, 0x48D0C6C7L, 0x4593E01EL, 0x4152FDA9L, |
185 |
|
0x5F15ADACL, 0x5BD4B01BL, 0x569796C2L, 0x52568B75L, |
186 |
|
0x6A1936C8L, 0x6ED82B7FL, 0x639B0DA6L, 0x675A1011L, |
187 |
|
0x791D4014L, 0x7DDC5DA3L, 0x709F7B7AL, 0x745E66CDL, |
188 |
|
0x9823B6E0L, 0x9CE2AB57L, 0x91A18D8EL, 0x95609039L, |
189 |
|
0x8B27C03CL, 0x8FE6DD8BL, 0x82A5FB52L, 0x8664E6E5L, |
190 |
|
0xBE2B5B58L, 0xBAEA46EFL, 0xB7A96036L, 0xB3687D81L, |
191 |
|
0xAD2F2D84L, 0xA9EE3033L, 0xA4AD16EAL, 0xA06C0B5DL, |
192 |
|
0xD4326D90L, 0xD0F37027L, 0xDDB056FEL, 0xD9714B49L, |
193 |
|
0xC7361B4CL, 0xC3F706FBL, 0xCEB42022L, 0xCA753D95L, |
194 |
|
0xF23A8028L, 0xF6FB9D9FL, 0xFBB8BB46L, 0xFF79A6F1L, |
195 |
|
0xE13EF6F4L, 0xE5FFEB43L, 0xE8BCCD9AL, 0xEC7DD02DL, |
196 |
|
0x34867077L, 0x30476DC0L, 0x3D044B19L, 0x39C556AEL, |
197 |
|
0x278206ABL, 0x23431B1CL, 0x2E003DC5L, 0x2AC12072L, |
198 |
|
0x128E9DCFL, 0x164F8078L, 0x1B0CA6A1L, 0x1FCDBB16L, |
199 |
|
0x018AEB13L, 0x054BF6A4L, 0x0808D07DL, 0x0CC9CDCAL, |
200 |
|
0x7897AB07L, 0x7C56B6B0L, 0x71159069L, 0x75D48DDEL, |
201 |
|
0x6B93DDDBL, 0x6F52C06CL, 0x6211E6B5L, 0x66D0FB02L, |
202 |
|
0x5E9F46BFL, 0x5A5E5B08L, 0x571D7DD1L, 0x53DC6066L, |
203 |
|
0x4D9B3063L, 0x495A2DD4L, 0x44190B0DL, 0x40D816BAL, |
204 |
|
0xACA5C697L, 0xA864DB20L, 0xA527FDF9L, 0xA1E6E04EL, |
205 |
|
0xBFA1B04BL, 0xBB60ADFCL, 0xB6238B25L, 0xB2E29692L, |
206 |
|
0x8AAD2B2FL, 0x8E6C3698L, 0x832F1041L, 0x87EE0DF6L, |
207 |
|
0x99A95DF3L, 0x9D684044L, 0x902B669DL, 0x94EA7B2AL, |
208 |
|
0xE0B41DE7L, 0xE4750050L, 0xE9362689L, 0xEDF73B3EL, |
209 |
|
0xF3B06B3BL, 0xF771768CL, 0xFA325055L, 0xFEF34DE2L, |
210 |
|
0xC6BCF05FL, 0xC27DEDE8L, 0xCF3ECB31L, 0xCBFFD686L, |
211 |
|
0xD5B88683L, 0xD1799B34L, 0xDC3ABDEDL, 0xD8FBA05AL, |
212 |
|
0x690CE0EEL, 0x6DCDFD59L, 0x608EDB80L, 0x644FC637L, |
213 |
|
0x7A089632L, 0x7EC98B85L, 0x738AAD5CL, 0x774BB0EBL, |
214 |
|
0x4F040D56L, 0x4BC510E1L, 0x46863638L, 0x42472B8FL, |
215 |
|
0x5C007B8AL, 0x58C1663DL, 0x558240E4L, 0x51435D53L, |
216 |
|
0x251D3B9EL, 0x21DC2629L, 0x2C9F00F0L, 0x285E1D47L, |
217 |
|
0x36194D42L, 0x32D850F5L, 0x3F9B762CL, 0x3B5A6B9BL, |
218 |
|
0x0315D626L, 0x07D4CB91L, 0x0A97ED48L, 0x0E56F0FFL, |
219 |
|
0x1011A0FAL, 0x14D0BD4DL, 0x19939B94L, 0x1D528623L, |
220 |
|
0xF12F560EL, 0xF5EE4BB9L, 0xF8AD6D60L, 0xFC6C70D7L, |
221 |
|
0xE22B20D2L, 0xE6EA3D65L, 0xEBA91BBCL, 0xEF68060BL, |
222 |
|
0xD727BBB6L, 0xD3E6A601L, 0xDEA580D8L, 0xDA649D6FL, |
223 |
|
0xC423CD6AL, 0xC0E2D0DDL, 0xCDA1F604L, 0xC960EBB3L, |
224 |
|
0xBD3E8D7EL, 0xB9FF90C9L, 0xB4BCB610L, 0xB07DABA7L, |
225 |
|
0xAE3AFBA2L, 0xAAFBE615L, 0xA7B8C0CCL, 0xA379DD7BL, |
226 |
|
0x9B3660C6L, 0x9FF77D71L, 0x92B45BA8L, 0x9675461FL, |
227 |
|
0x8832161AL, 0x8CF30BADL, 0x81B02D74L, 0x857130C3L, |
228 |
|
0x5D8A9099L, 0x594B8D2EL, 0x5408ABF7L, 0x50C9B640L, |
229 |
|
0x4E8EE645L, 0x4A4FFBF2L, 0x470CDD2BL, 0x43CDC09CL, |
230 |
|
0x7B827D21L, 0x7F436096L, 0x7200464FL, 0x76C15BF8L, |
231 |
|
0x68860BFDL, 0x6C47164AL, 0x61043093L, 0x65C52D24L, |
232 |
|
0x119B4BE9L, 0x155A565EL, 0x18197087L, 0x1CD86D30L, |
233 |
|
0x029F3D35L, 0x065E2082L, 0x0B1D065BL, 0x0FDC1BECL, |
234 |
|
0x3793A651L, 0x3352BBE6L, 0x3E119D3FL, 0x3AD08088L, |
235 |
|
0x2497D08DL, 0x2056CD3AL, 0x2D15EBE3L, 0x29D4F654L, |
236 |
|
0xC5A92679L, 0xC1683BCEL, 0xCC2B1D17L, 0xC8EA00A0L, |
237 |
|
0xD6AD50A5L, 0xD26C4D12L, 0xDF2F6BCBL, 0xDBEE767CL, |
238 |
|
0xE3A1CBC1L, 0xE760D676L, 0xEA23F0AFL, 0xEEE2ED18L, |
239 |
|
0xF0A5BD1DL, 0xF464A0AAL, 0xF9278673L, 0xFDE69BC4L, |
240 |
|
0x89B8FD09L, 0x8D79E0BEL, 0x803AC667L, 0x84FBDBD0L, |
241 |
|
0x9ABC8BD5L, 0x9E7D9662L, 0x933EB0BBL, 0x97FFAD0CL, |
242 |
|
0xAFB010B1L, 0xAB710D06L, 0xA6322BDFL, 0xA2F33668L, |
243 |
|
0xBCB4666DL, 0xB8757BDAL, 0xB5365D03L, 0xB1F740B4L |
244 |
|
|
245 |
|
}; |
246 |
|
|
247 |
|
uint32_t |
248 |
|
calc_crc(uint8_t *mem, int len, uint32_t crc) |
249 |
|
{ |
250 |
|
while( len >= 8) { |
251 |
|
DO8(mem, crc); |
252 |
|
len -= 8; |
253 |
|
} |
254 |
|
|
255 |
|
while( len ) { |
256 |
|
DO1(mem, crc); |
257 |
|
len--; |
258 |
|
} |
259 |
|
|
260 |
|
return crc; |
261 |
|
} |
262 |
|
|
263 |
|
void byte_swap(uint8_t *mem, int len, int element_size) { |
264 |
|
#ifdef ARCH_IS_BIG_ENDIAN |
265 |
|
int i; |
266 |
|
|
267 |
|
if(element_size == 1) { |
268 |
|
/* No need to swap */ |
269 |
|
} else if(element_size == 2) { |
270 |
|
uint8_t temp[2]; |
271 |
|
|
272 |
|
for(i=0; i < (len/2); i++ ) { |
273 |
|
temp[0] = mem[0]; |
274 |
|
temp[1] = mem[1]; |
275 |
|
mem[0] = temp[1]; |
276 |
|
mem[1] = temp[0]; |
277 |
|
|
278 |
|
mem += 2; |
279 |
|
} |
280 |
|
} else if(element_size == 4) { |
281 |
|
uint8_t temp[4]; |
282 |
|
|
283 |
|
for(i=0; i < (len/4); i++ ) { |
284 |
|
temp[0] = mem[0]; |
285 |
|
temp[1] = mem[1]; |
286 |
|
temp[2] = mem[2]; |
287 |
|
temp[3] = mem[3]; |
288 |
|
mem[0] = temp[3]; |
289 |
|
mem[1] = temp[2]; |
290 |
|
mem[2] = temp[1]; |
291 |
|
mem[3] = temp[0]; |
292 |
|
|
293 |
|
mem += 4; |
294 |
|
} |
295 |
|
} else { |
296 |
|
printf("ERROR: byte_swap unsupported element_size(%u)\n", element_size); |
297 |
|
} |
298 |
|
#endif |
299 |
|
} |
300 |
|
|
301 |
/********************************************************************* |
/********************************************************************* |
302 |
* test DCT |
* test DCT |
303 |
*********************************************************************/ |
*********************************************************************/ |
350 |
MSE += delta*delta; |
MSE += delta*delta; |
351 |
} |
} |
352 |
PSNR = (MSE==0.) ? 1.e6 : -4.3429448*log( MSE/64. ); |
PSNR = (MSE==0.) ? 1.e6 : -4.3429448*log( MSE/64. ); |
353 |
printf( "%s - %.3f usec PSNR=%.3f MSE=%.3f\n", |
printf( "%s - %.3f usec PSNR=%.3f MSE=%.3f %s\n", |
354 |
cpu->name, t, PSNR, MSE ); |
cpu->name, t, PSNR, MSE, |
355 |
if (ABS(MSE)>=64) printf( "*** CRC ERROR! ***\n" ); |
(ABS(MSE)>=64)? "| ERROR" :""); |
356 |
} |
} |
357 |
} |
} |
358 |
|
|
366 |
int tst; |
int tst; |
367 |
CPU *cpu; |
CPU *cpu; |
368 |
int i; |
int i; |
369 |
uint8_t Cur[16*16], Ref1[16*16], Ref2[16*16]; |
DECLARE_ALIGNED_MATRIX(Cur, 16, 16, uint8_t, 16); |
370 |
|
DECLARE_ALIGNED_MATRIX(Ref1, 16, 16, uint8_t, 16); |
371 |
|
DECLARE_ALIGNED_MATRIX(Ref2, 16, 16, uint8_t, 16); |
372 |
|
|
373 |
printf( "\n ====== test SAD ======\n" ); |
printf( "\n ====== test SAD ======\n" ); |
374 |
for(i=0; i<16*16;++i) { |
for(i=0; i<16*16;++i) { |
389 |
for(tst=0; tst<nb_tests; ++tst) s = sad8(Cur, Ref1, 16); |
for(tst=0; tst<nb_tests; ++tst) s = sad8(Cur, Ref1, 16); |
390 |
emms(); |
emms(); |
391 |
t = (gettime_usec() - t) / nb_tests; |
t = (gettime_usec() - t) / nb_tests; |
392 |
printf( "%s - sad8 %.3f usec sad=%d\n", cpu->name, t, s ); |
printf("%s - sad8 %.3f usec sad=%d %s\n", |
393 |
if (s!=3776) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, s, |
394 |
|
(s!=3776)?"| ERROR": "" ); |
395 |
|
|
396 |
t = gettime_usec(); |
t = gettime_usec(); |
397 |
emms(); |
emms(); |
398 |
for(tst=0; tst<nb_tests; ++tst) s = sad16(Cur, Ref1, 16, -1); |
for(tst=0; tst<nb_tests; ++tst) s = sad16(Cur, Ref1, 16, -1); |
399 |
emms(); |
emms(); |
400 |
t = (gettime_usec() - t) / nb_tests; |
t = (gettime_usec() - t) / nb_tests; |
401 |
printf( "%s - sad16 %.3f usec sad=%d\n", cpu->name, t, s ); |
printf("%s - sad16 %.3f usec sad=%d %s\n", |
402 |
if (s!=27214) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, s, |
403 |
|
(s!=27214)?"| ERROR": "" ); |
404 |
|
|
405 |
t = gettime_usec(); |
t = gettime_usec(); |
406 |
emms(); |
emms(); |
407 |
for(tst=0; tst<nb_tests; ++tst) s = sad16bi(Cur, Ref1, Ref2, 16); |
for(tst=0; tst<nb_tests; ++tst) s = sad16bi(Cur, Ref1, Ref2, 16); |
408 |
emms(); |
emms(); |
409 |
t = (gettime_usec() - t) / nb_tests; |
t = (gettime_usec() - t) / nb_tests; |
410 |
printf( "%s - sad16bi %.3f usec sad=%d\n", cpu->name, t, s ); |
printf( "%s - sad16bi %.3f usec sad=%d %s\n", |
411 |
if (s!=26274) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, s, |
412 |
|
(s!=26274)?"| ERROR": "" ); |
413 |
|
|
414 |
t = gettime_usec(); |
t = gettime_usec(); |
415 |
emms(); |
emms(); |
416 |
for(tst=0; tst<nb_tests; ++tst) s = dev16(Cur, 16); |
for(tst=0; tst<nb_tests; ++tst) s = dev16(Cur, 16); |
417 |
emms(); |
emms(); |
418 |
t = (gettime_usec() - t) / nb_tests; |
t = (gettime_usec() - t) / nb_tests; |
419 |
printf( "%s - dev16 %.3f usec sad=%d\n", cpu->name, t, s ); |
printf( "%s - dev16 %.3f usec sad=%d %s\n", |
420 |
if (s!=3344) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, s, |
421 |
|
(s!=3344)?"| ERROR": "" ); |
422 |
|
|
423 |
printf( " --- \n" ); |
printf( " --- \n" ); |
424 |
} |
} |
436 |
#define LEAVE \ |
#define LEAVE \ |
437 |
emms(); \ |
emms(); \ |
438 |
t = (gettime_usec() - t) / nb_tests; \ |
t = (gettime_usec() - t) / nb_tests; \ |
439 |
iCrc = 0; \ |
iCrc = calc_crc((uint8_t*)Dst, sizeof(Dst), CRC32_INITIAL) |
|
for(i=0; i<16*8; ++i) { iCrc += Dst[i]^i; } |
|
440 |
|
|
441 |
#define TEST_MB(FUNC, R) \ |
#define TEST_MB(FUNC, R) \ |
442 |
ENTER \ |
ENTER \ |
455 |
CPU *cpu; |
CPU *cpu; |
456 |
const uint8_t Src0[16*9] = { |
const uint8_t Src0[16*9] = { |
457 |
/* try to have every possible combinaison of rounding... */ |
/* try to have every possible combinaison of rounding... */ |
458 |
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, |
459 |
, 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, |
460 |
, 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, |
461 |
, 0, 3, 1, 3, 2, 3, 3, 3, 1 ,0,0,0, 0,0,0,0 |
0, 3, 1, 3, 2, 3, 3, 3, 1 ,0,0,0, 0,0,0,0, |
462 |
, 1, 3, 0, 2, 1, 0, 2, 3, 4 ,0,0,0, 0,0,0,0 |
1, 3, 0, 2, 1, 0, 2, 3, 4 ,0,0,0, 0,0,0,0, |
463 |
, 2, 2, 1, 2, 0, 1, 3, 5, 3 ,0,0,0, 0,0,0,0 |
2, 2, 1, 2, 0, 1, 3, 5, 3 ,0,0,0, 0,0,0,0, |
464 |
, 3, 1, 2, 3, 1, 2, 2, 6, 2 ,0,0,0, 0,0,0,0 |
3, 1, 2, 3, 1, 2, 2, 6, 2 ,0,0,0, 0,0,0,0, |
465 |
, 1, 0, 1, 3, 0, 3, 1, 6, 1 ,0,0,0, 0,0,0,0 |
1, 0, 1, 3, 0, 3, 1, 6, 1 ,0,0,0, 0,0,0,0, |
466 |
, 4, 3, 2, 1, 2, 3, 4, 0, 3 ,0,0,0, 0,0,0,0 |
4, 3, 2, 1, 2, 3, 4, 0, 3 ,0,0,0, 0,0,0,0 |
467 |
}; |
}; |
468 |
uint8_t Dst[16*8] = {0}; |
uint8_t Dst[16*8] = {0}; |
469 |
|
|
478 |
continue; |
continue; |
479 |
|
|
480 |
TEST_MB(interpolate8x8_halfpel_h, 0); |
TEST_MB(interpolate8x8_halfpel_h, 0); |
481 |
printf( "%s - interp- h-round0 %.3f usec iCrc=%d\n", cpu->name, t, iCrc ); |
printf("%s - interp- h-round0 %.3f usec crc32=0x%08x %s\n", |
482 |
if (iCrc!=8107) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, iCrc, |
483 |
|
(iCrc!=0x115381ba)?"| ERROR": "" ); |
484 |
|
|
485 |
TEST_MB(interpolate8x8_halfpel_h, 1); |
TEST_MB(interpolate8x8_halfpel_h, 1); |
486 |
printf( "%s - round1 %.3f usec iCrc=%d\n", cpu->name, t, iCrc ); |
printf("%s - round1 %.3f usec crc32=0x%08x %s\n", |
487 |
if (iCrc!=8100) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, iCrc, |
488 |
|
(iCrc!=0x2b1f528f)?"| ERROR": "" ); |
489 |
|
|
490 |
|
|
491 |
TEST_MB(interpolate8x8_halfpel_v, 0); |
TEST_MB(interpolate8x8_halfpel_v, 0); |
492 |
printf( "%s - interp- v-round0 %.3f usec iCrc=%d\n", cpu->name, t, iCrc ); |
printf("%s - interp- v-round0 %.3f usec crc32=0x%08x %s\n", |
493 |
if (iCrc!=8108) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, iCrc, |
494 |
|
(iCrc!=0x423cdcc7)?"| ERROR": "" ); |
495 |
|
|
496 |
TEST_MB(interpolate8x8_halfpel_v, 1); |
TEST_MB(interpolate8x8_halfpel_v, 1); |
497 |
printf( "%s - round1 %.3f usec iCrc=%d\n", cpu->name, t, iCrc ); |
printf("%s - round1 %.3f usec crc32=0x%08x %s\n", |
498 |
if (iCrc!=8105) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, iCrc, |
499 |
|
(iCrc!=0x42202efe)?"| ERROR": "" ); |
500 |
|
|
501 |
|
|
502 |
TEST_MB(interpolate8x8_halfpel_hv, 0); |
TEST_MB(interpolate8x8_halfpel_hv, 0); |
503 |
printf( "%s - interp-hv-round0 %.3f usec iCrc=%d\n", cpu->name, t, iCrc ); |
printf("%s - interp-hv-round0 %.3f usec crc32=0x%08x %s\n", |
504 |
if (iCrc!=8112) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, iCrc, |
505 |
|
(iCrc!=0xd198d387)?"| ERROR": "" ); |
506 |
|
|
507 |
TEST_MB(interpolate8x8_halfpel_hv, 1); |
TEST_MB(interpolate8x8_halfpel_hv, 1); |
508 |
printf( "%s - round1 %.3f usec iCrc=%d\n", cpu->name, t, iCrc ); |
printf("%s - round1 %.3f usec crc32=0x%08x %s\n", |
509 |
if (iCrc!=8103) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, iCrc, |
510 |
|
(iCrc!=0x9ecfd921)?"| ERROR": "" ); |
511 |
|
|
512 |
|
|
513 |
/* this is a new function, as of 06.06.2002 */ |
/* this is a new function, as of 06.06.2002 */ |
514 |
#if 0 |
#if 0 |
515 |
TEST_MB2(interpolate8x8_avrg); |
TEST_MB2(interpolate8x8_avrg); |
516 |
printf( "%s - interpolate8x8_c %.3f usec iCrc=%d\n", cpu->name, t, iCrc ); |
printf("%s - interpolate8x8_c %.3f usec crc32=0x%08x %s\n", |
517 |
if (iCrc!=8107) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, iCrc, |
518 |
|
(iCrc!=8107)?"| ERROR": "" ); |
519 |
#endif |
#endif |
520 |
|
|
521 |
printf( " --- \n" ); |
printf( " --- \n" ); |
551 |
} \ |
} \ |
552 |
emms(); \ |
emms(); \ |
553 |
t = (gettime_usec()-t -overhead) / nb_tests;\ |
t = (gettime_usec()-t -overhead) / nb_tests;\ |
554 |
s = 0; for(i=0; i<8*32; ++i) { s += (DST)[i]^i; } |
byte_swap((uint8_t*)(DST), 8*32*sizeof((DST)[0]), sizeof((DST)[0])); \ |
555 |
|
s = calc_crc((uint8_t*)(DST), 8*32*sizeof((DST)[0]), CRC32_INITIAL) |
556 |
|
|
557 |
#define TEST_TRANSFER(FUNC, DST, SRC) \ |
#define TEST_TRANSFER(FUNC, DST, SRC) \ |
558 |
TEST_TRANSFER_BEGIN(DST); \ |
TEST_TRANSFER_BEGIN(DST); \ |
578 |
} \ |
} \ |
579 |
emms(); \ |
emms(); \ |
580 |
t = (gettime_usec()-t -overhead) / nb_tests;\ |
t = (gettime_usec()-t -overhead) / nb_tests;\ |
581 |
s = 0; for(i=0; i<8*32; ++i) { s += (DST)[i]; } |
byte_swap((uint8_t*)(DST), 8*32*sizeof((DST)[0]), sizeof((DST)[0])); \ |
582 |
|
s = calc_crc((uint8_t*)(DST), 8*32*sizeof((DST)[0]), CRC32_INITIAL) |
583 |
|
|
584 |
#define TEST_TRANSFER2(FUNC, DST, SRC, R1) \ |
#define TEST_TRANSFER2(FUNC, DST, SRC, R1) \ |
585 |
TEST_TRANSFER2_BEGIN(DST,SRC); \ |
TEST_TRANSFER2_BEGIN(DST,SRC); \ |
596 |
const int nb_tests = 4000*speed_ref; |
const int nb_tests = 4000*speed_ref; |
597 |
int i; |
int i; |
598 |
CPU *cpu; |
CPU *cpu; |
599 |
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]; |
600 |
int16_t Src16[8*32], Dst16[8*32]; |
// int16_t Src16[8*32], Dst16[8*32]; |
601 |
|
DECLARE_ALIGNED_MATRIX(Src8, 8, 32, uint8_t, CACHE_LINE); |
602 |
|
DECLARE_ALIGNED_MATRIX(Dst8, 8, 32, uint8_t, CACHE_LINE); |
603 |
|
DECLARE_ALIGNED_MATRIX(Ref1, 8, 32, uint8_t, CACHE_LINE); |
604 |
|
DECLARE_ALIGNED_MATRIX(Ref2, 8, 32, uint8_t, CACHE_LINE); |
605 |
|
DECLARE_ALIGNED_MATRIX(Src16, 8, 32, uint16_t, CACHE_LINE); |
606 |
|
DECLARE_ALIGNED_MATRIX(Dst16, 8, 32, uint16_t, CACHE_LINE); |
607 |
|
|
608 |
printf( "\n === test transfer ===\n" ); |
printf( "\n === test transfer ===\n" ); |
609 |
|
|
616 |
continue; |
continue; |
617 |
|
|
618 |
TEST_TRANSFER(transfer_8to16copy, Dst16, Src8); |
TEST_TRANSFER(transfer_8to16copy, Dst16, Src8); |
619 |
printf( "%s - 8to16 %.3f usec crc=%d\n", cpu->name, t, s ); |
printf("%s - 8to16 %.3f usec crc32=0x%08x %s\n", |
620 |
if (s!=28288) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, s, |
621 |
|
(s!=0x115814bb)?"| ERROR": ""); |
622 |
|
|
623 |
TEST_TRANSFER(transfer_16to8copy, Dst8, Src16); |
TEST_TRANSFER(transfer_16to8copy, Dst8, Src16); |
624 |
printf( "%s - 16to8 %.3f usec crc=%d\n", cpu->name, t, s ); |
printf( "%s - 16to8 %.3f usec crc32=0x%08x %s\n", |
625 |
if (s!=28288) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, s, |
626 |
|
(s!=0xee7ccbb4)?"| ERROR": ""); |
627 |
|
|
628 |
TEST_TRANSFER(transfer8x8_copy, Dst8, Src8); |
TEST_TRANSFER(transfer8x8_copy, Dst8, Src8); |
629 |
printf( "%s - 8to8 %.3f usec crc=%d\n", cpu->name, t, s ); |
printf("%s - 8to8 %.3f usec crc32=0x%08x %s\n", |
630 |
if (s!=20352) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, s, |
631 |
|
(s!=0xd37b3295)?"| ERROR": ""); |
632 |
|
|
633 |
TEST_TRANSFER(transfer_16to8add, Dst8, Src16); |
TEST_TRANSFER(transfer_16to8add, Dst8, Src16); |
634 |
printf( "%s - 16to8add %.3f usec crc=%d\n", cpu->name, t, s ); |
printf("%s - 16to8add %.3f usec crc32=0x%08x %s\n", |
635 |
if (s!=25536) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, s, |
636 |
|
(s!=0xdd817bf4)?"| ERROR": "" ); |
637 |
|
|
638 |
TEST_TRANSFER2(transfer_8to16sub, Dst16, Src8, Ref1); |
TEST_TRANSFER2(transfer_8to16sub, Dst16, Src8, Ref1); |
639 |
printf( "%s - 8to16sub %.3f usec crc1=%d ", cpu->name, t, s ); |
{ |
640 |
if (s!=28064) printf( "*** CRC ERROR! ***\n" ); |
int s1, s2; |
641 |
s = 0; for(i=0; i<8*32; ++i) { s += (Src8[i]-Ref1[i])&i; } |
s1 = calc_crc((uint8_t*)Dst16, 8*32*sizeof(Dst16[0]), CRC32_INITIAL); |
642 |
printf( "crc2=%d\n", s); |
s2 = calc_crc((uint8_t*)Src8, 8*32*sizeof(Src8[0]), CRC32_INITIAL); |
643 |
if (s!=16256) printf( "*** CRC ERROR! ***\n" ); |
printf("%s - 8to16sub %.3f usec crc32(1)=0x%08x crc32(2)=0x%08x %s %s\n", |
644 |
|
cpu->name, t, s1, s2, |
645 |
|
(s1!=0xa1e07163)?"| ERROR1": "", |
646 |
|
(s2!=0xd86c5d23)?"| ERROR2": "" ); |
647 |
|
} |
648 |
|
|
649 |
TEST_TRANSFER3(transfer_8to16sub2, Dst16, Src8, Ref1, Ref2); |
TEST_TRANSFER3(transfer_8to16sub2, Dst16, Src8, Ref1, Ref2); |
650 |
printf( "%s - 8to16sub2 %.3f usec crc=%d\n", cpu->name, t, s ); |
printf("%s - 8to16sub2 %.3f usec crc32=0x%08x %s\n", |
651 |
if (s!=22368) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, s, |
652 |
|
(s!=0x99b6c4c7)?"| ERROR": "" ); |
653 |
|
|
654 |
printf( " --- \n" ); |
printf( " --- \n" ); |
655 |
} |
} |
661 |
|
|
662 |
#define TEST_QUANT(FUNC, DST, SRC) \ |
#define TEST_QUANT(FUNC, DST, SRC) \ |
663 |
t = gettime_usec(); \ |
t = gettime_usec(); \ |
664 |
for(s=0,qm=1; qm<=255; ++qm) { \ |
for(s=CRC32_INITIAL,qm=1; qm<=255; ++qm) { \ |
665 |
for(i=0; i<8*8; ++i) Quant[i] = qm; \ |
for(i=0; i<8*8; ++i) Quant[i] = qm; \ |
666 |
set_inter_matrix( Quant ); \ |
set_inter_matrix( mpeg_quant_matrices, Quant ); \ |
667 |
emms(); \ |
emms(); \ |
668 |
for(q=1; q<=max_Q; ++q) { \ |
for(q=1; q<=max_Q; ++q) { \ |
669 |
for(tst=0; tst<nb_tests; ++tst) \ |
for(tst=0; tst<nb_tests; ++tst) \ |
670 |
(FUNC)((DST), (SRC), q); \ |
(FUNC)((DST), (SRC), q, mpeg_quant_matrices); \ |
671 |
for(i=0; i<64; ++i) s+=(DST)[i]^i^qm; \ |
byte_swap((uint8_t*)(DST), 64*sizeof((DST)[0]), sizeof((DST)[0])); \ |
672 |
|
s = calc_crc((uint8_t*)(DST), 64*sizeof((DST)[0]), s); \ |
673 |
} \ |
} \ |
674 |
emms(); \ |
emms(); \ |
675 |
} \ |
} \ |
676 |
t = (gettime_usec()-t-overhead)/nb_tests/qm;\ |
t = (gettime_usec()-t-overhead)/nb_tests/qm |
|
s = (s&0xffff)^(s>>16) |
|
677 |
|
|
678 |
#define TEST_QUANT2(FUNC, DST, SRC) \ |
#define TEST_QUANT2(FUNC, DST, SRC) \ |
679 |
t = gettime_usec(); \ |
t = gettime_usec(); \ |
680 |
for(s=0,qm=1; qm<=255; ++qm) { \ |
for(s=CRC32_INITIAL,qm=1; qm<=255; ++qm) { \ |
681 |
for(i=0; i<8*8; ++i) Quant[i] = qm; \ |
for(i=0; i<8*8; ++i) Quant[i] = qm; \ |
682 |
set_intra_matrix( Quant ); \ |
set_intra_matrix( mpeg_quant_matrices, Quant ); \ |
683 |
emms(); \ |
emms(); \ |
684 |
for(q=1; q<=max_Q; ++q) { \ |
for(q=1; q<=max_Q; ++q) { \ |
685 |
for(tst=0; tst<nb_tests; ++tst) \ |
for(tst=0; tst<nb_tests; ++tst) \ |
686 |
(FUNC)((DST), (SRC), q, q); \ |
(FUNC)((DST), (SRC), q, q, mpeg_quant_matrices); \ |
687 |
for(i=0; i<64; ++i) s+=(DST)[i]^i^qm; \ |
byte_swap((uint8_t*)(DST), 64*sizeof((DST)[0]), sizeof((DST)[0])); \ |
688 |
|
s = calc_crc((uint8_t*)(DST), 64*sizeof((DST)[0]), s); \ |
689 |
} \ |
} \ |
690 |
emms(); \ |
emms(); \ |
691 |
} \ |
} \ |
692 |
t = (gettime_usec()-t-overhead)/nb_tests/qm;\ |
t = (gettime_usec()-t-overhead)/nb_tests/qm |
693 |
s = (s&0xffff)^(s>>16) |
|
694 |
|
#define TEST_INTRA(REFFUNC, NEWFUNC, RANGE) \ |
695 |
|
{ int i,q,s;\ |
696 |
|
DECLARE_ALIGNED_MATRIX(Src, 8, 8, int16_t, 16); \ |
697 |
|
DECLARE_ALIGNED_MATRIX(Dst, 8, 8, int16_t, 16); \ |
698 |
|
DECLARE_ALIGNED_MATRIX(Dst2,8, 8, int16_t, 16); \ |
699 |
|
for(q=1;q<=max_Q;q++) \ |
700 |
|
for(s=-RANGE;s<RANGE;s++) { \ |
701 |
|
for(i=0;i<64;i++) Src[i]=s; \ |
702 |
|
(REFFUNC)((Dst),(Src),q,q,mpeg_quant_matrices); \ |
703 |
|
(NEWFUNC)((Dst2),(Src),q,q,mpeg_quant_matrices); \ |
704 |
|
for(i=0;i<64;i++) \ |
705 |
|
if(Dst[i]!=Dst2[i]) printf("ERROR : " #NEWFUNC " i%d quant:%d input:%d C_result:%d ASM_result:%d\n",i,q,s,Dst[i],Dst2[i]); \ |
706 |
|
} \ |
707 |
|
} |
708 |
|
|
709 |
|
#define TEST_INTER(REFFUNC, NEWFUNC, RANGE) \ |
710 |
|
{ int i,q,s; \ |
711 |
|
DECLARE_ALIGNED_MATRIX(Src, 8, 8, int16_t, 16); \ |
712 |
|
DECLARE_ALIGNED_MATRIX(Dst, 8, 8, int16_t, 16); \ |
713 |
|
DECLARE_ALIGNED_MATRIX(Dst2,8, 8, int16_t, 16); \ |
714 |
|
for(q=1;q<=max_Q;q++) \ |
715 |
|
for(s=-RANGE;s<RANGE;s++) { \ |
716 |
|
for(i=0;i<64;i++) Src[i]=s; \ |
717 |
|
(REFFUNC)((Dst),(Src),q,mpeg_quant_matrices); \ |
718 |
|
(NEWFUNC)((Dst2),(Src),q,mpeg_quant_matrices); \ |
719 |
|
emms(); \ |
720 |
|
for(i=0;i<64;i++) \ |
721 |
|
if(Dst[i]!=Dst2[i]) printf("ERROR : " #NEWFUNC " i%d quant:%d input:%d C_result:%d ASM_result:%d\n",i,q,s,Dst[i],Dst2[i]); \ |
722 |
|
} \ |
723 |
|
} |
724 |
|
|
725 |
void test_quant() |
void test_quant() |
726 |
{ |
{ |
727 |
const int nb_tests = 1*speed_ref; |
const int nb_tests = 1*speed_ref; |
728 |
const int max_Q = 31; |
const int max_Q = 31; |
729 |
|
DECLARE_ALIGNED_MATRIX(mpeg_quant_matrices, 8, 64, uint16_t, 16); |
730 |
|
|
731 |
int i, qm; |
int i, qm; |
732 |
CPU *cpu; |
CPU *cpu; |
733 |
int16_t Src[8*8], Dst[8*8]; |
DECLARE_ALIGNED_MATRIX(Src, 8, 8, int16_t, 16); |
734 |
|
DECLARE_ALIGNED_MATRIX(Dst, 8, 8, int16_t, 16); |
735 |
|
DECLARE_ALIGNED_MATRIX(Dst2,8, 8, int16_t, 16); |
736 |
uint8_t Quant[8*8]; |
uint8_t Quant[8*8]; |
737 |
|
|
738 |
printf( "\n ===== test quant =====\n" ); |
printf( "\n ===== test quant =====\n" ); |
753 |
if (!init_cpu(cpu)) |
if (!init_cpu(cpu)) |
754 |
continue; |
continue; |
755 |
|
|
756 |
|
// exhaustive tests to compare against the (ref) C-version |
757 |
|
TEST_INTRA(quant_h263_intra_c, quant_h263_intra, 2048); |
758 |
|
TEST_INTRA(dequant_h263_intra_c, dequant_h263_intra , 512 ); |
759 |
|
TEST_INTER(quant_h263_inter_c, quant_h263_inter , 2048); |
760 |
|
TEST_INTER(dequant_h263_inter_c, dequant_h263_inter , 512 ); |
761 |
|
|
762 |
overhead = -gettime_usec(); |
overhead = -gettime_usec(); |
763 |
for(s=0,qm=1; qm<=255; ++qm) { |
for(s=0,qm=1; qm<=255; ++qm) { |
764 |
for(i=0; i<8*8; ++i) Quant[i] = qm; |
for(i=0; i<8*8; ++i) Quant[i] = qm; |
765 |
set_inter_matrix( Quant ); |
set_inter_matrix(mpeg_quant_matrices, Quant ); |
766 |
for(q=1; q<=max_Q; ++q) |
for(q=1; q<=max_Q; ++q) |
767 |
for(i=0; i<64; ++i) s+=Dst[i]^i^qm; |
for(i=0; i<64; ++i) s+=Dst[i]^i^qm; |
768 |
} |
} |
769 |
overhead += gettime_usec(); |
overhead += gettime_usec(); |
770 |
|
|
|
#if 1 |
|
771 |
TEST_QUANT2(quant_mpeg_intra, Dst, Src); |
TEST_QUANT2(quant_mpeg_intra, Dst, Src); |
772 |
printf( "%s - quant_mpeg_intra %.3f usec crc=%d\n", cpu->name, t, s ); |
printf("%s - quant_mpeg_intra %.3f usec crc32=0x%08x %s\n", |
773 |
if (s!=29809) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, s, |
774 |
|
(s!=0xfd6a21a4)? "| ERROR": ""); |
775 |
|
|
776 |
TEST_QUANT(quant_mpeg_inter, Dst, Src); |
TEST_QUANT(quant_mpeg_inter, Dst, Src); |
777 |
printf( "%s - quant_mpeg_inter %.3f usec crc=%d\n", cpu->name, t, s ); |
printf("%s - quant_mpeg_inter %.3f usec crc32=0x%08x %s\n", |
778 |
if (s!=12574) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, s, |
779 |
#endif |
(s!=0xf6de7757)?"| ERROR": ""); |
780 |
#if 1 |
|
781 |
TEST_QUANT2(dequant_mpeg_intra, Dst, Src); |
TEST_QUANT2(dequant_mpeg_intra, Dst, Src); |
782 |
printf( "%s - dequant_mpeg_intra %.3f usec crc=%d\n", cpu->name, t, s ); |
printf("%s - dequant_mpeg_intra %.3f usec crc32=0x%08x %s\n", |
783 |
if (s!=24052) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, s, |
784 |
|
(s!=0x2def7bc7)?"| ERROR": ""); |
785 |
|
|
786 |
TEST_QUANT(dequant_mpeg_inter, Dst, Src); |
TEST_QUANT(dequant_mpeg_inter, Dst, Src); |
787 |
printf( "%s - dequant_mpeg_inter %.3f usec crc=%d\n", cpu->name, t, s ); |
printf("%s - dequant_mpeg_inter %.3f usec crc32=0x%08x %s\n", |
788 |
if (s!=63847) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, s, |
789 |
#endif |
(s!=0xd878c722)?"| ERROR": ""); |
790 |
#if 1 |
|
791 |
TEST_QUANT2(quant_h263_intra, Dst, Src); |
TEST_QUANT2(quant_h263_intra, Dst, Src); |
792 |
printf( "%s - quant_h263_intra %.3f usec crc=%d\n", cpu->name, t, s ); |
printf("%s - quant_h263_intra %.3f usec crc32=0x%08x %s\n", |
793 |
if (s!=25662) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, s, |
794 |
|
(s!=0x2eba9d43)?"| ERROR": ""); |
795 |
|
|
796 |
TEST_QUANT(quant_h263_inter, Dst, Src); |
TEST_QUANT(quant_h263_inter, Dst, Src); |
797 |
printf( "%s - quant_h263_inter %.3f usec crc=%d\n", cpu->name, t, s ); |
printf("%s - quant_h263_inter %.3f usec crc32=0x%08x %s\n", |
798 |
if (s!=23972) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, s, |
799 |
#endif |
(s!=0xbd315a7e)?"| ERROR": ""); |
800 |
#if 1 |
|
801 |
TEST_QUANT2(dequant_h263_intra, Dst, Src); |
TEST_QUANT2(dequant_h263_intra, Dst, Src); |
802 |
printf( "%s - dequant_h263_intra %.3f usec crc=%d\n", cpu->name, t, s ); |
printf("%s - dequant_h263_intra %.3f usec crc32=0x%08x %s\n", |
803 |
if (s!=49900) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, s, |
804 |
|
(s!=0x9841212a)?"| ERROR": ""); |
805 |
|
|
806 |
TEST_QUANT(dequant_h263_inter, Dst, Src); |
TEST_QUANT(dequant_h263_inter, Dst, Src); |
807 |
printf( "%s - dequant_h263_inter %.3f usec crc=%d\n", cpu->name, t, s ); |
printf("%s - dequant_h263_inter %.3f usec crc32=0x%08x %s\n", |
808 |
if (s!=48899) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, s, |
809 |
#endif |
(s!=0xe7df8fba)?"| ERROR": ""); |
810 |
|
|
811 |
|
printf( " --- \n" ); |
812 |
|
} |
813 |
|
} |
814 |
|
|
815 |
|
/********************************************************************* |
816 |
|
* test distortion operators |
817 |
|
*********************************************************************/ |
818 |
|
|
819 |
|
static void ieee_reseed(long s); |
820 |
|
static long ieee_rand(int Min, int Max); |
821 |
|
|
822 |
|
#define TEST_SSE(FUNCTION, SRC1, SRC2, STRIDE) \ |
823 |
|
do { \ |
824 |
|
t = gettime_usec(); \ |
825 |
|
tst = nb_tests; \ |
826 |
|
while((tst--)>0) sse = (FUNCTION)((SRC1), (SRC2), (STRIDE)); \ |
827 |
|
emms(); \ |
828 |
|
t = (gettime_usec() - t)/(double)nb_tests; \ |
829 |
|
} while(0) |
830 |
|
|
831 |
|
|
832 |
|
void test_sse() |
833 |
|
{ |
834 |
|
const int nb_tests = 100000*speed_ref; |
835 |
|
int i; |
836 |
|
CPU *cpu; |
837 |
|
DECLARE_ALIGNED_MATRIX(Src1, 8, 8, int16_t, 16); |
838 |
|
DECLARE_ALIGNED_MATRIX(Src2, 8, 8, int16_t, 16); |
839 |
|
DECLARE_ALIGNED_MATRIX(Src3, 8, 8, int16_t, 16); |
840 |
|
DECLARE_ALIGNED_MATRIX(Src4, 8, 8, int16_t, 16); |
841 |
|
|
842 |
|
printf( "\n ===== test sse =====\n" ); |
843 |
|
|
844 |
|
ieee_reseed(1); |
845 |
|
for(i=0; i<64; ++i) { |
846 |
|
Src1[i] = ieee_rand(-2048, 2047); |
847 |
|
Src2[i] = ieee_rand(-2048, 2047); |
848 |
|
Src3[i] = ieee_rand(-2048, 2047); |
849 |
|
Src4[i] = ieee_rand(-2048, 2047); |
850 |
|
} |
851 |
|
|
852 |
|
for(cpu = cpu_list; cpu->name!=0; ++cpu) |
853 |
|
{ |
854 |
|
double t; |
855 |
|
int tst, sse; |
856 |
|
|
857 |
|
if (!init_cpu(cpu)) |
858 |
|
continue; |
859 |
|
|
860 |
|
/* 16 bit element blocks */ |
861 |
|
TEST_SSE(sse8_16bit, Src1, Src2, 16); |
862 |
|
printf("%s - sse8_16bit#1 %.3f usec sse=%d %s\n", |
863 |
|
cpu->name, t, sse, (sse!=182013834)?"| ERROR": ""); |
864 |
|
TEST_SSE(sse8_16bit, Src1, Src3, 16); |
865 |
|
printf("%s - sse8_16bit#2 %.3f usec sse=%d %s\n", |
866 |
|
cpu->name, t, sse, (sse!=142545203)?"| ERROR": ""); |
867 |
|
TEST_SSE(sse8_16bit, Src1, Src4, 16); |
868 |
|
printf("%s - sse8_16bit#3 %.3f usec sse=%d %s\n", |
869 |
|
cpu->name, t, sse, (sse!=146340935)?"| ERROR": ""); |
870 |
|
TEST_SSE(sse8_16bit, Src2, Src3, 16); |
871 |
|
printf("%s - sse8_16bit#4 %.3f usec sse=%d %s\n", |
872 |
|
cpu->name, t, sse, (sse!=130136661)?"| ERROR": ""); |
873 |
|
TEST_SSE(sse8_16bit, Src2, Src4, 16); |
874 |
|
printf("%s - sse8_16bit#5 %.3f usec sse=%d %s\n", |
875 |
|
cpu->name, t, sse, (sse!=136870353)?"| ERROR": ""); |
876 |
|
TEST_SSE(sse8_16bit, Src3, Src4, 16); |
877 |
|
printf("%s - sse8_16bit#6 %.3f usec sse=%d %s\n", |
878 |
|
cpu->name, t, sse, (sse!=164107772)?"| ERROR": ""); |
879 |
|
|
880 |
|
/* 8 bit element blocks */ |
881 |
|
TEST_SSE(sse8_8bit, (int8_t*)Src1, (int8_t*)Src2, 8); |
882 |
|
printf("%s - sse8_8bit#1 %.3f usec sse=%d %s\n", |
883 |
|
cpu->name, t, sse, (sse!=1356423)?"| ERROR": ""); |
884 |
|
TEST_SSE(sse8_8bit, (int8_t*)Src1, (int8_t*)Src3, 8); |
885 |
|
printf("%s - sse8_8bit#2 %.3f usec sse=%d %s\n", |
886 |
|
cpu->name, t, sse, (sse!=1173074)?"| ERROR": ""); |
887 |
|
TEST_SSE(sse8_8bit, (int8_t*)Src1, (int8_t*)Src4, 8); |
888 |
|
printf("%s - sse8_8bit#3 %.3f usec sse=%d %s\n", |
889 |
|
cpu->name, t, sse, (sse!=1092357)?"| ERROR": ""); |
890 |
|
TEST_SSE(sse8_8bit, (int8_t*)Src2, (int8_t*)Src3, 8); |
891 |
|
printf("%s - sse8_8bit#4 %.3f usec sse=%d %s\n", |
892 |
|
cpu->name, t, sse, (sse!=1360239)?"| ERROR": ""); |
893 |
|
TEST_SSE(sse8_8bit, (int8_t*)Src2, (int8_t*)Src4, 8); |
894 |
|
printf("%s - sse8_8bit#5 %.3f usec sse=%d %s\n", |
895 |
|
cpu->name, t, sse, (sse!=1208414)?"| ERROR": ""); |
896 |
|
TEST_SSE(sse8_8bit, (int8_t*)Src3, (int8_t*)Src4, 8); |
897 |
|
printf("%s - sse8_8bit#6 %.3f usec sse=%d %s\n", |
898 |
|
cpu->name, t, sse, (sse!=1099285)?"| ERROR": ""); |
899 |
|
|
900 |
printf( " --- \n" ); |
printf( " --- \n" ); |
901 |
} |
} |
902 |
} |
} |
905 |
* test non-zero AC counting |
* test non-zero AC counting |
906 |
*********************************************************************/ |
*********************************************************************/ |
907 |
|
|
908 |
#define TEST_CBP(FUNC, SRC) \ |
#define TEST_CBP(FUNC, SRC, NB) \ |
909 |
t = gettime_usec(); \ |
t = gettime_usec(); \ |
910 |
emms(); \ |
emms(); \ |
911 |
for(tst=0; tst<nb_tests; ++tst) { \ |
for(tst=0; tst<NB; ++tst) { \ |
912 |
cbp = (FUNC)((SRC)); \ |
cbp = (FUNC)((SRC)); \ |
913 |
} \ |
} \ |
914 |
emms(); \ |
emms(); \ |
917 |
void test_cbp() |
void test_cbp() |
918 |
{ |
{ |
919 |
const int nb_tests = 10000*speed_ref; |
const int nb_tests = 10000*speed_ref; |
920 |
int i; |
int i, n, m; |
921 |
CPU *cpu; |
CPU *cpu; |
922 |
int16_t Src1[6*64], Src2[6*64], Src3[6*64], Src4[6*64]; |
DECLARE_ALIGNED_MATRIX(Src1, 6, 64, int16_t, 16); |
923 |
|
DECLARE_ALIGNED_MATRIX(Src2, 6, 64, int16_t, 16); |
924 |
|
DECLARE_ALIGNED_MATRIX(Src3, 6, 64, int16_t, 16); |
925 |
|
DECLARE_ALIGNED_MATRIX(Src4, 6, 64, int16_t, 16); |
926 |
|
DECLARE_ALIGNED_MATRIX(Src5, 6, 64, int16_t, 16); |
927 |
|
|
928 |
printf( "\n ===== test cbp =====\n" ); |
printf( "\n ===== test cbp =====\n" ); |
929 |
|
|
932 |
Src2[i] = (i<3*64); /* half-full */ |
Src2[i] = (i<3*64); /* half-full */ |
933 |
Src3[i] = ((i+32)>3*64); |
Src3[i] = ((i+32)>3*64); |
934 |
Src4[i] = (i==(3*64+2) || i==(5*64+9)); |
Src4[i] = (i==(3*64+2) || i==(5*64+9)); |
935 |
|
Src5[i] = ieee_rand(0,1) ? -1 : 1; /* +/- test */ |
936 |
} |
} |
937 |
|
|
938 |
for(cpu = cpu_list; cpu->name!=0; ++cpu) |
for(cpu = cpu_list; cpu->name!=0; ++cpu) |
943 |
if (!init_cpu(cpu)) |
if (!init_cpu(cpu)) |
944 |
continue; |
continue; |
945 |
|
|
946 |
TEST_CBP(calc_cbp, Src1); |
TEST_CBP(calc_cbp, Src1, nb_tests); |
947 |
printf( "%s - calc_cbp#1 %.3f usec cbp=0x%x\n", cpu->name, t, cbp ); |
printf("%s - calc_cbp#1 %.3f usec cbp=0x%02x %s\n", |
948 |
if (cbp!=0x15) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, cbp, (cbp!=0x15)?"| ERROR": ""); |
949 |
TEST_CBP(calc_cbp, Src2); |
TEST_CBP(calc_cbp, Src2, nb_tests); |
950 |
printf( "%s - calc_cbp#2 %.3f usec cbp=0x%x\n", cpu->name, t, cbp ); |
printf("%s - calc_cbp#2 %.3f usec cbp=0x%02x %s\n", |
951 |
if (cbp!=0x38) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, cbp, (cbp!=0x38)?"| ERROR": ""); |
952 |
TEST_CBP(calc_cbp, Src3); |
TEST_CBP(calc_cbp, Src3, nb_tests); |
953 |
printf( "%s - calc_cbp#3 %.3f usec cbp=0x%x\n", cpu->name, t, cbp ); |
printf("%s - calc_cbp#3 %.3f usec cbp=0x%02x %s\n", |
954 |
if (cbp!=0x0f) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, cbp, (cbp!=0x0f)?"| ERROR": "" ); |
955 |
TEST_CBP(calc_cbp, Src4); |
TEST_CBP(calc_cbp, Src4, nb_tests); |
956 |
printf( "%s - calc_cbp#4 %.3f usec cbp=0x%x\n", cpu->name, t, cbp ); |
printf("%s - calc_cbp#4 %.3f usec cbp=0x%02x %s\n", |
957 |
if (cbp!=0x05) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, cbp, (cbp!=0x05)?"| ERROR": "" ); |
958 |
|
TEST_CBP(calc_cbp, Src5, nb_tests); |
959 |
|
printf("%s - calc_cbp#4 %.3f usec cbp=0x%02x %s\n", |
960 |
|
cpu->name, t, cbp, (cbp!=0x3f)?"| ERROR": "" ); |
961 |
printf( " --- \n" ); |
printf( " --- \n" ); |
962 |
} |
} |
963 |
|
|
964 |
|
for(cpu = cpu_list; cpu->name!=0; ++cpu) /* bench suggested by Carlo (carlo dot bramix at libero dot it) */ |
965 |
|
{ |
966 |
|
double t; |
967 |
|
int tst, cbp, err; |
968 |
|
|
969 |
|
if (!init_cpu(cpu)) |
970 |
|
continue; |
971 |
|
|
972 |
|
err = 0; |
973 |
|
for(n=0; n<6; ++n) |
974 |
|
{ |
975 |
|
for(m=0; m<64; ++m) |
976 |
|
{ |
977 |
|
for(i=0; i<6*64; ++i) |
978 |
|
Src1[i] = (i== (m + n*64)); |
979 |
|
|
980 |
|
TEST_CBP(calc_cbp, Src1, 1); |
981 |
|
if (cbp!= (((m!=0)<<(5-n)))) |
982 |
|
{ |
983 |
|
printf( "%s - calc_cbp#5: ERROR at pos %d / %d!\n", cpu->name, n, m); |
984 |
|
err = 1; |
985 |
|
break; |
986 |
|
} |
987 |
|
} |
988 |
|
} |
989 |
|
if (!err) |
990 |
|
printf( " %s - calc_cbp#5 : OK\n", cpu->name ); |
991 |
|
|
992 |
|
} |
993 |
} |
} |
994 |
|
|
995 |
/********************************************************************* |
/********************************************************************* |
1383 |
* measure raw decoding speed |
* measure raw decoding speed |
1384 |
*********************************************************************/ |
*********************************************************************/ |
1385 |
|
|
1386 |
void test_dec(const char *name, int width, int height, int with_chksum) |
void test_dec(const char *name, int width, int height, int ref_chksum) |
1387 |
{ |
{ |
1388 |
FILE *f = 0; |
FILE *f = 0; |
1389 |
void *dechandle = 0; |
void *dechandle = 0; |
1394 |
double t = 0.; |
double t = 0.; |
1395 |
int nb = 0; |
int nb = 0; |
1396 |
uint8_t *buf = 0; |
uint8_t *buf = 0; |
1397 |
uint8_t *rgb_out = 0; |
uint8_t *yuv_out = 0; |
1398 |
int buf_size, pos; |
int buf_size, pos; |
1399 |
uint32_t chksum = 0; |
uint32_t chksum = 0; |
1400 |
|
int bps = (width+31) & ~31; |
1401 |
|
|
1402 |
memset(&xinit, 0, sizeof(xinit)); |
memset(&xinit, 0, sizeof(xinit)); |
1403 |
xinit.cpu_flags = XVID_CPU_MMX | XVID_CPU_FORCE; |
xinit.cpu_flags = cpu_mask; |
1404 |
xinit.version = XVID_VERSION; |
xinit.version = XVID_VERSION; |
1405 |
xvid_global(NULL, 0, &xinit, NULL); |
xvid_global(NULL, 0, &xinit, NULL); |
1406 |
|
|
1410 |
xparam.version = XVID_VERSION; |
xparam.version = XVID_VERSION; |
1411 |
xerr = xvid_decore(NULL, XVID_DEC_CREATE, &xparam, NULL); |
xerr = xvid_decore(NULL, XVID_DEC_CREATE, &xparam, NULL); |
1412 |
if (xerr==XVID_ERR_FAIL) { |
if (xerr==XVID_ERR_FAIL) { |
1413 |
printf("can't init decoder (err=%d)\n", xerr); |
printf("ERROR: can't init decoder (err=%d)\n", xerr); |
1414 |
return; |
return; |
1415 |
} |
} |
1416 |
dechandle = xparam.handle; |
dechandle = xparam.handle; |
1418 |
|
|
1419 |
f = fopen(name, "rb"); |
f = fopen(name, "rb"); |
1420 |
if (f==0) { |
if (f==0) { |
1421 |
printf( "can't open file '%s'\n", name); |
printf( "ERROR: can't open file '%s'\n", name); |
1422 |
return; |
return; |
1423 |
} |
} |
1424 |
fseek(f, 0, SEEK_END); |
fseek(f, 0, SEEK_END); |
1425 |
buf_size = ftell(f); |
buf_size = ftell(f); |
1426 |
fseek(f, 0, SEEK_SET); |
fseek(f, 0, SEEK_SET); |
1427 |
if (buf_size<=0) { |
if (buf_size<=0) { |
1428 |
printf("error while stating file\n"); |
printf("ERROR: error while stating file\n"); |
1429 |
goto End; |
goto End; |
1430 |
} |
} |
|
else printf( "Input size: %d\n", buf_size); |
|
1431 |
|
|
1432 |
buf = malloc(buf_size); /* should be enuf' */ |
buf = malloc(buf_size); |
1433 |
rgb_out = calloc(4, width*height); /* <-room for _RGB24 */ |
yuv_out = calloc(1, bps*height*3/2 + 15); |
1434 |
if (buf==0 || rgb_out==0) { |
if (buf==0 || yuv_out==0) { |
1435 |
printf( "malloc failed!\n" ); |
printf( "ERROR: malloc failed!\n" ); |
1436 |
goto End; |
goto End; |
1437 |
} |
} |
1438 |
|
|
1439 |
if (fread(buf, buf_size, 1, f)!=1) { |
if (fread(buf, buf_size, 1, f)!=1) { |
1440 |
printf( "file-read failed\n" ); |
printf( "ERROR: file-read failed\n" ); |
1441 |
goto End; |
goto End; |
1442 |
} |
} |
1443 |
|
|
1445 |
pos = 0; |
pos = 0; |
1446 |
t = -gettime_usec(); |
t = -gettime_usec(); |
1447 |
while(1) { |
while(1) { |
1448 |
|
int y; |
1449 |
|
|
1450 |
memset(&xframe, 0, sizeof(xframe)); |
memset(&xframe, 0, sizeof(xframe)); |
1451 |
xframe.version = XVID_VERSION; |
xframe.version = XVID_VERSION; |
1452 |
xframe.bitstream = buf + pos; |
xframe.bitstream = buf + pos; |
1453 |
xframe.length = buf_size - pos; |
xframe.length = buf_size - pos; |
1454 |
xframe.output.plane[0] = rgb_out; |
xframe.output.plane[0] = (uint8_t*)(((size_t)yuv_out + 15) & ~15); |
1455 |
xframe.output.stride[0] = width; |
xframe.output.plane[1] = (uint8_t*)xframe.output.plane[0] + bps*height; |
1456 |
xframe.output.csp = XVID_CSP_BGR; |
xframe.output.plane[2] = (uint8_t*)xframe.output.plane[1] + bps/2; |
1457 |
|
xframe.output.stride[0] = bps; |
1458 |
|
xframe.output.stride[1] = bps; |
1459 |
|
xframe.output.stride[2] = bps; |
1460 |
|
xframe.output.csp = XVID_CSP_I420; |
1461 |
xerr = xvid_decore(dechandle, XVID_DEC_DECODE, &xframe, 0); |
xerr = xvid_decore(dechandle, XVID_DEC_DECODE, &xframe, 0); |
1462 |
|
if (xerr<0) { |
1463 |
|
printf("ERROR: decoding failed for frame #%d (err=%d)!\n", nb, xerr); |
1464 |
|
break; |
1465 |
|
} |
1466 |
|
else if (xerr==0) |
1467 |
|
break; |
1468 |
|
else if (verbose>0) printf("#%d %d\n", nb, xerr ); |
1469 |
|
|
1470 |
|
pos += xerr; |
1471 |
nb++; |
nb++; |
1472 |
pos += xframe.length; |
|
1473 |
if (with_chksum) { |
for(y=0; y<height/2; ++y) { |
1474 |
int k = width*height; |
chksum = calc_crc((uint8_t*)xframe.output.plane[0] + (2*y+0)*bps, width, chksum); |
1475 |
uint32_t *ptr = (uint32_t *)rgb_out; |
chksum = calc_crc((uint8_t*)xframe.output.plane[0] + (2*y+1)*bps, width, chksum); |
1476 |
while(k-->0) chksum += *ptr++; |
chksum = calc_crc((uint8_t*)xframe.output.plane[1] + y*bps, width/2, chksum); |
1477 |
|
chksum = calc_crc((uint8_t*)xframe.output.plane[2] + y*bps, width/2, chksum); |
1478 |
} |
} |
1479 |
if (pos==buf_size) |
if (pos==buf_size) |
1480 |
break; |
break; |
|
if (xerr==XVID_ERR_FAIL) { |
|
|
printf("decoding failed for frame #%d (err=%d)!\n", nb, xerr); |
|
|
break; |
|
|
} |
|
1481 |
} |
} |
1482 |
t += gettime_usec(); |
t += gettime_usec(); |
1483 |
|
if (ref_chksum==0) { |
1484 |
if (t>0.) |
if (t>0.) |
1485 |
printf( "%d frames decoded in %.3f s -> %.1f FPS\n", nb, t*1.e-6f, (float)(nb*1.e6f/t) ); |
printf( "%d frames decoded in %.3f s -> %.1f FPS Checksum:0x%.8x\n", nb, t*1.e-6f, (float)(nb*1.e6f/t), chksum ); |
1486 |
if (with_chksum) |
} |
1487 |
printf("checksum: 0x%.8x\n", chksum); |
else { |
1488 |
|
printf("FPS:%.1f Checksum: 0x%.8x Expected:0x%.8x | %s\n", |
1489 |
|
t>0. ? (float)(nb*1.e6f/t) : 0.f, chksum, ref_chksum, (chksum==ref_chksum) ? "OK" : "ERROR"); |
1490 |
|
} |
1491 |
|
|
1492 |
End: |
End: |
1493 |
if (rgb_out!=0) free(rgb_out); |
if (yuv_out!=0) free(yuv_out); |
1494 |
if (buf!=0) free(buf); |
if (buf!=0) free(buf); |
1495 |
if (dechandle!=0) { |
if (dechandle!=0) { |
1496 |
xerr= xvid_decore(dechandle, XVID_DEC_DESTROY, NULL, NULL); |
xerr= xvid_decore(dechandle, XVID_DEC_DESTROY, NULL, NULL); |
1497 |
if (xerr==XVID_ERR_FAIL) |
if (xerr==XVID_ERR_FAIL) |
1498 |
printf("destroy-decoder failed (err=%d)!\n", xerr); |
printf("ERROR: destroy-decoder failed (err=%d)!\n", xerr); |
1499 |
} |
} |
1500 |
if (f!=0) fclose(f); |
if (f!=0) fclose(f); |
1501 |
} |
} |
1507 |
void test_bugs1() |
void test_bugs1() |
1508 |
{ |
{ |
1509 |
CPU *cpu; |
CPU *cpu; |
1510 |
|
uint16_t mpeg_quant_matrices[64*8]; |
1511 |
|
|
1512 |
printf( "\n ===== (de)quant4_intra saturation bug? =====\n" ); |
printf( "\n ===== (de)quant4_intra saturation bug? =====\n" ); |
1513 |
|
|
1520 |
continue; |
continue; |
1521 |
|
|
1522 |
for(i=0; i<64; ++i) Src[i] = i-32; |
for(i=0; i<64; ++i) Src[i] = i-32; |
1523 |
set_intra_matrix( get_default_intra_matrix() ); |
set_intra_matrix( mpeg_quant_matrices, get_default_intra_matrix() ); |
1524 |
dequant_mpeg_intra(Dst, Src, 31, 5); |
dequant_mpeg_intra(Dst, Src, 31, 5, mpeg_quant_matrices); |
1525 |
printf( "dequant_mpeg_intra with CPU=%s: ", cpu->name); |
printf( "dequant_mpeg_intra with CPU=%s: ", cpu->name); |
1526 |
printf( " Out[]= " ); |
printf( " Out[]= " ); |
1527 |
for(i=0; i<64; ++i) printf( "[%d]", Dst[i]); |
for(i=0; i<64; ++i) printf( "[%d]", Dst[i]); |
1539 |
continue; |
continue; |
1540 |
|
|
1541 |
for(i=0; i<64; ++i) Src[i] = i-32; |
for(i=0; i<64; ++i) Src[i] = i-32; |
1542 |
set_inter_matrix( get_default_inter_matrix() ); |
set_inter_matrix( mpeg_quant_matrices, get_default_inter_matrix() ); |
1543 |
dequant_mpeg_inter(Dst, Src, 31); |
dequant_mpeg_inter(Dst, Src, 31, mpeg_quant_matrices); |
1544 |
printf( "dequant_mpeg_inter with CPU=%s: ", cpu->name); |
printf( "dequant_mpeg_inter with CPU=%s: ", cpu->name); |
1545 |
printf( " Out[]= " ); |
printf( " Out[]= " ); |
1546 |
for(i=0; i<64; ++i) printf( "[%d]", Dst[i]); |
for(i=0; i<64; ++i) printf( "[%d]", Dst[i]); |
1551 |
void test_dct_precision_diffs() |
void test_dct_precision_diffs() |
1552 |
{ |
{ |
1553 |
CPU *cpu; |
CPU *cpu; |
1554 |
short Blk[8*8], Blk0[8*8]; |
DECLARE_ALIGNED_MATRIX(Blk, 8, 8, int16_t, 16); |
1555 |
|
DECLARE_ALIGNED_MATRIX(Blk0, 8, 8, int16_t, 16); |
1556 |
|
|
1557 |
printf( "\n ===== fdct/idct precision diffs =====\n" ); |
printf( "\n ===== fdct/idct precision diffs =====\n" ); |
1558 |
|
|
1585 |
const int max_Q = 31; |
const int max_Q = 31; |
1586 |
int i, n, qm, q; |
int i, n, qm, q; |
1587 |
CPU *cpu; |
CPU *cpu; |
1588 |
int16_t Src[8*8], Dst[8*8]; |
DECLARE_ALIGNED_MATRIX(Src, 8, 8, int16_t, 16); |
1589 |
|
DECLARE_ALIGNED_MATRIX(Dst, 8, 8, int16_t, 16); |
1590 |
uint8_t Quant[8*8]; |
uint8_t Quant[8*8]; |
1591 |
CPU cpu_bug_list[] = { { "PLAINC", 0 }, { "MMX ", XVID_CPU_MMX }, {0,0} }; |
CPU cpu_bug_list[] = { { "PLAINC", 0 }, { "MMX ", XVID_CPU_MMX }, {0,0} }; |
1592 |
uint16_t Crcs_Inter[2][32]; |
uint16_t Crcs_Inter[2][32]; |
1593 |
uint16_t Crcs_Intra[2][32]; |
uint16_t Crcs_Intra[2][32]; |
1594 |
|
DECLARE_ALIGNED_MATRIX(mpeg_quant_matrices, 8, 64, uint16_t, 16); |
1595 |
|
|
1596 |
printf( "\n ===== test MPEG4-quantize bug =====\n" ); |
printf( "\n ===== test MPEG4-quantize bug =====\n" ); |
1597 |
|
|
1598 |
for(i=0; i<64; ++i) Src[i] = 2048*(i-32)/32; |
for(i=0; i<64; ++i) Src[i] = 2048*(i-32)/32; |
1601 |
for(qm=1; qm<=255; ++qm) |
for(qm=1; qm<=255; ++qm) |
1602 |
{ |
{ |
1603 |
for(i=0; i<8*8; ++i) Quant[i] = qm; |
for(i=0; i<8*8; ++i) Quant[i] = qm; |
1604 |
set_inter_matrix( Quant ); |
set_inter_matrix( mpeg_quant_matrices, Quant ); |
1605 |
|
|
1606 |
for(n=0, cpu = cpu_bug_list; cpu->name!=0; ++cpu, ++n) |
for(n=0, cpu = cpu_bug_list; cpu->name!=0; ++cpu, ++n) |
1607 |
{ |
{ |
1612 |
|
|
1613 |
for(q=1; q<=max_Q; ++q) { |
for(q=1; q<=max_Q; ++q) { |
1614 |
emms(); |
emms(); |
1615 |
quant_mpeg_inter( Dst, Src, q ); |
quant_mpeg_inter( Dst, Src, q, mpeg_quant_matrices ); |
1616 |
emms(); |
emms(); |
1617 |
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; |
1618 |
Crcs_Inter[n][q] = s; |
Crcs_Inter[n][q] = s; |
1631 |
for(qm=1; qm<=255; ++qm) |
for(qm=1; qm<=255; ++qm) |
1632 |
{ |
{ |
1633 |
for(i=0; i<8*8; ++i) Quant[i] = qm; |
for(i=0; i<8*8; ++i) Quant[i] = qm; |
1634 |
set_intra_matrix( Quant ); |
set_intra_matrix( mpeg_quant_matrices, Quant ); |
1635 |
|
|
1636 |
for(n=0, cpu = cpu_bug_list; cpu->name!=0; ++cpu, ++n) |
for(n=0, cpu = cpu_bug_list; cpu->name!=0; ++cpu, ++n) |
1637 |
{ |
{ |
1642 |
|
|
1643 |
for(q=1; q<=max_Q; ++q) { |
for(q=1; q<=max_Q; ++q) { |
1644 |
emms(); |
emms(); |
1645 |
quant_mpeg_intra( Dst, Src, q, q); |
quant_mpeg_intra( Dst, Src, q, q, mpeg_quant_matrices); |
1646 |
emms(); |
emms(); |
1647 |
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; |
1648 |
Crcs_Intra[n][q] = s; |
Crcs_Intra[n][q] = s; |
1657 |
} |
} |
1658 |
#endif |
#endif |
1659 |
} |
} |
1660 |
|
/*********************************************************************/ |
1661 |
|
|
1662 |
|
static uint32_t __inline log2bin_v1(uint32_t value) |
1663 |
|
{ |
1664 |
|
int n = 0; |
1665 |
|
while (value) { |
1666 |
|
value >>= 1; |
1667 |
|
n++; |
1668 |
|
} |
1669 |
|
return n; |
1670 |
|
} |
1671 |
|
|
1672 |
|
static const uint8_t log2_tab_16[16] = { 0, 1, 2, 2, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4 }; |
1673 |
|
|
1674 |
|
static uint32_t __inline log2bin_v2(uint32_t value) |
1675 |
|
{ |
1676 |
|
int n = 0; |
1677 |
|
if (value & 0xffff0000) { |
1678 |
|
value >>= 16; |
1679 |
|
n += 16; |
1680 |
|
} |
1681 |
|
if (value & 0xff00) { |
1682 |
|
value >>= 8; |
1683 |
|
n += 8; |
1684 |
|
} |
1685 |
|
if (value & 0xf0) { |
1686 |
|
value >>= 4; |
1687 |
|
n += 4; |
1688 |
|
} |
1689 |
|
return n + log2_tab_16[value]; |
1690 |
|
} |
1691 |
|
|
1692 |
|
void test_log2bin() |
1693 |
|
{ |
1694 |
|
const int nb_tests = 3000*speed_ref; |
1695 |
|
int n, crc1=0, crc2=0; |
1696 |
|
uint32_t s, s0; |
1697 |
|
double t1, t2; |
1698 |
|
|
1699 |
|
t1 = gettime_usec(); |
1700 |
|
s0 = (int)(t1*31.241); |
1701 |
|
for(s=s0, n=0; n<nb_tests; ++n, s=(s*12363+31)&0x7fffffff) |
1702 |
|
crc1 += log2bin_v1(s); |
1703 |
|
t1 = (gettime_usec()-t1) / nb_tests; |
1704 |
|
|
1705 |
|
t2 = gettime_usec(); |
1706 |
|
for(s=s0, n=0; n<nb_tests; ++n, s=(s*12363+31)&0x7fffffff) |
1707 |
|
crc2 += log2bin_v2(s); |
1708 |
|
t2 = (gettime_usec() - t2) / nb_tests; |
1709 |
|
|
1710 |
|
printf( "log2bin_v1: %.3f sec crc=%d\n", t1, crc1 ); |
1711 |
|
printf( "log2bin_v2: %.3f sec crc=%d\n", t2, crc2 ); |
1712 |
|
if (crc1!=crc2) printf( " CRC ERROR !\n" ); |
1713 |
|
} |
1714 |
|
|
1715 |
|
/*********************************************************************/ |
1716 |
|
|
1717 |
|
static void __inline old_gcd(int *num, int *den) |
1718 |
|
{ |
1719 |
|
int i = *num; |
1720 |
|
while (i > 1) { |
1721 |
|
if (*num % i == 0 && *den % i == 0) { |
1722 |
|
*num /= i; |
1723 |
|
*den /= i; |
1724 |
|
i = *num; |
1725 |
|
continue; |
1726 |
|
} |
1727 |
|
i--; |
1728 |
|
} |
1729 |
|
} |
1730 |
|
|
1731 |
|
static uint32_t gcd(int num, int den) |
1732 |
|
{ |
1733 |
|
int tmp; |
1734 |
|
while( (tmp=num%den) ) { num = den; den = tmp; } |
1735 |
|
return den; |
1736 |
|
} |
1737 |
|
static void __inline new_gcd(int *num, int *den) |
1738 |
|
{ |
1739 |
|
const int div = gcd(*num, *den); |
1740 |
|
if (num) { |
1741 |
|
*num /= div; |
1742 |
|
*den /= div; |
1743 |
|
} |
1744 |
|
} |
1745 |
|
|
1746 |
|
void test_gcd() |
1747 |
|
{ |
1748 |
|
const int nb_tests = 10*speed_ref; |
1749 |
|
int i; |
1750 |
|
uint32_t crc1=0, crc2=0; |
1751 |
|
uint32_t n0, n, d0, d; |
1752 |
|
double t1, t2; |
1753 |
|
|
1754 |
|
t1 = gettime_usec(); |
1755 |
|
n0 = 0xfffff & (int)(t1*31.241); |
1756 |
|
d0 = 0xfffff & (int)( ((n0*4123)%17) | 1 ); |
1757 |
|
for(n=n0, d=d0, i=0; i<nb_tests; ++i) { |
1758 |
|
old_gcd(&n, &d); |
1759 |
|
crc1 = (((crc1>>4)^d) + ((crc1<<2)^n) ) & 0xffffff; |
1760 |
|
n = d; |
1761 |
|
d = (d*12363+31) & 0xffff; |
1762 |
|
d |= !d; |
1763 |
|
} |
1764 |
|
t1 = (gettime_usec()-t1) / nb_tests; |
1765 |
|
|
1766 |
|
t2 = gettime_usec(); |
1767 |
|
for(n=n0, d=d0, i=0; i<nb_tests; ++i) { |
1768 |
|
new_gcd(&n, &d); |
1769 |
|
crc2 = (((crc2>>4)^d) + ((crc2<<2)^n) ) & 0xffffff; |
1770 |
|
n = d; |
1771 |
|
d = (d*12363+31) & 0xffff; |
1772 |
|
d |= !d; |
1773 |
|
} |
1774 |
|
t2 = (gettime_usec() - t2) / nb_tests; |
1775 |
|
|
1776 |
|
printf( "old_gcd: %.3f sec crc=%d\n", t1, crc1 ); |
1777 |
|
printf( "new_gcd: %.3f sec crc=%d\n", t2, crc2 ); |
1778 |
|
if (crc1!=crc2) printf( " CRC ERROR !\n" ); |
1779 |
|
} |
1780 |
|
|
1781 |
/********************************************************************* |
/********************************************************************* |
1782 |
* main |
* main |
1783 |
*********************************************************************/ |
*********************************************************************/ |
1784 |
|
|
1785 |
int main(int argc, char *argv[]) |
static void arg_missing(const char *opt) |
1786 |
|
{ |
1787 |
|
printf( "missing argument after option '%s'\n", opt); |
1788 |
|
exit(-1); |
1789 |
|
} |
1790 |
|
|
1791 |
|
int main(int argc, const char *argv[]) |
1792 |
{ |
{ |
1793 |
int what = 0; |
int c, what = 0; |
1794 |
if (argc>1) what = atoi(argv[1]); |
int width, height; |
1795 |
|
uint32_t chksum = 0; |
1796 |
|
const char * test_bitstream = 0; |
1797 |
|
|
1798 |
|
cpu_mask = 0; // default => will use autodectect |
1799 |
|
for(c=1; c<argc; ++c) |
1800 |
|
{ |
1801 |
|
if (!strcmp(argv[c], "-v")) verbose++; |
1802 |
|
else if (!strcmp(argv[c], "-c")) cpu_mask = 0 /* PLAIN_C */ | XVID_CPU_FORCE; |
1803 |
|
else if (!strcmp(argv[c], "-mmx")) cpu_mask = XVID_CPU_MMX | XVID_CPU_FORCE; |
1804 |
|
else if (!strcmp(argv[c], "-mmxext")) cpu_mask = XVID_CPU_MMXEXT | XVID_CPU_MMX | XVID_CPU_FORCE; |
1805 |
|
else if (!strcmp(argv[c], "-sse2")) cpu_mask = XVID_CPU_SSE2 | XVID_CPU_MMXEXT | XVID_CPU_MMX | XVID_CPU_FORCE; |
1806 |
|
else if (!strcmp(argv[c], "-3dnow")) cpu_mask = XVID_CPU_3DNOW | XVID_CPU_FORCE; |
1807 |
|
else if (!strcmp(argv[c], "-3dnowe")) cpu_mask = XVID_CPU_3DNOW | XVID_CPU_3DNOWEXT | XVID_CPU_FORCE; |
1808 |
|
else if (!strcmp(argv[c], "-altivec")) cpu_mask = XVID_CPU_ALTIVEC | XVID_CPU_FORCE; |
1809 |
|
else if (!strcmp(argv[c], "-spd")) { |
1810 |
|
if (++c==argc) arg_missing( argv[argc-1] ); |
1811 |
|
speed_ref = atoi(argv[c]); |
1812 |
|
} |
1813 |
|
else if (argv[c][0]!='-') { |
1814 |
|
what = atoi(argv[c]); |
1815 |
|
if (what==9) { |
1816 |
|
if (c+4>argc) { |
1817 |
|
printf("usage: %s %d bitstream width height (checksum)\n", argv[0], what); |
1818 |
|
exit(-1); |
1819 |
|
} |
1820 |
|
test_bitstream = argv[++c]; |
1821 |
|
width = atoi(argv[++c]); |
1822 |
|
height = atoi(argv[++c]); |
1823 |
|
if (c+1<argc && argv[c+1][0]!='-') { |
1824 |
|
if (sscanf(argv[c+1], "0x%x", &chksum)!=1) { |
1825 |
|
printf( "can't read checksum value.\n" ); |
1826 |
|
exit(-1); |
1827 |
|
} |
1828 |
|
else c++; |
1829 |
|
} |
1830 |
|
// printf( "[%s] %dx%d (0x%.8x)\n", test_bitstream, width, height, chksum); |
1831 |
|
} |
1832 |
|
} |
1833 |
|
else { |
1834 |
|
printf( "unrecognized option '%s'\n", argv[c]); |
1835 |
|
exit(-1); |
1836 |
|
} |
1837 |
|
} |
1838 |
|
|
1839 |
|
|
1840 |
if (what==0 || what==1) test_dct(); |
if (what==0 || what==1) test_dct(); |
1841 |
if (what==0 || what==2) test_mb(); |
if (what==0 || what==2) test_mb(); |
1842 |
if (what==0 || what==3) test_sad(); |
if (what==0 || what==3) test_sad(); |
1843 |
if (what==0 || what==4) test_transfer(); |
if (what==0 || what==4) test_transfer(); |
1844 |
if (what==0 || what==5) test_quant(); |
if (what==0 || what==5) test_quant(); |
1845 |
if (what==0 || what==6) test_cbp(); |
if (what==0 || what==6) test_cbp(); |
1846 |
|
if (what==0 || what==10) test_sse(); |
1847 |
|
if (what==0 || what==11) test_log2bin(); |
1848 |
|
if (what==0 || what==12) test_gcd(); |
1849 |
|
|
1850 |
|
|
1851 |
if (what==7) { |
if (what==7) { |
1852 |
test_IEEE1180_compliance(-256, 255, 1); |
test_IEEE1180_compliance(-256, 255, 1); |
1858 |
} |
} |
1859 |
if (what==8) test_dct_saturation(-256, 255); |
if (what==8) test_dct_saturation(-256, 255); |
1860 |
|
|
1861 |
if (what==9) { |
if (test_bitstream) |
1862 |
int width, height; |
test_dec(test_bitstream, width, height, chksum); |
|
if (argc<5) { |
|
|
printf("usage: %s %d [bitstream] [width] [height]\n", argv[0], what); |
|
|
return 1; |
|
|
} |
|
|
width = atoi(argv[3]); |
|
|
height = atoi(argv[4]); |
|
|
test_dec(argv[2], width, height, (argc>5)); |
|
|
} |
|
|
|
|
1863 |
if (what==-1) { |
if (what==-1) { |
1864 |
test_dct_precision_diffs(); |
test_dct_precision_diffs(); |
1865 |
test_bugs1(); |
test_bugs1(); |
1867 |
if (what==-2) |
if (what==-2) |
1868 |
test_quant_bug(); |
test_quant_bug(); |
1869 |
|
|
1870 |
return 0; |
if ((what >= 0 && what <= 6) || what == 10) { |
1871 |
|
printf("\n\n" |
1872 |
|
"NB: If a function isn't optimised for a specific set of intructions,\n" |
1873 |
|
" a C function is used instead. So don't panic if some functions\n" |
1874 |
|
" may appear to be slow.\n"); |
1875 |
} |
} |
1876 |
|
|
1877 |
/********************************************************************* |
#ifdef ARCH_IS_IA32 |
1878 |
* 'Reference' output (except for timing) on an Athlon XP 2200+ |
if (what == 0 || what == 5) { |
1879 |
*********************************************************************/ |
printf("\n" |
1880 |
|
"NB: MMX mpeg4 quantization is known to have very small errors (+/-1 magnitude)\n" |
1881 |
/* as of 07/01/2002, there's a problem with MMX mpeg4-quantization */ |
" for 1 or 2 coefficients a block. This is mainly caused by the fact the unit\n" |
1882 |
/* |
" test goes far behind the usual limits of real encoding. Please do not report\n" |
1883 |
|
" this error to the developers.\n"); |
1884 |
|
} |
1885 |
|
#endif |
1886 |
|
|
1887 |
===== test fdct/idct ===== |
return 0; |
1888 |
PLAINC - 1.446 usec PSNR=13.291 MSE=3.000 |
} |
|
MMX - -0.260 usec PSNR=9.611 MSE=7.000 |
|
|
MMXEXT - -0.293 usec PSNR=9.611 MSE=7.000 |
|
|
3DNOW - 1.535 usec PSNR=13.291 MSE=3.000 |
|
|
3DNOWE - 0.376 usec PSNR=13.291 MSE=3.000 |
|
|
|
|
|
=== test block motion === |
|
|
PLAINC - interp- h-round0 0.126 usec iCrc=8107 |
|
|
PLAINC - round1 0.136 usec iCrc=8100 |
|
|
PLAINC - interp- v-round0 0.121 usec iCrc=8108 |
|
|
PLAINC - round1 0.127 usec iCrc=8105 |
|
|
PLAINC - interp-hv-round0 0.192 usec iCrc=8112 |
|
|
PLAINC - round1 0.213 usec iCrc=8103 |
|
|
--- |
|
|
MMX - interp- h-round0 0.048 usec iCrc=8107 |
|
|
MMX - round1 0.048 usec iCrc=8100 |
|
|
MMX - interp- v-round0 0.046 usec iCrc=8108 |
|
|
MMX - round1 0.047 usec iCrc=8105 |
|
|
MMX - interp-hv-round0 0.074 usec iCrc=8112 |
|
|
MMX - round1 0.074 usec iCrc=8103 |
|
|
--- |
|
|
MMXEXT - interp- h-round0 0.019 usec iCrc=8107 |
|
|
MMXEXT - round1 0.025 usec iCrc=8100 |
|
|
MMXEXT - interp- v-round0 0.015 usec iCrc=8108 |
|
|
MMXEXT - round1 0.024 usec iCrc=8105 |
|
|
MMXEXT - interp-hv-round0 0.039 usec iCrc=8112 |
|
|
MMXEXT - round1 0.037 usec iCrc=8103 |
|
|
--- |
|
|
3DNOW - interp- h-round0 0.019 usec iCrc=8107 |
|
|
3DNOW - round1 0.023 usec iCrc=8100 |
|
|
3DNOW - interp- v-round0 0.015 usec iCrc=8108 |
|
|
3DNOW - round1 0.024 usec iCrc=8105 |
|
|
3DNOW - interp-hv-round0 0.037 usec iCrc=8112 |
|
|
3DNOW - round1 0.038 usec iCrc=8103 |
|
|
--- |
|
|
3DNOWE - interp- h-round0 0.022 usec iCrc=8107 |
|
|
3DNOWE - round1 0.023 usec iCrc=8100 |
|
|
3DNOWE - interp- v-round0 0.016 usec iCrc=8108 |
|
|
3DNOWE - round1 0.021 usec iCrc=8105 |
|
|
3DNOWE - interp-hv-round0 0.036 usec iCrc=8112 |
|
|
3DNOWE - round1 0.036 usec iCrc=8103 |
|
|
--- |
|
|
|
|
|
====== test SAD ====== |
|
|
PLAINC - sad8 0.165 usec sad=3776 |
|
|
PLAINC - sad16 0.587 usec sad=27214 |
|
|
PLAINC - sad16bi 1.290 usec sad=26274 |
|
|
PLAINC - dev16 1.535 usec sad=3344 |
|
|
--- |
|
|
MMX - sad8 0.036 usec sad=3776 |
|
|
MMX - sad16 0.113 usec sad=27214 |
|
|
MMX - sad16bi 0.250 usec sad=26274 |
|
|
MMX - dev16 0.187 usec sad=3344 |
|
|
--- |
|
|
MMXEXT - sad8 0.015 usec sad=3776 |
|
|
MMXEXT - sad16 0.046 usec sad=27214 |
|
|
MMXEXT - sad16bi 0.059 usec sad=26274 |
|
|
MMXEXT - dev16 0.088 usec sad=3344 |
|
|
--- |
|
|
3DNOW - sad8 0.165 usec sad=3776 |
|
|
3DNOW - sad16 0.589 usec sad=27214 |
|
|
3DNOW - sad16bi 0.119 usec sad=26274 |
|
|
3DNOW - dev16 1.541 usec sad=3344 |
|
|
--- |
|
|
3DNOWE - sad8 0.018 usec sad=3776 |
|
|
3DNOWE - sad16 0.039 usec sad=27214 |
|
|
3DNOWE - sad16bi 0.051 usec sad=26274 |
|
|
3DNOWE - dev16 0.070 usec sad=3344 |
|
|
--- |
|
|
|
|
|
=== test transfer === |
|
|
PLAINC - 8to16 0.207 usec crc=28288 |
|
|
PLAINC - 16to8 0.357 usec crc=28288 |
|
|
PLAINC - 8to8 0.154 usec crc=20352 |
|
|
PLAINC - 16to8add 0.391 usec crc=25536 |
|
|
PLAINC - 8to16sub 0.562 usec crc1=28064 crc2=16256 |
|
|
PLAINC - 8to16sub2 0.519 usec crc=22368 |
|
|
--- |
|
|
MMX - 8to16 0.048 usec crc=28288 |
|
|
MMX - 16to8 0.205 usec crc=28288 |
|
|
MMX - 8to8 -0.158 usec crc=20352 |
|
|
MMX - 16to8add 0.015 usec crc=25536 |
|
|
MMX - 8to16sub 0.203 usec crc1=28064 crc2=16256 |
|
|
MMX - 8to16sub2 0.087 usec crc=22368 |
|
|
--- |
|
|
MMXEXT - 8to16 0.013 usec crc=28288 |
|
|
MMXEXT - 16to8 0.011 usec crc=28288 |
|
|
MMXEXT - 8to8 -0.023 usec crc=20352 |
|
|
MMXEXT - 16to8add 0.023 usec crc=25536 |
|
|
MMXEXT - 8to16sub 0.072 usec crc1=28064 crc2=16256 |
|
|
MMXEXT - 8to16sub2 0.093 usec crc=22368 |
|
|
--- |
|
|
3DNOW - 8to16 0.192 usec crc=28288 |
|
|
3DNOW - 16to8 0.367 usec crc=28288 |
|
|
3DNOW - 8to8 0.132 usec crc=20352 |
|
|
3DNOW - 16to8add 0.440 usec crc=25536 |
|
|
3DNOW - 8to16sub 0.557 usec crc1=28064 crc2=16256 |
|
|
3DNOW - 8to16sub2 0.691 usec crc=22368 |
|
|
--- |
|
|
3DNOWE - 8to16 0.004 usec crc=28288 |
|
|
3DNOWE - 16to8 0.019 usec crc=28288 |
|
|
3DNOWE - 8to8 -0.294 usec crc=20352 |
|
|
3DNOWE - 16to8add 0.028 usec crc=25536 |
|
|
3DNOWE - 8to16sub 0.065 usec crc1=28064 crc2=16256 |
|
|
3DNOWE - 8to16sub2 0.027 usec crc=22368 |
|
|
--- |
|
|
|
|
|
===== test quant ===== |
|
|
PLAINC - quant_mpeg_intra 67.757 usec crc=29809 |
|
|
PLAINC - quant_mpeg_inter 68.482 usec crc=12574 |
|
|
PLAINC - dequant_mpeg_intra 20.764 usec crc=24052 |
|
|
PLAINC - dequant_mpeg_inter 24.413 usec crc=63847 |
|
|
PLAINC - quant_h263_intra 16.446 usec crc=25662 |
|
|
PLAINC - quant_h263_inter 14.499 usec crc=23972 |
|
|
PLAINC - dequant_h263_intra 16.419 usec crc=49900 |
|
|
PLAINC - dequant_h263_inter 18.679 usec crc=48899 |
|
|
--- |
|
|
MMX - quant_mpeg_intra 8.299 usec crc=3459 |
|
|
*** CRC ERROR! *** |
|
|
MMX - quant_mpeg_inter 7.078 usec crc=13247 |
|
|
*** CRC ERROR! *** |
|
|
MMX - dequant_mpeg_intra 3.455 usec crc=24052 |
|
|
MMX - dequant_mpeg_inter 4.483 usec crc=63847 |
|
|
MMX - quant_h263_intra 2.597 usec crc=25662 |
|
|
MMX - quant_h263_inter 2.151 usec crc=23972 |
|
|
MMX - dequant_h263_intra 2.684 usec crc=49900 |
|
|
MMX - dequant_h263_inter 2.647 usec crc=48899 |
|
|
--- |
|
|
MMXEXT - quant_mpeg_intra 3.878 usec crc=29809 |
|
|
MMXEXT - quant_mpeg_inter 4.112 usec crc=12574 |
|
|
MMXEXT - dequant_mpeg_intra 3.452 usec crc=24052 |
|
|
MMXEXT - dequant_mpeg_inter 4.473 usec crc=63847 |
|
|
MMXEXT - quant_h263_intra 2.608 usec crc=25662 |
|
|
MMXEXT - quant_h263_inter 2.145 usec crc=23972 |
|
|
MMXEXT - dequant_h263_intra 2.478 usec crc=49900 |
|
|
MMXEXT - dequant_h263_inter 2.450 usec crc=48899 |
|
|
--- |
|
|
3DNOW - quant_mpeg_intra 66.051 usec crc=29809 |
|
|
3DNOW - quant_mpeg_inter 73.494 usec crc=12574 |
|
|
3DNOW - dequant_mpeg_intra 20.374 usec crc=24052 |
|
|
3DNOW - dequant_mpeg_inter 23.645 usec crc=63847 |
|
|
3DNOW - quant_h263_intra 16.292 usec crc=25662 |
|
|
3DNOW - quant_h263_inter 14.322 usec crc=23972 |
|
|
3DNOW - dequant_h263_intra 16.613 usec crc=49900 |
|
|
3DNOW - dequant_h263_inter 18.382 usec crc=48899 |
|
|
--- |
|
|
3DNOWE - quant_mpeg_intra 66.140 usec crc=29809 |
|
|
3DNOWE - quant_mpeg_inter 68.454 usec crc=12574 |
|
|
3DNOWE - dequant_mpeg_intra 2.881 usec crc=24052 |
|
|
3DNOWE - dequant_mpeg_inter 4.155 usec crc=63847 |
|
|
3DNOWE - quant_h263_intra 1.451 usec crc=25662 |
|
|
3DNOWE - quant_h263_inter 1.849 usec crc=23972 |
|
|
3DNOWE - dequant_h263_intra 2.101 usec crc=49900 |
|
|
3DNOWE - dequant_h263_inter 2.109 usec crc=48899 |
|
|
--- |
|
|
|
|
|
===== test cbp ===== |
|
|
PLAINC - calc_cbp#1 0.090 usec cbp=0x15 |
|
|
PLAINC - calc_cbp#2 0.086 usec cbp=0x38 |
|
|
PLAINC - calc_cbp#3 0.087 usec cbp=0xf |
|
|
PLAINC - calc_cbp#4 0.114 usec cbp=0x5 |
|
|
--- |
|
|
MMX - calc_cbp#1 0.061 usec cbp=0x15 |
|
|
MMX - calc_cbp#2 0.063 usec cbp=0x38 |
|
|
MMX - calc_cbp#3 0.061 usec cbp=0xf |
|
|
MMX - calc_cbp#4 0.060 usec cbp=0x5 |
|
|
--- |
|
|
MMXEXT - calc_cbp#1 0.062 usec cbp=0x15 |
|
|
MMXEXT - calc_cbp#2 0.060 usec cbp=0x38 |
|
|
MMXEXT - calc_cbp#3 0.062 usec cbp=0xf |
|
|
MMXEXT - calc_cbp#4 0.061 usec cbp=0x5 |
|
|
--- |
|
|
3DNOW - calc_cbp#1 0.089 usec cbp=0x15 |
|
|
3DNOW - calc_cbp#2 0.087 usec cbp=0x38 |
|
|
3DNOW - calc_cbp#3 0.087 usec cbp=0xf |
|
|
3DNOW - calc_cbp#4 0.116 usec cbp=0x5 |
|
|
--- |
|
|
3DNOWE - calc_cbp#1 0.050 usec cbp=0x15 |
|
|
3DNOWE - calc_cbp#2 0.051 usec cbp=0x38 |
|
|
3DNOWE - calc_cbp#3 0.050 usec cbp=0xf |
|
|
3DNOWE - calc_cbp#4 0.049 usec cbp=0x5 |
|
|
--- |
|
1889 |
|
|
1890 |
*/ |
/*********************************************************************/ |