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 1.13 2004-04-05 20:36:36 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 |
30 |
* bullet-proof (should plug some .md5 here)... |
* bullet-proof (should plug some .md5 here)... |
31 |
* |
* |
32 |
* compiles best at xvidcore/src-dir with something like |
* compiles with something like: |
33 |
* |
* gcc -o xvid_bench xvid_bench.c -I../src/ -lxvidcore -lm |
|
* gcc -DARCH_IS_IA32 -DARCH_IS_32BIT -o xvid_bench xvid_bench.c \ |
|
|
* ../build/generic/libxvidcore.a -lm |
|
|
* |
|
|
* History: |
|
34 |
* |
* |
35 |
* 06.06.2002 initial coding -Skal- |
****************************************************************************/ |
|
* 27.02.2003 minor changes (compile, sad16v) <gruel@web.de> |
|
|
* |
|
|
*************************************************************************/ |
|
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_3DNOW | XVID_CPU_3DNOWEXT } |
123 |
, { "IA64 ", XVID_CPU_IA64 } |
#endif |
124 |
|
#ifdef ARCH_IS_PPC |
125 |
|
, { "ALTIVEC", XVID_CPU_ALTIVEC } |
126 |
|
#endif |
127 |
|
//, { "IA64 ", XVID_CPU_IA64 } |
128 |
//, { "TSC ", XVID_CPU_TSC } |
//, { "TSC ", XVID_CPU_TSC } |
129 |
, { 0, 0 } } |
, { 0, 0 } }; |
130 |
|
|
131 |
, cpu_short_list[] = |
CPU cpu_short_list[] = |
132 |
{ { "PLAINC", 0 } |
{ { "PLAINC", 0 } |
133 |
|
#ifdef ARCH_IS_IA32 |
134 |
, { "MMX ", XVID_CPU_MMX } |
, { "MMX ", XVID_CPU_MMX } |
135 |
//, { "MMXEXT", XVID_CPU_MMXEXT | XVID_CPU_MMX } |
//, { "MMXEXT", XVID_CPU_MMXEXT | XVID_CPU_MMX } |
136 |
, { "IA64 ", XVID_CPU_IA64 } |
#endif |
137 |
, { 0, 0 } } |
//, { "IA64 ", XVID_CPU_IA64 } |
138 |
|
, { 0, 0 } }; |
139 |
|
|
140 |
, cpu_short_list2[] = |
CPU cpu_short_list2[] = |
141 |
{ { "PLAINC", 0 } |
{ { "PLAINC", 0 } |
142 |
|
#ifdef ARCH_IS_IA32 |
143 |
, { "MMX ", XVID_CPU_MMX } |
, { "MMX ", XVID_CPU_MMX } |
144 |
, { "SSE2 ", XVID_CPU_SSE2 | XVID_CPU_MMX } |
, { "SSE2 ", XVID_CPU_SSE2 | XVID_CPU_MMX } |
145 |
|
#endif |
146 |
, { 0, 0 } }; |
, { 0, 0 } }; |
147 |
|
|
148 |
|
|
149 |
int init_cpu(CPU *cpu) |
int init_cpu(CPU *cpu) |
150 |
{ |
{ |
151 |
int xerr, cpu_type; |
xvid_gbl_info_t xinfo; |
152 |
XVID_INIT_PARAM xinit; |
|
153 |
|
/* Get the available CPU flags */ |
154 |
|
memset(&xinfo, 0, sizeof(xinfo)); |
155 |
|
xinfo.version = XVID_VERSION; |
156 |
|
xvid_global(NULL, XVID_GBL_INFO, &xinfo, NULL); |
157 |
|
|
158 |
cpu_type = check_cpu_features() & cpu->cpu; |
/* Are we trying to test a subset of the host CPU features */ |
159 |
xinit.cpu_flags = cpu_type | XVID_CPU_FORCE; |
if ((xinfo.cpu_flags & cpu->cpu) == cpu->cpu) { |
160 |
// xinit.cpu_flags = XVID_CPU_MMX | XVID_CPU_FORCE; |
int xerr; |
161 |
xerr = xvid_init(NULL, 0, &xinit, NULL); |
xvid_gbl_init_t xinit; |
162 |
if (cpu->cpu>0 && (cpu_type==0 || xerr!=XVID_ERR_OK)) { |
memset(&xinit, 0, sizeof(xinit)); |
163 |
printf( "%s - skipped...\n", cpu->name ); |
xinit.cpu_flags = cpu->cpu | XVID_CPU_FORCE; |
164 |
|
xinit.version = XVID_VERSION; |
165 |
|
xerr = xvid_global(NULL, XVID_GBL_INIT, &xinit, NULL); |
166 |
|
if (xerr==XVID_ERR_FAIL) { |
167 |
|
/* libxvidcore failed to init */ |
168 |
return 0; |
return 0; |
169 |
} |
} |
170 |
|
} else { |
171 |
|
/* The host CPU doesn't support some required feature for this test */ |
172 |
|
return(0); |
173 |
|
} |
174 |
return 1; |
return 1; |
175 |
} |
} |
176 |
|
|
177 |
|
#define CRC32_REMAINDER 0xCBF43926 |
178 |
|
#define CRC32_INITIAL 0xffffffff |
179 |
|
|
180 |
|
#define DO1(c, crc) ((crc) = crc32tab[((unsigned int)((crc)>>24) ^ (*c++)) & 0xff] ^ ((crc) << 8)) |
181 |
|
#define DO2(c, crc) DO1(c, crc); DO1(c, crc); |
182 |
|
#define DO4(c, crc) DO2(c, crc); DO2(c, crc); |
183 |
|
#define DO8(c, crc) DO4(c, crc); DO4(c, crc); |
184 |
|
|
185 |
|
/****************************************************************************** |
186 |
|
* Precomputed AAL5 CRC32 lookup table |
187 |
|
******************************************************************************/ |
188 |
|
|
189 |
|
static unsigned long crc32tab[256] = { |
190 |
|
|
191 |
|
0x00000000L, 0x04C11DB7L, 0x09823B6EL, 0x0D4326D9L, |
192 |
|
0x130476DCL, 0x17C56B6BL, 0x1A864DB2L, 0x1E475005L, |
193 |
|
0x2608EDB8L, 0x22C9F00FL, 0x2F8AD6D6L, 0x2B4BCB61L, |
194 |
|
0x350C9B64L, 0x31CD86D3L, 0x3C8EA00AL, 0x384FBDBDL, |
195 |
|
0x4C11DB70L, 0x48D0C6C7L, 0x4593E01EL, 0x4152FDA9L, |
196 |
|
0x5F15ADACL, 0x5BD4B01BL, 0x569796C2L, 0x52568B75L, |
197 |
|
0x6A1936C8L, 0x6ED82B7FL, 0x639B0DA6L, 0x675A1011L, |
198 |
|
0x791D4014L, 0x7DDC5DA3L, 0x709F7B7AL, 0x745E66CDL, |
199 |
|
0x9823B6E0L, 0x9CE2AB57L, 0x91A18D8EL, 0x95609039L, |
200 |
|
0x8B27C03CL, 0x8FE6DD8BL, 0x82A5FB52L, 0x8664E6E5L, |
201 |
|
0xBE2B5B58L, 0xBAEA46EFL, 0xB7A96036L, 0xB3687D81L, |
202 |
|
0xAD2F2D84L, 0xA9EE3033L, 0xA4AD16EAL, 0xA06C0B5DL, |
203 |
|
0xD4326D90L, 0xD0F37027L, 0xDDB056FEL, 0xD9714B49L, |
204 |
|
0xC7361B4CL, 0xC3F706FBL, 0xCEB42022L, 0xCA753D95L, |
205 |
|
0xF23A8028L, 0xF6FB9D9FL, 0xFBB8BB46L, 0xFF79A6F1L, |
206 |
|
0xE13EF6F4L, 0xE5FFEB43L, 0xE8BCCD9AL, 0xEC7DD02DL, |
207 |
|
0x34867077L, 0x30476DC0L, 0x3D044B19L, 0x39C556AEL, |
208 |
|
0x278206ABL, 0x23431B1CL, 0x2E003DC5L, 0x2AC12072L, |
209 |
|
0x128E9DCFL, 0x164F8078L, 0x1B0CA6A1L, 0x1FCDBB16L, |
210 |
|
0x018AEB13L, 0x054BF6A4L, 0x0808D07DL, 0x0CC9CDCAL, |
211 |
|
0x7897AB07L, 0x7C56B6B0L, 0x71159069L, 0x75D48DDEL, |
212 |
|
0x6B93DDDBL, 0x6F52C06CL, 0x6211E6B5L, 0x66D0FB02L, |
213 |
|
0x5E9F46BFL, 0x5A5E5B08L, 0x571D7DD1L, 0x53DC6066L, |
214 |
|
0x4D9B3063L, 0x495A2DD4L, 0x44190B0DL, 0x40D816BAL, |
215 |
|
0xACA5C697L, 0xA864DB20L, 0xA527FDF9L, 0xA1E6E04EL, |
216 |
|
0xBFA1B04BL, 0xBB60ADFCL, 0xB6238B25L, 0xB2E29692L, |
217 |
|
0x8AAD2B2FL, 0x8E6C3698L, 0x832F1041L, 0x87EE0DF6L, |
218 |
|
0x99A95DF3L, 0x9D684044L, 0x902B669DL, 0x94EA7B2AL, |
219 |
|
0xE0B41DE7L, 0xE4750050L, 0xE9362689L, 0xEDF73B3EL, |
220 |
|
0xF3B06B3BL, 0xF771768CL, 0xFA325055L, 0xFEF34DE2L, |
221 |
|
0xC6BCF05FL, 0xC27DEDE8L, 0xCF3ECB31L, 0xCBFFD686L, |
222 |
|
0xD5B88683L, 0xD1799B34L, 0xDC3ABDEDL, 0xD8FBA05AL, |
223 |
|
0x690CE0EEL, 0x6DCDFD59L, 0x608EDB80L, 0x644FC637L, |
224 |
|
0x7A089632L, 0x7EC98B85L, 0x738AAD5CL, 0x774BB0EBL, |
225 |
|
0x4F040D56L, 0x4BC510E1L, 0x46863638L, 0x42472B8FL, |
226 |
|
0x5C007B8AL, 0x58C1663DL, 0x558240E4L, 0x51435D53L, |
227 |
|
0x251D3B9EL, 0x21DC2629L, 0x2C9F00F0L, 0x285E1D47L, |
228 |
|
0x36194D42L, 0x32D850F5L, 0x3F9B762CL, 0x3B5A6B9BL, |
229 |
|
0x0315D626L, 0x07D4CB91L, 0x0A97ED48L, 0x0E56F0FFL, |
230 |
|
0x1011A0FAL, 0x14D0BD4DL, 0x19939B94L, 0x1D528623L, |
231 |
|
0xF12F560EL, 0xF5EE4BB9L, 0xF8AD6D60L, 0xFC6C70D7L, |
232 |
|
0xE22B20D2L, 0xE6EA3D65L, 0xEBA91BBCL, 0xEF68060BL, |
233 |
|
0xD727BBB6L, 0xD3E6A601L, 0xDEA580D8L, 0xDA649D6FL, |
234 |
|
0xC423CD6AL, 0xC0E2D0DDL, 0xCDA1F604L, 0xC960EBB3L, |
235 |
|
0xBD3E8D7EL, 0xB9FF90C9L, 0xB4BCB610L, 0xB07DABA7L, |
236 |
|
0xAE3AFBA2L, 0xAAFBE615L, 0xA7B8C0CCL, 0xA379DD7BL, |
237 |
|
0x9B3660C6L, 0x9FF77D71L, 0x92B45BA8L, 0x9675461FL, |
238 |
|
0x8832161AL, 0x8CF30BADL, 0x81B02D74L, 0x857130C3L, |
239 |
|
0x5D8A9099L, 0x594B8D2EL, 0x5408ABF7L, 0x50C9B640L, |
240 |
|
0x4E8EE645L, 0x4A4FFBF2L, 0x470CDD2BL, 0x43CDC09CL, |
241 |
|
0x7B827D21L, 0x7F436096L, 0x7200464FL, 0x76C15BF8L, |
242 |
|
0x68860BFDL, 0x6C47164AL, 0x61043093L, 0x65C52D24L, |
243 |
|
0x119B4BE9L, 0x155A565EL, 0x18197087L, 0x1CD86D30L, |
244 |
|
0x029F3D35L, 0x065E2082L, 0x0B1D065BL, 0x0FDC1BECL, |
245 |
|
0x3793A651L, 0x3352BBE6L, 0x3E119D3FL, 0x3AD08088L, |
246 |
|
0x2497D08DL, 0x2056CD3AL, 0x2D15EBE3L, 0x29D4F654L, |
247 |
|
0xC5A92679L, 0xC1683BCEL, 0xCC2B1D17L, 0xC8EA00A0L, |
248 |
|
0xD6AD50A5L, 0xD26C4D12L, 0xDF2F6BCBL, 0xDBEE767CL, |
249 |
|
0xE3A1CBC1L, 0xE760D676L, 0xEA23F0AFL, 0xEEE2ED18L, |
250 |
|
0xF0A5BD1DL, 0xF464A0AAL, 0xF9278673L, 0xFDE69BC4L, |
251 |
|
0x89B8FD09L, 0x8D79E0BEL, 0x803AC667L, 0x84FBDBD0L, |
252 |
|
0x9ABC8BD5L, 0x9E7D9662L, 0x933EB0BBL, 0x97FFAD0CL, |
253 |
|
0xAFB010B1L, 0xAB710D06L, 0xA6322BDFL, 0xA2F33668L, |
254 |
|
0xBCB4666DL, 0xB8757BDAL, 0xB5365D03L, 0xB1F740B4L |
255 |
|
|
256 |
|
}; |
257 |
|
|
258 |
|
uint32_t |
259 |
|
calc_crc(uint8_t *mem, int len, uint32_t initial) |
260 |
|
{ |
261 |
|
|
262 |
|
register unsigned int crc; |
263 |
|
|
264 |
|
crc = initial; |
265 |
|
|
266 |
|
while( len >= 8) { |
267 |
|
DO8(mem, crc); |
268 |
|
len -= 8; |
269 |
|
} |
270 |
|
|
271 |
|
while( len ) { |
272 |
|
DO1(mem, crc); |
273 |
|
len--; |
274 |
|
} |
275 |
|
|
276 |
|
return(crc); |
277 |
|
|
278 |
|
} |
279 |
|
|
280 |
/********************************************************************* |
/********************************************************************* |
281 |
* test DCT |
* test DCT |
282 |
*********************************************************************/ |
*********************************************************************/ |
289 |
int tst; |
int tst; |
290 |
CPU *cpu; |
CPU *cpu; |
291 |
int i; |
int i; |
292 |
short iDst0[8*8], iDst[8*8], fDst[8*8]; |
DECLARE_ALIGNED_MATRIX(iDst0, 8, 8, short, 16); |
293 |
|
DECLARE_ALIGNED_MATRIX(iDst, 8, 8, short, 16); |
294 |
|
DECLARE_ALIGNED_MATRIX(fDst, 8, 8, short, 16); |
295 |
double overhead; |
double overhead; |
296 |
|
|
297 |
printf( "\n ===== test fdct/idct =====\n" ); |
printf( "\n ===== test fdct/idct =====\n" ); |
329 |
MSE += delta*delta; |
MSE += delta*delta; |
330 |
} |
} |
331 |
PSNR = (MSE==0.) ? 1.e6 : -4.3429448*log( MSE/64. ); |
PSNR = (MSE==0.) ? 1.e6 : -4.3429448*log( MSE/64. ); |
332 |
printf( "%s - %.3f usec PSNR=%.3f MSE=%.3f\n", |
printf( "%s - %.3f usec PSNR=%.3f MSE=%.3f %s\n", |
333 |
cpu->name, t, PSNR, MSE ); |
cpu->name, t, PSNR, MSE, |
334 |
if (ABS(MSE)>=64) printf( "*** CRC ERROR! ***\n" ); |
(ABS(MSE)>=64)? "| ERROR" :""); |
335 |
} |
} |
336 |
} |
} |
337 |
|
|
342 |
void test_sad() |
void test_sad() |
343 |
{ |
{ |
344 |
const int nb_tests = 2000*speed_ref; |
const int nb_tests = 2000*speed_ref; |
345 |
int tst,dummy[4]; |
int tst; |
346 |
CPU *cpu; |
CPU *cpu; |
347 |
int i; |
int i; |
348 |
uint8_t Cur[16*16], Ref1[16*16], Ref2[16*16]; |
DECLARE_ALIGNED_MATRIX(Cur, 16, 16, uint8_t, 16); |
349 |
|
DECLARE_ALIGNED_MATRIX(Ref1, 16, 16, uint8_t, 16); |
350 |
|
DECLARE_ALIGNED_MATRIX(Ref2, 16, 16, uint8_t, 16); |
351 |
|
|
352 |
printf( "\n ====== test SAD ======\n" ); |
printf( "\n ====== test SAD ======\n" ); |
353 |
for(i=0; i<16*16;++i) { |
for(i=0; i<16*16;++i) { |
368 |
for(tst=0; tst<nb_tests; ++tst) s = sad8(Cur, Ref1, 16); |
for(tst=0; tst<nb_tests; ++tst) s = sad8(Cur, Ref1, 16); |
369 |
emms(); |
emms(); |
370 |
t = (gettime_usec() - t) / nb_tests; |
t = (gettime_usec() - t) / nb_tests; |
371 |
printf( "%s - sad8 %.3f usec sad=%d\n", cpu->name, t, s ); |
printf("%s - sad8 %.3f usec sad=%d %s\n", |
372 |
if (s!=3776) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, s, |
373 |
|
(s!=3776)?"| ERROR": "" ); |
374 |
|
|
375 |
t = gettime_usec(); |
t = gettime_usec(); |
376 |
emms(); |
emms(); |
377 |
for(tst=0; tst<nb_tests; ++tst) s = sad16(Cur, Ref1, 16, 65535); |
for(tst=0; tst<nb_tests; ++tst) s = sad16(Cur, Ref1, 16, -1); |
378 |
emms(); |
emms(); |
379 |
t = (gettime_usec() - t) / nb_tests; |
t = (gettime_usec() - t) / nb_tests; |
380 |
printf( "%s - sad16 %.3f usec sad=%d\n", cpu->name, t, s ); |
printf("%s - sad16 %.3f usec sad=%d %s\n", |
381 |
if (s!=27214) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, s, |
382 |
|
(s!=27214)?"| ERROR": "" ); |
|
t = gettime_usec(); |
|
|
emms(); |
|
|
for(tst=0; tst<nb_tests; ++tst) s = sad16v(Cur, Ref1, 16, dummy); |
|
|
emms(); |
|
|
t = (gettime_usec() - t) / nb_tests; |
|
|
printf( "%s - sad16v %.3f usec sad=%d\n", cpu->name, t, s ); |
|
|
if (s!=27214) printf( "*** CRC ERROR! ***\n" ); |
|
383 |
|
|
384 |
t = gettime_usec(); |
t = gettime_usec(); |
385 |
emms(); |
emms(); |
386 |
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); |
387 |
emms(); |
emms(); |
388 |
t = (gettime_usec() - t) / nb_tests; |
t = (gettime_usec() - t) / nb_tests; |
389 |
printf( "%s - sad16bi %.3f usec sad=%d\n", cpu->name, t, s ); |
printf( "%s - sad16bi %.3f usec sad=%d %s\n", |
390 |
if (s!=26274) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, s, |
391 |
|
(s!=26274)?"| ERROR": "" ); |
392 |
|
|
393 |
t = gettime_usec(); |
t = gettime_usec(); |
394 |
emms(); |
emms(); |
395 |
for(tst=0; tst<nb_tests; ++tst) s = dev16(Cur, 16); |
for(tst=0; tst<nb_tests; ++tst) s = dev16(Cur, 16); |
396 |
emms(); |
emms(); |
397 |
t = (gettime_usec() - t) / nb_tests; |
t = (gettime_usec() - t) / nb_tests; |
398 |
printf( "%s - dev16 %.3f usec sad=%d\n", cpu->name, t, s ); |
printf( "%s - dev16 %.3f usec sad=%d %s\n", |
399 |
if (s!=3344) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, s, |
400 |
|
(s!=3344)?"| ERROR": "" ); |
401 |
|
|
402 |
printf( " --- \n" ); |
printf( " --- \n" ); |
403 |
} |
} |
415 |
#define LEAVE \ |
#define LEAVE \ |
416 |
emms(); \ |
emms(); \ |
417 |
t = (gettime_usec() - t) / nb_tests; \ |
t = (gettime_usec() - t) / nb_tests; \ |
418 |
iCrc = 0; \ |
iCrc = calc_crc((uint8_t*)Dst, sizeof(Dst), CRC32_INITIAL) |
|
for(i=0; i<16*8; ++i) { iCrc += Dst[i]^i; } |
|
419 |
|
|
420 |
#define TEST_MB(FUNC, R) \ |
#define TEST_MB(FUNC, R) \ |
421 |
ENTER \ |
ENTER \ |
433 |
const int nb_tests = 2000*speed_ref; |
const int nb_tests = 2000*speed_ref; |
434 |
CPU *cpu; |
CPU *cpu; |
435 |
const uint8_t Src0[16*9] = { |
const uint8_t Src0[16*9] = { |
436 |
// try to have every possible combinaison of rounding... |
/* try to have every possible combinaison of rounding... */ |
437 |
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, |
438 |
, 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, |
439 |
, 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, |
440 |
, 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, |
441 |
, 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, |
442 |
, 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, |
443 |
, 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, |
444 |
, 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, |
445 |
, 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 |
446 |
}; |
}; |
447 |
uint8_t Dst[16*8] = {0}; |
uint8_t Dst[16*8] = {0}; |
448 |
|
|
457 |
continue; |
continue; |
458 |
|
|
459 |
TEST_MB(interpolate8x8_halfpel_h, 0); |
TEST_MB(interpolate8x8_halfpel_h, 0); |
460 |
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", |
461 |
if (iCrc!=8107) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, iCrc, |
462 |
|
(iCrc!=0x115381ba)?"| ERROR": "" ); |
463 |
|
|
464 |
TEST_MB(interpolate8x8_halfpel_h, 1); |
TEST_MB(interpolate8x8_halfpel_h, 1); |
465 |
printf( "%s - round1 %.3f usec iCrc=%d\n", cpu->name, t, iCrc ); |
printf("%s - round1 %.3f usec crc32=0x%08x %s\n", |
466 |
if (iCrc!=8100) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, iCrc, |
467 |
|
(iCrc!=0x2b1f528f)?"| ERROR": "" ); |
468 |
|
|
469 |
|
|
470 |
TEST_MB(interpolate8x8_halfpel_v, 0); |
TEST_MB(interpolate8x8_halfpel_v, 0); |
471 |
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", |
472 |
if (iCrc!=8108) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, iCrc, |
473 |
|
(iCrc!=0x423cdcc7)?"| ERROR": "" ); |
474 |
|
|
475 |
TEST_MB(interpolate8x8_halfpel_v, 1); |
TEST_MB(interpolate8x8_halfpel_v, 1); |
476 |
printf( "%s - round1 %.3f usec iCrc=%d\n", cpu->name, t, iCrc ); |
printf("%s - round1 %.3f usec crc32=0x%08x %s\n", |
477 |
if (iCrc!=8105) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, iCrc, |
478 |
|
(iCrc!=0x42202efe)?"| ERROR": "" ); |
479 |
|
|
480 |
|
|
481 |
TEST_MB(interpolate8x8_halfpel_hv, 0); |
TEST_MB(interpolate8x8_halfpel_hv, 0); |
482 |
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", |
483 |
if (iCrc!=8112) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, iCrc, |
484 |
|
(iCrc!=0xd198d387)?"| ERROR": "" ); |
485 |
|
|
486 |
TEST_MB(interpolate8x8_halfpel_hv, 1); |
TEST_MB(interpolate8x8_halfpel_hv, 1); |
487 |
printf( "%s - round1 %.3f usec iCrc=%d\n", cpu->name, t, iCrc ); |
printf("%s - round1 %.3f usec crc32=0x%08x %s\n", |
488 |
if (iCrc!=8103) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, iCrc, |
489 |
|
(iCrc!=0x9ecfd921)?"| ERROR": "" ); |
490 |
|
|
491 |
|
|
492 |
// this is a new function, as of 06.06.2002 |
/* this is a new function, as of 06.06.2002 */ |
493 |
#if 0 |
#if 0 |
494 |
TEST_MB2(interpolate8x8_avrg); |
TEST_MB2(interpolate8x8_avrg); |
495 |
printf( "%s - interpolate8x8_c %.3f usec iCrc=%d\n", cpu->name, t, iCrc ); |
printf("%s - interpolate8x8_c %.3f usec crc32=0x%08x %s\n", |
496 |
if (iCrc!=8107) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, iCrc, |
497 |
|
(iCrc!=8107)?"| ERROR": "" ); |
498 |
#endif |
#endif |
499 |
|
|
500 |
printf( " --- \n" ); |
printf( " --- \n" ); |
530 |
} \ |
} \ |
531 |
emms(); \ |
emms(); \ |
532 |
t = (gettime_usec()-t -overhead) / nb_tests;\ |
t = (gettime_usec()-t -overhead) / nb_tests;\ |
533 |
s = 0; for(i=0; i<8*32; ++i) { s += (DST)[i]^i; } |
s = calc_crc((uint8_t*)(DST), sizeof((DST)), CRC32_INITIAL) |
534 |
|
|
535 |
#define TEST_TRANSFER(FUNC, DST, SRC) \ |
#define TEST_TRANSFER(FUNC, DST, SRC) \ |
536 |
TEST_TRANSFER_BEGIN(DST); \ |
TEST_TRANSFER_BEGIN(DST); \ |
556 |
} \ |
} \ |
557 |
emms(); \ |
emms(); \ |
558 |
t = (gettime_usec()-t -overhead) / nb_tests;\ |
t = (gettime_usec()-t -overhead) / nb_tests;\ |
559 |
s = 0; for(i=0; i<8*32; ++i) { s += (DST)[i]; } |
s = calc_crc((uint8_t*)(DST), sizeof((DST)), CRC32_INITIAL) |
560 |
|
|
561 |
#define TEST_TRANSFER2(FUNC, DST, SRC, R1) \ |
#define TEST_TRANSFER2(FUNC, DST, SRC, R1) \ |
562 |
TEST_TRANSFER2_BEGIN(DST,SRC); \ |
TEST_TRANSFER2_BEGIN(DST,SRC); \ |
578 |
|
|
579 |
printf( "\n === test transfer ===\n" ); |
printf( "\n === test transfer ===\n" ); |
580 |
|
|
581 |
for(cpu = cpu_short_list; cpu->name!=0; ++cpu) |
for(cpu = cpu_list; cpu->name!=0; ++cpu) |
582 |
{ |
{ |
583 |
double t, overhead; |
double t, overhead; |
584 |
int tst, s; |
int tst, s; |
587 |
continue; |
continue; |
588 |
|
|
589 |
TEST_TRANSFER(transfer_8to16copy, Dst16, Src8); |
TEST_TRANSFER(transfer_8to16copy, Dst16, Src8); |
590 |
printf( "%s - 8to16 %.3f usec crc=%d\n", cpu->name, t, s ); |
printf("%s - 8to16 %.3f usec crc32=0x%08x %s\n", |
591 |
if (s!=28288) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, s, |
592 |
|
(s!=0x115814bb)?"| ERROR": ""); |
593 |
|
|
594 |
TEST_TRANSFER(transfer_16to8copy, Dst8, Src16); |
TEST_TRANSFER(transfer_16to8copy, Dst8, Src16); |
595 |
printf( "%s - 16to8 %.3f usec crc=%d\n", cpu->name, t, s ); |
printf( "%s - 16to8 %.3f usec crc32=0x%08x %s\n", |
596 |
if (s!=28288) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, s, |
597 |
|
(s!=0xee7ccbb4)?"| ERROR": ""); |
598 |
|
|
599 |
TEST_TRANSFER(transfer8x8_copy, Dst8, Src8); |
TEST_TRANSFER(transfer8x8_copy, Dst8, Src8); |
600 |
printf( "%s - 8to8 %.3f usec crc=%d\n", cpu->name, t, s ); |
printf("%s - 8to8 %.3f usec crc32=0x%08x %s\n", |
601 |
if (s!=20352) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, s, |
602 |
|
(s!=0xd37b3295)?"| ERROR": ""); |
603 |
|
|
604 |
TEST_TRANSFER(transfer_16to8add, Dst8, Src16); |
TEST_TRANSFER(transfer_16to8add, Dst8, Src16); |
605 |
printf( "%s - 16to8add %.3f usec crc=%d\n", cpu->name, t, s ); |
printf("%s - 16to8add %.3f usec crc32=0x%08x %s\n", |
606 |
if (s!=25536) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, s, |
607 |
|
(s!=0xdd817bf4)?"| ERROR": "" ); |
608 |
|
|
609 |
TEST_TRANSFER2(transfer_8to16sub, Dst16, Src8, Ref1); |
TEST_TRANSFER2(transfer_8to16sub, Dst16, Src8, Ref1); |
610 |
printf( "%s - 8to16sub %.3f usec crc1=%d ", cpu->name, t, s ); |
{ |
611 |
if (s!=28064) printf( "*** CRC ERROR! ***\n" ); |
int s1, s2; |
612 |
s = 0; for(i=0; i<8*32; ++i) { s += (Src8[i]-Ref1[i])&i; } |
s1 = calc_crc((uint8_t*)Dst16, sizeof(Dst16), CRC32_INITIAL); |
613 |
printf( "crc2=%d\n", s); |
s2 = calc_crc((uint8_t*)Src8, sizeof(Src8), CRC32_INITIAL); |
614 |
if (s!=16256) printf( "*** CRC ERROR! ***\n" ); |
printf("%s - 8to16sub %.3f usec crc32(1)=0x%08x crc32(2)=0x%08x %s %s\n", |
615 |
#if 1 |
cpu->name, t, s1, s2, |
616 |
|
(s1!=0xa1e07163)?"| ERROR1": "", |
617 |
|
(s2!=0xd86c5d23)?"| ERROR2": "" ); |
618 |
|
} |
619 |
|
|
620 |
TEST_TRANSFER3(transfer_8to16sub2, Dst16, Src8, Ref1, Ref2); |
TEST_TRANSFER3(transfer_8to16sub2, Dst16, Src8, Ref1, Ref2); |
621 |
printf( "%s - 8to16sub2 %.3f usec crc=%d\n", cpu->name, t, s ); |
printf("%s - 8to16sub2 %.3f usec crc32=0x%08x %s\n", |
622 |
if (s!=20384) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, s, |
623 |
// for(i=0; i<64; ++i) printf( "[%d]", Dst16[i]); |
(s!=0x99b6c4c7)?"| ERROR": "" ); |
624 |
// printf("\n"); |
|
|
#endif |
|
625 |
printf( " --- \n" ); |
printf( " --- \n" ); |
626 |
} |
} |
627 |
} |
} |
632 |
|
|
633 |
#define TEST_QUANT(FUNC, DST, SRC) \ |
#define TEST_QUANT(FUNC, DST, SRC) \ |
634 |
t = gettime_usec(); \ |
t = gettime_usec(); \ |
635 |
for(s=0,qm=1; qm<=255; ++qm) { \ |
for(s=CRC32_INITIAL,qm=1; qm<=255; ++qm) { \ |
636 |
for(i=0; i<8*8; ++i) Quant[i] = qm; \ |
for(i=0; i<8*8; ++i) Quant[i] = qm; \ |
637 |
set_inter_matrix( Quant ); \ |
set_inter_matrix( mpeg_quant_matrices, Quant ); \ |
638 |
emms(); \ |
emms(); \ |
639 |
for(q=1; q<=max_Q; ++q) { \ |
for(q=1; q<=max_Q; ++q) { \ |
640 |
for(tst=0; tst<nb_tests; ++tst) \ |
for(tst=0; tst<nb_tests; ++tst) \ |
641 |
(FUNC)((DST), (SRC), q); \ |
(FUNC)((DST), (SRC), q, mpeg_quant_matrices); \ |
642 |
for(i=0; i<64; ++i) s+=(DST)[i]^i^qm; \ |
s = calc_crc((uint8_t*)(DST), 64*sizeof(int16_t), s); \ |
643 |
} \ |
} \ |
644 |
emms(); \ |
emms(); \ |
645 |
} \ |
} \ |
646 |
t = (gettime_usec()-t-overhead)/nb_tests/qm;\ |
t = (gettime_usec()-t-overhead)/nb_tests/qm |
|
s = (s&0xffff)^(s>>16) |
|
647 |
|
|
648 |
#define TEST_QUANT2(FUNC, DST, SRC) \ |
#define TEST_QUANT2(FUNC, DST, SRC) \ |
649 |
t = gettime_usec(); \ |
t = gettime_usec(); \ |
650 |
for(s=0,qm=1; qm<=255; ++qm) { \ |
for(s=CRC32_INITIAL,qm=1; qm<=255; ++qm) { \ |
651 |
for(i=0; i<8*8; ++i) Quant[i] = qm; \ |
for(i=0; i<8*8; ++i) Quant[i] = qm; \ |
652 |
set_intra_matrix( Quant ); \ |
set_intra_matrix( mpeg_quant_matrices, Quant ); \ |
653 |
emms(); \ |
emms(); \ |
654 |
for(q=1; q<=max_Q; ++q) { \ |
for(q=1; q<=max_Q; ++q) { \ |
655 |
for(tst=0; tst<nb_tests; ++tst) \ |
for(tst=0; tst<nb_tests; ++tst) \ |
656 |
(FUNC)((DST), (SRC), q, q); \ |
(FUNC)((DST), (SRC), q, q, mpeg_quant_matrices); \ |
657 |
for(i=0; i<64; ++i) s+=(DST)[i]^i^qm; \ |
s = calc_crc((uint8_t*)(DST), 64*sizeof(int16_t), s); \ |
658 |
|
} \ |
659 |
|
emms(); \ |
660 |
|
} \ |
661 |
|
t = (gettime_usec()-t-overhead)/nb_tests/qm |
662 |
|
|
663 |
|
#define TEST_INTRA(REFFUNC, NEWFUNC, RANGE) \ |
664 |
|
{ int i,q,s;\ |
665 |
|
DECLARE_ALIGNED_MATRIX(Src, 8, 8, int16_t, 16); \ |
666 |
|
DECLARE_ALIGNED_MATRIX(Dst, 8, 8, int16_t, 16); \ |
667 |
|
DECLARE_ALIGNED_MATRIX(Dst2,8, 8, int16_t, 16); \ |
668 |
|
for(q=1;q<=max_Q;q++) \ |
669 |
|
for(s=-RANGE;s<RANGE;s++) { \ |
670 |
|
for(i=0;i<64;i++) Src[i]=s; \ |
671 |
|
(REFFUNC)((Dst),(Src),q,q,mpeg_quant_matrices); \ |
672 |
|
(NEWFUNC)((Dst2),(Src),q,q,mpeg_quant_matrices); \ |
673 |
|
for(i=0;i<64;i++) \ |
674 |
|
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]); \ |
675 |
} \ |
} \ |
676 |
|
} |
677 |
|
|
678 |
|
#define TEST_INTER(REFFUNC, NEWFUNC, RANGE) \ |
679 |
|
{ int i,q,s; \ |
680 |
|
DECLARE_ALIGNED_MATRIX(Src, 8, 8, int16_t, 16); \ |
681 |
|
DECLARE_ALIGNED_MATRIX(Dst, 8, 8, int16_t, 16); \ |
682 |
|
DECLARE_ALIGNED_MATRIX(Dst2,8, 8, int16_t, 16); \ |
683 |
|
for(q=1;q<=max_Q;q++) \ |
684 |
|
for(s=-RANGE;s<RANGE;s++) { \ |
685 |
|
for(i=0;i<64;i++) Src[i]=s; \ |
686 |
|
(REFFUNC)((Dst),(Src),q,mpeg_quant_matrices); \ |
687 |
|
(NEWFUNC)((Dst2),(Src),q,mpeg_quant_matrices); \ |
688 |
emms(); \ |
emms(); \ |
689 |
|
for(i=0;i<64;i++) \ |
690 |
|
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]); \ |
691 |
} \ |
} \ |
692 |
t = (gettime_usec()-t-overhead)/nb_tests/qm;\ |
} |
|
s = (s&0xffff)^(s>>16) |
|
693 |
|
|
694 |
void test_quant() |
void test_quant() |
695 |
{ |
{ |
696 |
const int nb_tests = 1*speed_ref; |
const int nb_tests = 1*speed_ref; |
697 |
const int max_Q = 31; |
const int max_Q = 31; |
698 |
|
DECLARE_ALIGNED_MATRIX(mpeg_quant_matrices, 8, 64, uint16_t, 16); |
699 |
|
|
700 |
int i, qm; |
int i, qm; |
701 |
CPU *cpu; |
CPU *cpu; |
702 |
int16_t Src[8*8], Dst[8*8]; |
DECLARE_ALIGNED_MATRIX(Src, 8, 8, int16_t, 16); |
703 |
|
DECLARE_ALIGNED_MATRIX(Dst, 8, 8, int16_t, 16); |
704 |
|
DECLARE_ALIGNED_MATRIX(Dst2,8, 8, int16_t, 16); |
705 |
uint8_t Quant[8*8]; |
uint8_t Quant[8*8]; |
706 |
|
|
707 |
printf( "\n ===== test quant =====\n" ); |
printf( "\n ===== test quant =====\n" ); |
708 |
|
|
709 |
// we deliberately enfringe the norm's specified range [-127,127], |
/* we deliberately enfringe the norm's specified range [-127,127], */ |
710 |
// to test the robustness of the iquant module |
/* to test the robustness of the iquant module */ |
711 |
for(i=0; i<64; ++i) { |
for(i=0; i<64; ++i) { |
712 |
Src[i] = 1 + (i-32) * (i&6); |
Src[i] = 1 + (i-32) * (i&6); |
713 |
Dst[i] = 0; |
Dst[i] = 0; |
714 |
} |
} |
715 |
|
|
716 |
for(cpu = cpu_short_list; cpu->name!=0; ++cpu) |
for(cpu = cpu_list; cpu->name!=0; ++cpu) |
717 |
{ |
{ |
718 |
double t, overhead; |
double t, overhead; |
719 |
int tst, q; |
int tst, q; |
722 |
if (!init_cpu(cpu)) |
if (!init_cpu(cpu)) |
723 |
continue; |
continue; |
724 |
|
|
725 |
|
// exhaustive tests to compare against the (ref) C-version |
726 |
|
TEST_INTRA(quant_h263_intra_c, quant_h263_intra, 2048); |
727 |
|
TEST_INTRA(dequant_h263_intra_c, dequant_h263_intra , 512 ); |
728 |
|
TEST_INTER(quant_h263_inter_c, quant_h263_inter , 2048); |
729 |
|
TEST_INTER(dequant_h263_inter_c, dequant_h263_inter , 512 ); |
730 |
|
|
731 |
overhead = -gettime_usec(); |
overhead = -gettime_usec(); |
732 |
for(s=0,qm=1; qm<=255; ++qm) { |
for(s=0,qm=1; qm<=255; ++qm) { |
733 |
for(i=0; i<8*8; ++i) Quant[i] = qm; |
for(i=0; i<8*8; ++i) Quant[i] = qm; |
734 |
set_inter_matrix( Quant ); |
set_inter_matrix(mpeg_quant_matrices, Quant ); |
735 |
for(q=1; q<=max_Q; ++q) |
for(q=1; q<=max_Q; ++q) |
736 |
for(i=0; i<64; ++i) s+=Dst[i]^i^qm; |
for(i=0; i<64; ++i) s+=Dst[i]^i^qm; |
737 |
} |
} |
738 |
overhead += gettime_usec(); |
overhead += gettime_usec(); |
739 |
|
|
740 |
#if 1 |
TEST_QUANT2(quant_mpeg_intra, Dst, Src); |
741 |
TEST_QUANT2(quant4_intra, Dst, Src); |
printf("%s - quant_mpeg_intra %.3f usec crc32=0x%08x %s\n", |
742 |
printf( "%s - quant4_intra %.3f usec crc=%d\n", cpu->name, t, s ); |
cpu->name, t, s, |
743 |
if (s!=29809) printf( "*** CRC ERROR! ***\n" ); |
(s!=0xfd6a21a4)? "| ERROR": ""); |
744 |
|
|
745 |
TEST_QUANT(quant4_inter, Dst, Src); |
TEST_QUANT(quant_mpeg_inter, Dst, Src); |
746 |
printf( "%s - quant4_inter %.3f usec crc=%d\n", cpu->name, t, s ); |
printf("%s - quant_mpeg_inter %.3f usec crc32=0x%08x %s\n", |
747 |
if (s!=12574) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, s, |
748 |
#endif |
(s!=0xf6de7757)?"| ERROR": ""); |
749 |
#if 1 |
|
750 |
TEST_QUANT2(dequant4_intra, Dst, Src); |
TEST_QUANT2(dequant_mpeg_intra, Dst, Src); |
751 |
printf( "%s - dequant4_intra %.3f usec crc=%d\n", cpu->name, t, s ); |
printf("%s - dequant_mpeg_intra %.3f usec crc32=0x%08x %s\n", |
752 |
if (s!=24052) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, s, |
753 |
|
(s!=0x2def7bc7)?"| ERROR": ""); |
754 |
TEST_QUANT(dequant4_inter, Dst, Src); |
|
755 |
printf( "%s - dequant4_inter %.3f usec crc=%d\n", cpu->name, t, s ); |
TEST_QUANT(dequant_mpeg_inter, Dst, Src); |
756 |
if (s!=63847) printf( "*** CRC ERROR! ***\n" ); |
printf("%s - dequant_mpeg_inter %.3f usec crc32=0x%08x %s\n", |
757 |
#endif |
cpu->name, t, s, |
758 |
#if 1 |
(s!=0xd878c722)?"| ERROR": ""); |
759 |
TEST_QUANT2(quant_intra, Dst, Src); |
|
760 |
printf( "%s - quant_intra %.3f usec crc=%d\n", cpu->name, t, s ); |
TEST_QUANT2(quant_h263_intra, Dst, Src); |
761 |
if (s!=25662) printf( "*** CRC ERROR! ***\n" ); |
printf("%s - quant_h263_intra %.3f usec crc32=0x%08x %s\n", |
762 |
|
cpu->name, t, s, |
763 |
TEST_QUANT(quant_inter, Dst, Src); |
(s!=0x2eba9d43)?"| ERROR": ""); |
764 |
printf( "%s - quant_inter %.3f usec crc=%d\n", cpu->name, t, s ); |
|
765 |
if (s!=23972) printf( "*** CRC ERROR! ***\n" ); |
TEST_QUANT(quant_h263_inter, Dst, Src); |
766 |
#endif |
printf("%s - quant_h263_inter %.3f usec crc32=0x%08x %s\n", |
767 |
#if 1 |
cpu->name, t, s, |
768 |
TEST_QUANT2(dequant_intra, Dst, Src); |
(s!=0xbd315a7e)?"| ERROR": ""); |
769 |
printf( "%s - dequant_intra %.3f usec crc=%d\n", cpu->name, t, s ); |
|
770 |
if (s!=49900) printf( "*** CRC ERROR! ***\n" ); |
TEST_QUANT2(dequant_h263_intra, Dst, Src); |
771 |
|
printf("%s - dequant_h263_intra %.3f usec crc32=0x%08x %s\n", |
772 |
TEST_QUANT(dequant_inter, Dst, Src); |
cpu->name, t, s, |
773 |
printf( "%s - dequant_inter %.3f usec crc=%d\n", cpu->name, t, s ); |
(s!=0x9841212a)?"| ERROR": ""); |
774 |
if (s!=48899) printf( "*** CRC ERROR! ***\n" ); |
|
775 |
#endif |
TEST_QUANT(dequant_h263_inter, Dst, Src); |
776 |
|
printf("%s - dequant_h263_inter %.3f usec crc32=0x%08x %s\n", |
777 |
|
cpu->name, t, s, |
778 |
|
(s!=0xe7df8fba)?"| ERROR": ""); |
779 |
|
|
780 |
printf( " --- \n" ); |
printf( " --- \n" ); |
781 |
} |
} |
782 |
} |
} |
799 |
const int nb_tests = 10000*speed_ref; |
const int nb_tests = 10000*speed_ref; |
800 |
int i; |
int i; |
801 |
CPU *cpu; |
CPU *cpu; |
802 |
int16_t Src1[6*64], Src2[6*64], Src3[6*64], Src4[6*64]; |
DECLARE_ALIGNED_MATRIX(Src1, 6, 64, int16_t, 16); |
803 |
|
DECLARE_ALIGNED_MATRIX(Src2, 6, 64, int16_t, 16); |
804 |
|
DECLARE_ALIGNED_MATRIX(Src3, 6, 64, int16_t, 16); |
805 |
|
DECLARE_ALIGNED_MATRIX(Src4, 6, 64, int16_t, 16); |
806 |
|
|
807 |
printf( "\n ===== test cbp =====\n" ); |
printf( "\n ===== test cbp =====\n" ); |
808 |
|
|
809 |
for(i=0; i<6*64; ++i) { |
for(i=0; i<6*64; ++i) { |
810 |
Src1[i] = (i*i*3/8192)&(i/64)&1; // 'random' |
Src1[i] = (i*i*3/8192)&(i/64)&1; /* 'random' */ |
811 |
Src2[i] = (i<3*64); // half-full |
Src2[i] = (i<3*64); /* half-full */ |
812 |
Src3[i] = ((i+32)>3*64); |
Src3[i] = ((i+32)>3*64); |
813 |
Src4[i] = (i==(3*64+2) || i==(5*64+9)); |
Src4[i] = (i==(3*64+2) || i==(5*64+9)); |
814 |
} |
} |
815 |
|
|
816 |
for(cpu = cpu_short_list2; cpu->name!=0; ++cpu) |
for(cpu = cpu_list; cpu->name!=0; ++cpu) |
817 |
{ |
{ |
818 |
double t; |
double t; |
819 |
int tst, cbp; |
int tst, cbp; |
822 |
continue; |
continue; |
823 |
|
|
824 |
TEST_CBP(calc_cbp, Src1); |
TEST_CBP(calc_cbp, Src1); |
825 |
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\n", |
826 |
if (cbp!=0x15) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, cbp, (cbp!=0x15)?"| ERROR": ""); |
827 |
TEST_CBP(calc_cbp, Src2); |
TEST_CBP(calc_cbp, Src2); |
828 |
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\n", |
829 |
if (cbp!=0x38) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, cbp, (cbp!=0x38)?"| ERROR": ""); |
830 |
TEST_CBP(calc_cbp, Src3); |
TEST_CBP(calc_cbp, Src3); |
831 |
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\n", |
832 |
if (cbp!=0x0f) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, cbp, (cbp!=0x0f)?"| ERROR": "" ); |
833 |
TEST_CBP(calc_cbp, Src4); |
TEST_CBP(calc_cbp, Src4); |
834 |
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\n", |
835 |
if (cbp!=0x05) printf( "*** CRC ERROR! ***\n" ); |
cpu->name, t, cbp, (cbp!=0x05)?"| ERROR": "" ); |
836 |
printf( " --- \n" ); |
printf( " --- \n" ); |
837 |
} |
} |
838 |
} |
} |
963 |
} |
} |
964 |
} |
} |
965 |
|
|
966 |
////////////////////////////////////////////////////////// |
///* ////////////////////////////////////////////////////// */ |
967 |
/* Pseudo-random generator specified by IEEE 1180 */ |
/* Pseudo-random generator specified by IEEE 1180 */ |
968 |
|
|
969 |
static long ieee_seed = 1; |
static long ieee_seed = 1; |
1050 |
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 }; |
1051 |
int Loops = 10000; |
int Loops = 10000; |
1052 |
int i, m, n; |
int i, m, n; |
1053 |
short Blk0[64]; // reference |
DECLARE_ALIGNED_MATRIX(Blk0, 8, 8, short, 16); /* reference */ |
1054 |
short Blk[64], iBlk[64]; |
DECLARE_ALIGNED_MATRIX(Blk, 8, 8, short, 16); |
1055 |
short Ref_FDCT[64]; |
DECLARE_ALIGNED_MATRIX(iBlk, 8, 8, short, 16); |
1056 |
short Ref_IDCT[64]; |
DECLARE_ALIGNED_MATRIX(Ref_FDCT, 8, 8, short, 16); |
1057 |
|
DECLARE_ALIGNED_MATRIX(Ref_IDCT, 8, 8, short, 16); |
1058 |
|
|
1059 |
STATS_8x8 FStats; // forward dct stats |
STATS_8x8 FStats; /* forward dct stats */ |
1060 |
STATS_8x8 IStats; // inverse dct stats |
STATS_8x8 IStats; /* inverse dct stats */ |
1061 |
|
|
1062 |
CPU *cpu; |
CPU *cpu; |
1063 |
|
|
1080 |
for(i=0; i<64; ++i) |
for(i=0; i<64; ++i) |
1081 |
Blk0[i] = (short)ieee_rand(Min,Max) * Sign; |
Blk0[i] = (short)ieee_rand(Min,Max) * Sign; |
1082 |
|
|
1083 |
// hmm, I'm not quite sure this is exactly |
/* hmm, I'm not quite sure this is exactly */ |
1084 |
// the tests described in the norm. check... |
/* the tests described in the norm. check... */ |
1085 |
|
|
1086 |
memcpy(Ref_FDCT, Blk0, 64*sizeof(short)); |
memcpy(Ref_FDCT, Blk0, 64*sizeof(short)); |
1087 |
ref_fdct(Ref_FDCT); |
ref_fdct(Ref_FDCT); |
1108 |
|
|
1109 |
printf( "\n -- FDCT report --\n" ); |
printf( "\n -- FDCT report --\n" ); |
1110 |
// print_stats(&FStats); |
// print_stats(&FStats); |
1111 |
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 */ |
1112 |
|
|
1113 |
for(i=0; i<64; i++) Blk[i] = 0; |
for(i=0; i<64; i++) Blk[i] = 0; |
1114 |
emms(); fdct(Blk); emms(); |
emms(); fdct(Blk); emms(); |
1130 |
|
|
1131 |
void test_dct_saturation(int Min, int Max) |
void test_dct_saturation(int Min, int Max) |
1132 |
{ |
{ |
1133 |
// test behaviour on input range fringe |
/* test behaviour on input range fringe */ |
1134 |
|
|
1135 |
int i, n, p; |
int i, n, p; |
1136 |
CPU *cpu; |
CPU *cpu; |
1137 |
// const short IDCT_MAX = 2047; // 12bits input |
// const short IDCT_MAX = 2047; /* 12bits input */ |
1138 |
// const short IDCT_MIN = -2048; |
// const short IDCT_MIN = -2048; |
1139 |
// const short IDCT_OUT = 256; // 9bits ouput |
// const short IDCT_OUT = 256; /* 9bits ouput */ |
1140 |
const int Partitions = 4; |
const int Partitions = 4; |
1141 |
const int Loops = 10000 / Partitions; |
const int Loops = 10000 / Partitions; |
1142 |
|
|
1153 |
printf( "\n===== IEEE test for %s Min=%d Max=%d =====\n", |
printf( "\n===== IEEE test for %s Min=%d Max=%d =====\n", |
1154 |
cpu->name, Min, Max ); |
cpu->name, Min, Max ); |
1155 |
|
|
1156 |
// FDCT tests // |
/* FDCT tests // */ |
1157 |
|
|
1158 |
init_stats(&Stats); |
init_stats(&Stats); |
1159 |
|
|
1160 |
// test each computation channels separately |
/* test each computation channels separately */ |
1161 |
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; |
1162 |
ref_fdct(Blk0); |
ref_fdct(Blk0); |
1163 |
emms(); fdct(Blk); emms(); |
emms(); fdct(Blk); emms(); |
1168 |
emms(); fdct(Blk); emms(); |
emms(); fdct(Blk); emms(); |
1169 |
store_stats(&Stats, Blk, Blk0); |
store_stats(&Stats, Blk, Blk0); |
1170 |
|
|
1171 |
// randomly saturated inputs |
/* randomly saturated inputs */ |
1172 |
for(p=0; p<Partitions; ++p) |
for(p=0; p<Partitions; ++p) |
1173 |
{ |
{ |
1174 |
for(n=0; n<Loops; ++n) |
for(n=0; n<Loops; ++n) |
1184 |
report_stats(&Stats, 0); |
report_stats(&Stats, 0); |
1185 |
|
|
1186 |
|
|
1187 |
// IDCT tests // |
/* IDCT tests // */ |
1188 |
#if 0 |
#if 0 |
1189 |
// no finished yet |
/* no finished yet */ |
1190 |
|
|
1191 |
init_stats(&Stats); |
init_stats(&Stats); |
1192 |
|
|
1193 |
// test each computation channel separately |
/* test each computation channel separately */ |
1194 |
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; |
1195 |
ref_idct(Blk0); |
ref_idct(Blk0); |
1196 |
emms(); idct(Blk); emms(); |
emms(); idct(Blk); emms(); |
1203 |
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); } |
1204 |
store_stats(&Stats, Blk, Blk0); |
store_stats(&Stats, Blk, Blk0); |
1205 |
|
|
1206 |
// randomly saturated inputs |
/* randomly saturated inputs */ |
1207 |
for(p=0; p<Partitions; ++p) |
for(p=0; p<Partitions; ++p) |
1208 |
{ |
{ |
1209 |
for(n=0; n<Loops; ++n) |
for(n=0; n<Loops; ++n) |
1233 |
FILE *f = 0; |
FILE *f = 0; |
1234 |
void *dechandle = 0; |
void *dechandle = 0; |
1235 |
int xerr; |
int xerr; |
1236 |
XVID_INIT_PARAM xinit; |
xvid_gbl_init_t xinit; |
1237 |
XVID_DEC_PARAM xparam; |
xvid_dec_create_t xparam; |
1238 |
XVID_DEC_FRAME xframe; |
xvid_dec_frame_t xframe; |
1239 |
double t = 0.; |
double t = 0.; |
1240 |
int nb = 0; |
int nb = 0; |
1241 |
uint8_t *buf = 0; |
uint8_t *buf = 0; |
1243 |
int buf_size, pos; |
int buf_size, pos; |
1244 |
uint32_t chksum = 0; |
uint32_t chksum = 0; |
1245 |
|
|
1246 |
|
memset(&xinit, 0, sizeof(xinit)); |
1247 |
xinit.cpu_flags = XVID_CPU_MMX | XVID_CPU_FORCE; |
xinit.cpu_flags = XVID_CPU_MMX | XVID_CPU_FORCE; |
1248 |
xvid_init(NULL, 0, &xinit, NULL); |
xinit.version = XVID_VERSION; |
1249 |
printf( "API version: %d, core build:%d\n", xinit.api_version, xinit.core_build); |
xvid_global(NULL, 0, &xinit, NULL); |
|
|
|
1250 |
|
|
1251 |
|
memset(&xparam, 0, sizeof(xparam)); |
1252 |
xparam.width = width; |
xparam.width = width; |
1253 |
xparam.height = height; |
xparam.height = height; |
1254 |
|
xparam.version = XVID_VERSION; |
1255 |
xerr = xvid_decore(NULL, XVID_DEC_CREATE, &xparam, NULL); |
xerr = xvid_decore(NULL, XVID_DEC_CREATE, &xparam, NULL); |
1256 |
if (xerr!=XVID_ERR_OK) { |
if (xerr==XVID_ERR_FAIL) { |
1257 |
printf("can't init decoder (err=%d)\n", xerr); |
printf("can't init decoder (err=%d)\n", xerr); |
1258 |
return; |
return; |
1259 |
} |
} |
1274 |
} |
} |
1275 |
else printf( "Input size: %d\n", buf_size); |
else printf( "Input size: %d\n", buf_size); |
1276 |
|
|
1277 |
buf = malloc(buf_size); // should be enuf' |
buf = malloc(buf_size); /* should be enuf' */ |
1278 |
rgb_out = calloc(4, width*height); // <-room for _RGB24 |
rgb_out = calloc(4, width*height); /* <-room for _RGB24 */ |
1279 |
if (buf==0 || rgb_out==0) { |
if (buf==0 || rgb_out==0) { |
1280 |
printf( "malloc failed!\n" ); |
printf( "malloc failed!\n" ); |
1281 |
goto End; |
goto End; |
1290 |
pos = 0; |
pos = 0; |
1291 |
t = -gettime_usec(); |
t = -gettime_usec(); |
1292 |
while(1) { |
while(1) { |
1293 |
|
memset(&xframe, 0, sizeof(xframe)); |
1294 |
|
xframe.version = XVID_VERSION; |
1295 |
xframe.bitstream = buf + pos; |
xframe.bitstream = buf + pos; |
1296 |
xframe.length = buf_size - pos; |
xframe.length = buf_size - pos; |
1297 |
xframe.image = rgb_out; |
xframe.output.plane[0] = rgb_out; |
1298 |
xframe.stride = width; |
xframe.output.stride[0] = width; |
1299 |
xframe.colorspace = XVID_CSP_RGB24; |
xframe.output.csp = XVID_CSP_BGR; |
1300 |
xerr = xvid_decore(dechandle, XVID_DEC_DECODE, &xframe, 0); |
xerr = xvid_decore(dechandle, XVID_DEC_DECODE, &xframe, 0); |
1301 |
nb++; |
nb++; |
1302 |
pos += xframe.length; |
pos += xframe.length; |
1307 |
} |
} |
1308 |
if (pos==buf_size) |
if (pos==buf_size) |
1309 |
break; |
break; |
1310 |
if (xerr!=XVID_ERR_OK) { |
if (xerr==XVID_ERR_FAIL) { |
1311 |
printf("decoding failed for frame #%d (err=%d)!\n", nb, xerr); |
printf("decoding failed for frame #%d (err=%d)!\n", nb, xerr); |
1312 |
break; |
break; |
1313 |
} |
} |
1323 |
if (buf!=0) free(buf); |
if (buf!=0) free(buf); |
1324 |
if (dechandle!=0) { |
if (dechandle!=0) { |
1325 |
xerr= xvid_decore(dechandle, XVID_DEC_DESTROY, NULL, NULL); |
xerr= xvid_decore(dechandle, XVID_DEC_DESTROY, NULL, NULL); |
1326 |
if (xerr!=XVID_ERR_OK) |
if (xerr==XVID_ERR_FAIL) |
1327 |
printf("destroy-decoder failed (err=%d)!\n", xerr); |
printf("destroy-decoder failed (err=%d)!\n", xerr); |
1328 |
} |
} |
1329 |
if (f!=0) fclose(f); |
if (f!=0) fclose(f); |
1336 |
void test_bugs1() |
void test_bugs1() |
1337 |
{ |
{ |
1338 |
CPU *cpu; |
CPU *cpu; |
1339 |
|
uint16_t mpeg_quant_matrices[64*8]; |
1340 |
|
|
1341 |
printf( "\n ===== (de)quant4_intra saturation bug? =====\n" ); |
printf( "\n ===== (de)quant4_intra saturation bug? =====\n" ); |
1342 |
|
|
1343 |
for(cpu = cpu_short_list; cpu->name!=0; ++cpu) |
for(cpu = cpu_list; cpu->name!=0; ++cpu) |
1344 |
{ |
{ |
1345 |
int i; |
int i; |
1346 |
int16_t Src[8*8], Dst[8*8]; |
int16_t Src[8*8], Dst[8*8]; |
1349 |
continue; |
continue; |
1350 |
|
|
1351 |
for(i=0; i<64; ++i) Src[i] = i-32; |
for(i=0; i<64; ++i) Src[i] = i-32; |
1352 |
set_intra_matrix( get_default_intra_matrix() ); |
set_intra_matrix( mpeg_quant_matrices, get_default_intra_matrix() ); |
1353 |
dequant4_intra(Dst, Src, 31, 5); |
dequant_mpeg_intra(Dst, Src, 31, 5, mpeg_quant_matrices); |
1354 |
printf( "dequant4_intra with CPU=%s: ", cpu->name); |
printf( "dequant_mpeg_intra with CPU=%s: ", cpu->name); |
1355 |
printf( " Out[]= " ); |
printf( " Out[]= " ); |
1356 |
for(i=0; i<64; ++i) printf( "[%d]", Dst[i]); |
for(i=0; i<64; ++i) printf( "[%d]", Dst[i]); |
1357 |
printf( "\n" ); |
printf( "\n" ); |
1359 |
|
|
1360 |
printf( "\n ===== (de)quant4_inter saturation bug? =====\n" ); |
printf( "\n ===== (de)quant4_inter saturation bug? =====\n" ); |
1361 |
|
|
1362 |
for(cpu = cpu_short_list; cpu->name!=0; ++cpu) |
for(cpu = cpu_list; cpu->name!=0; ++cpu) |
1363 |
{ |
{ |
1364 |
int i; |
int i; |
1365 |
int16_t Src[8*8], Dst[8*8]; |
int16_t Src[8*8], Dst[8*8]; |
1368 |
continue; |
continue; |
1369 |
|
|
1370 |
for(i=0; i<64; ++i) Src[i] = i-32; |
for(i=0; i<64; ++i) Src[i] = i-32; |
1371 |
set_inter_matrix( get_default_inter_matrix() ); |
set_inter_matrix( mpeg_quant_matrices, get_default_inter_matrix() ); |
1372 |
dequant4_inter(Dst, Src, 31); |
dequant_mpeg_inter(Dst, Src, 31, mpeg_quant_matrices); |
1373 |
printf( "dequant4_inter with CPU=%s: ", cpu->name); |
printf( "dequant_mpeg_inter with CPU=%s: ", cpu->name); |
1374 |
printf( " Out[]= " ); |
printf( " Out[]= " ); |
1375 |
for(i=0; i<64; ++i) printf( "[%d]", Dst[i]); |
for(i=0; i<64; ++i) printf( "[%d]", Dst[i]); |
1376 |
printf( "\n" ); |
printf( "\n" ); |
1380 |
void test_dct_precision_diffs() |
void test_dct_precision_diffs() |
1381 |
{ |
{ |
1382 |
CPU *cpu; |
CPU *cpu; |
1383 |
short Blk[8*8], Blk0[8*8]; |
DECLARE_ALIGNED_MATRIX(Blk, 8, 8, int16_t, 16); |
1384 |
|
DECLARE_ALIGNED_MATRIX(Blk0, 8, 8, int16_t, 16); |
1385 |
|
|
1386 |
printf( "\n ===== fdct/idct precision diffs =====\n" ); |
printf( "\n ===== fdct/idct precision diffs =====\n" ); |
1387 |
|
|
1388 |
for(cpu = cpu_short_list; cpu->name!=0; ++cpu) |
for(cpu = cpu_list; cpu->name!=0; ++cpu) |
1389 |
{ |
{ |
1390 |
int i; |
int i; |
1391 |
|
|
1414 |
const int max_Q = 31; |
const int max_Q = 31; |
1415 |
int i, n, qm, q; |
int i, n, qm, q; |
1416 |
CPU *cpu; |
CPU *cpu; |
1417 |
int16_t Src[8*8], Dst[8*8]; |
DECLARE_ALIGNED_MATRIX(Src, 8, 8, int16_t, 16); |
1418 |
|
DECLARE_ALIGNED_MATRIX(Dst, 8, 8, int16_t, 16); |
1419 |
uint8_t Quant[8*8]; |
uint8_t Quant[8*8]; |
1420 |
CPU cpu_bug_list[] = { { "PLAINC", 0 }, { "MMX ", XVID_CPU_MMX }, {0,0} }; |
CPU cpu_bug_list[] = { { "PLAINC", 0 }, { "MMX ", XVID_CPU_MMX }, {0,0} }; |
1421 |
uint16_t Crcs_Inter[2][32]; |
uint16_t Crcs_Inter[2][32]; |
1422 |
uint16_t Crcs_Intra[2][32]; |
uint16_t Crcs_Intra[2][32]; |
1423 |
|
DECLARE_ALIGNED_MATRIX(mpeg_quant_matrices, 8, 64, uint16_t, 16); |
1424 |
|
|
1425 |
printf( "\n ===== test MPEG4-quantize bug =====\n" ); |
printf( "\n ===== test MPEG4-quantize bug =====\n" ); |
1426 |
|
|
1427 |
for(i=0; i<64; ++i) Src[i] = 2048*(i-32)/32; |
for(i=0; i<64; ++i) Src[i] = 2048*(i-32)/32; |
1430 |
for(qm=1; qm<=255; ++qm) |
for(qm=1; qm<=255; ++qm) |
1431 |
{ |
{ |
1432 |
for(i=0; i<8*8; ++i) Quant[i] = qm; |
for(i=0; i<8*8; ++i) Quant[i] = qm; |
1433 |
set_inter_matrix( Quant ); |
set_inter_matrix( mpeg_quant_matrices, Quant ); |
1434 |
|
|
1435 |
for(n=0, cpu = cpu_bug_list; cpu->name!=0; ++cpu, ++n) |
for(n=0, cpu = cpu_bug_list; cpu->name!=0; ++cpu, ++n) |
1436 |
{ |
{ |
1441 |
|
|
1442 |
for(q=1; q<=max_Q; ++q) { |
for(q=1; q<=max_Q; ++q) { |
1443 |
emms(); |
emms(); |
1444 |
quant4_inter( Dst, Src, q ); |
quant_mpeg_inter( Dst, Src, q, mpeg_quant_matrices ); |
1445 |
emms(); |
emms(); |
1446 |
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; |
1447 |
Crcs_Inter[n][q] = s; |
Crcs_Inter[n][q] = s; |
1460 |
for(qm=1; qm<=255; ++qm) |
for(qm=1; qm<=255; ++qm) |
1461 |
{ |
{ |
1462 |
for(i=0; i<8*8; ++i) Quant[i] = qm; |
for(i=0; i<8*8; ++i) Quant[i] = qm; |
1463 |
set_intra_matrix( Quant ); |
set_intra_matrix( mpeg_quant_matrices, Quant ); |
1464 |
|
|
1465 |
for(n=0, cpu = cpu_bug_list; cpu->name!=0; ++cpu, ++n) |
for(n=0, cpu = cpu_bug_list; cpu->name!=0; ++cpu, ++n) |
1466 |
{ |
{ |
1471 |
|
|
1472 |
for(q=1; q<=max_Q; ++q) { |
for(q=1; q<=max_Q; ++q) { |
1473 |
emms(); |
emms(); |
1474 |
quant4_intra( Dst, Src, q, q); |
quant_mpeg_intra( Dst, Src, q, q, mpeg_quant_matrices); |
1475 |
emms(); |
emms(); |
1476 |
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; |
1477 |
Crcs_Intra[n][q] = s; |
Crcs_Intra[n][q] = s; |
1504 |
|
|
1505 |
if (what==7) { |
if (what==7) { |
1506 |
test_IEEE1180_compliance(-256, 255, 1); |
test_IEEE1180_compliance(-256, 255, 1); |
|
#if 0 |
|
1507 |
test_IEEE1180_compliance(-256, 255,-1); |
test_IEEE1180_compliance(-256, 255,-1); |
1508 |
test_IEEE1180_compliance( -5, 5, 1); |
test_IEEE1180_compliance( -5, 5, 1); |
1509 |
test_IEEE1180_compliance( -5, 5,-1); |
test_IEEE1180_compliance( -5, 5,-1); |
1510 |
test_IEEE1180_compliance(-300, 300, 1); |
test_IEEE1180_compliance(-300, 300, 1); |
1511 |
test_IEEE1180_compliance(-300, 300,-1); |
test_IEEE1180_compliance(-300, 300,-1); |
|
#endif |
|
1512 |
} |
} |
1513 |
if (what==8) test_dct_saturation(-256, 255); |
if (what==8) test_dct_saturation(-256, 255); |
1514 |
|
|
1530 |
if (what==-2) |
if (what==-2) |
1531 |
test_quant_bug(); |
test_quant_bug(); |
1532 |
|
|
1533 |
|
if (what >= 0 && what <= 6) { |
1534 |
|
printf("\n\n" |
1535 |
|
"NB: If a function isn't optimised for a specific set of intructions,\n" |
1536 |
|
" a C function is used instead. So don't panic if some functions\n" |
1537 |
|
" may appear to be slow.\n"); |
1538 |
|
} |
1539 |
|
|
1540 |
|
#ifdef ARCH_IS_IA32 |
1541 |
|
if (what == 0 || what == 5) { |
1542 |
|
printf("\n" |
1543 |
|
"NB: MMX mpeg4 quantization is known to have very small errors (+/-1 magnitude)\n" |
1544 |
|
" for 1 or 2 coefficients a block. This is mainly caused by the fact the unit\n" |
1545 |
|
" test goes far behind the usual limits of real encoding. Please do not report\n" |
1546 |
|
" this error to the developers.\n"); |
1547 |
|
} |
1548 |
|
#endif |
1549 |
|
|
1550 |
return 0; |
return 0; |
1551 |
} |
} |
1552 |
|
|
1553 |
/********************************************************************* |
/********************************************************************* |
1554 |
* 'Reference' output (except for timing) on a PIII 1.13Ghz/linux |
* 'Reference' output (except for timing) on an Athlon XP 2200+ |
1555 |
*********************************************************************/ |
*********************************************************************/ |
1556 |
|
|
1557 |
/* as of 07/01/2002, there's a problem with mpeg4-quantization */ |
/* as of 2002-01-07, there's a problem with MMX mpeg4-quantization */ |
1558 |
/* |
/* as of 2003-11-30, the problem is still here */ |
1559 |
|
|
1560 |
|
/********************************************************************* |
1561 |
|
|
1562 |
|
|
1563 |
===== test fdct/idct ===== |
===== test fdct/idct ===== |
1564 |
PLAINC - 3.312 usec PSNR=13.291 MSE=3.000 |
PLAINC - 2.867 usec PSNR=13.291 MSE=3.000 |
1565 |
MMX - 0.591 usec PSNR=13.291 MSE=3.000 |
MMX - -0.211 usec PSNR=9.611 MSE=7.000 |
1566 |
MMXEXT - 0.577 usec PSNR=13.291 MSE=3.000 |
MMXEXT - -0.256 usec PSNR=9.611 MSE=7.000 |
1567 |
SSE2 - 0.588 usec PSNR=13.291 MSE=3.000 |
3DNOW - 2.855 usec PSNR=13.291 MSE=3.000 |
1568 |
3DNOW - skipped... |
3DNOWE - 1.429 usec PSNR=13.291 MSE=3.000 |
|
3DNOWE - skipped... |
|
1569 |
|
|
1570 |
=== test block motion === |
=== test block motion === |
1571 |
PLAINC - interp- h-round0 0.911 usec iCrc=8107 |
PLAINC - interp- h-round0 0.538 usec crc32=0x115381ba |
1572 |
PLAINC - round1 0.863 usec iCrc=8100 |
PLAINC - round1 0.527 usec crc32=0x2b1f528f |
1573 |
PLAINC - interp- v-round0 0.860 usec iCrc=8108 |
PLAINC - interp- v-round0 0.554 usec crc32=0x423cdcc7 |
1574 |
PLAINC - round1 0.857 usec iCrc=8105 |
PLAINC - round1 0.551 usec crc32=0x42202efe |
1575 |
PLAINC - interp-hv-round0 2.103 usec iCrc=8112 |
PLAINC - interp-hv-round0 1.041 usec crc32=0xd198d387 |
1576 |
PLAINC - round1 2.050 usec iCrc=8103 |
PLAINC - round1 1.038 usec crc32=0x9ecfd921 |
1577 |
--- |
--- |
1578 |
MMX - interp- h-round0 0.105 usec iCrc=8107 |
MMX - interp- h-round0 0.051 usec crc32=0x115381ba |
1579 |
MMX - round1 0.106 usec iCrc=8100 |
MMX - round1 0.053 usec crc32=0x2b1f528f |
1580 |
MMX - interp- v-round0 0.106 usec iCrc=8108 |
MMX - interp- v-round0 0.048 usec crc32=0x423cdcc7 |
1581 |
MMX - round1 0.106 usec iCrc=8105 |
MMX - round1 0.048 usec crc32=0x42202efe |
1582 |
MMX - interp-hv-round0 0.145 usec iCrc=8112 |
MMX - interp-hv-round0 0.074 usec crc32=0xd198d387 |
1583 |
MMX - round1 0.145 usec iCrc=8103 |
MMX - round1 0.073 usec crc32=0x9ecfd921 |
1584 |
--- |
--- |
1585 |
MMXEXT - interp- h-round0 0.028 usec iCrc=8107 |
MMXEXT - interp- h-round0 0.020 usec crc32=0x115381ba |
1586 |
MMXEXT - round1 0.041 usec iCrc=8100 |
MMXEXT - round1 0.025 usec crc32=0x2b1f528f |
1587 |
MMXEXT - interp- v-round0 0.027 usec iCrc=8108 |
MMXEXT - interp- v-round0 0.016 usec crc32=0x423cdcc7 |
1588 |
MMXEXT - round1 0.041 usec iCrc=8105 |
MMXEXT - round1 0.024 usec crc32=0x42202efe |
1589 |
MMXEXT - interp-hv-round0 0.066 usec iCrc=8112 |
MMXEXT - interp-hv-round0 0.037 usec crc32=0xd198d387 |
1590 |
MMXEXT - round1 0.065 usec iCrc=8103 |
MMXEXT - round1 0.037 usec crc32=0x9ecfd921 |
1591 |
--- |
--- |
1592 |
SSE2 - interp- h-round0 0.109 usec iCrc=8107 |
3DNOW - interp- h-round0 0.020 usec crc32=0x115381ba |
1593 |
SSE2 - round1 0.105 usec iCrc=8100 |
3DNOW - round1 0.029 usec crc32=0x2b1f528f |
1594 |
SSE2 - interp- v-round0 0.106 usec iCrc=8108 |
3DNOW - interp- v-round0 0.016 usec crc32=0x423cdcc7 |
1595 |
SSE2 - round1 0.109 usec iCrc=8105 |
3DNOW - round1 0.024 usec crc32=0x42202efe |
1596 |
SSE2 - interp-hv-round0 0.145 usec iCrc=8112 |
3DNOW - interp-hv-round0 0.038 usec crc32=0xd198d387 |
1597 |
SSE2 - round1 0.145 usec iCrc=8103 |
3DNOW - round1 0.039 usec crc32=0x9ecfd921 |
1598 |
|
--- |
1599 |
|
3DNOWE - interp- h-round0 0.020 usec crc32=0x115381ba |
1600 |
|
3DNOWE - round1 0.024 usec crc32=0x2b1f528f |
1601 |
|
3DNOWE - interp- v-round0 0.016 usec crc32=0x423cdcc7 |
1602 |
|
3DNOWE - round1 0.021 usec crc32=0x42202efe |
1603 |
|
3DNOWE - interp-hv-round0 0.037 usec crc32=0xd198d387 |
1604 |
|
3DNOWE - round1 0.036 usec crc32=0x9ecfd921 |
1605 |
--- |
--- |
|
3DNOW - skipped... |
|
|
3DNOWE - skipped... |
|
1606 |
|
|
1607 |
====== test SAD ====== |
====== test SAD ====== |
1608 |
PLAINC - sad8 0.251 usec sad=3776 |
PLAINC - sad8 0.505 usec sad=3776 |
1609 |
PLAINC - sad16 1.601 usec sad=27214 |
PLAINC - sad16 1.941 usec sad=27214 |
1610 |
PLAINC - sad16bi 2.371 usec sad=26274 |
PLAINC - sad16bi 4.925 usec sad=26274 |
1611 |
PLAINC - dev16 1.564 usec sad=3344 |
PLAINC - dev16 4.254 usec sad=3344 |
1612 |
--- |
--- |
1613 |
MMX - sad8 0.057 usec sad=3776 |
MMX - sad8 0.036 usec sad=3776 |
1614 |
MMX - sad16 0.182 usec sad=27214 |
MMX - sad16 0.107 usec sad=27214 |
1615 |
MMX - sad16bi 2.462 usec sad=26274 |
MMX - sad16bi 0.259 usec sad=26274 |
1616 |
MMX - dev16 0.311 usec sad=3344 |
MMX - dev16 0.187 usec sad=3344 |
1617 |
--- |
--- |
1618 |
MMXEXT - sad8 0.036 usec sad=3776 |
MMXEXT - sad8 0.016 usec sad=3776 |
1619 |
MMXEXT - sad16 0.109 usec sad=27214 |
MMXEXT - sad16 0.050 usec sad=27214 |
1620 |
MMXEXT - sad16bi 0.143 usec sad=26274 |
MMXEXT - sad16bi 0.060 usec sad=26274 |
1621 |
MMXEXT - dev16 0.192 usec sad=3344 |
MMXEXT - dev16 0.086 usec sad=3344 |
1622 |
--- |
--- |
1623 |
SSE2 - sad8 0.057 usec sad=3776 |
3DNOW - sad8 0.506 usec sad=3776 |
1624 |
SSE2 - sad16 0.179 usec sad=27214 |
3DNOW - sad16 1.954 usec sad=27214 |
1625 |
SSE2 - sad16bi 2.456 usec sad=26274 |
3DNOW - sad16bi 0.119 usec sad=26274 |
1626 |
SSE2 - dev16 0.321 usec sad=3344 |
3DNOW - dev16 4.252 usec sad=3344 |
1627 |
|
--- |
1628 |
|
3DNOWE - sad8 0.017 usec sad=3776 |
1629 |
|
3DNOWE - sad16 0.038 usec sad=27214 |
1630 |
|
3DNOWE - sad16bi 0.052 usec sad=26274 |
1631 |
|
3DNOWE - dev16 0.067 usec sad=3344 |
1632 |
--- |
--- |
|
3DNOW - skipped... |
|
|
3DNOWE - skipped... |
|
1633 |
|
|
1634 |
=== test transfer === |
=== test transfer === |
1635 |
PLAINC - 8to16 0.151 usec crc=28288 |
PLAINC - 8to16 0.603 usec crc32=0x115814bb |
1636 |
PLAINC - 16to8 1.113 usec crc=28288 |
PLAINC - 16to8 1.077 usec crc32=0xee7ccbb4 |
1637 |
PLAINC - 8to8 0.043 usec crc=20352 |
PLAINC - 8to8 0.679 usec crc32=0xd37b3295 |
1638 |
PLAINC - 16to8add 1.069 usec crc=25536 |
PLAINC - 16to8add 1.341 usec crc32=0xdd817bf4 |
1639 |
PLAINC - 8to16sub 0.631 usec crc1=28064 crc2=16256 |
PLAINC - 8to16sub 1.566 usec crc32(1)=0xa1e07163 crc32(2)=0xd86c5d23 |
1640 |
PLAINC - 8to16sub2 0.597 usec crc=20384 |
PLAINC - 8to16sub2 2.206 usec crc32=0x99b6c4c7 |
1641 |
--- |
--- |
1642 |
MMX - 8to16 0.032 usec crc=28288 |
MMX - 8to16 -0.025 usec crc32=0x115814bb |
1643 |
MMX - 16to8 0.024 usec crc=28288 |
MMX - 16to8 -0.049 usec crc32=0xee7ccbb4 |
1644 |
MMX - 8to8 0.020 usec crc=20352 |
MMX - 8to8 0.014 usec crc32=0xd37b3295 |
1645 |
MMX - 16to8add 0.043 usec crc=25536 |
MMX - 16to8add 0.011 usec crc32=0xdd817bf4 |
1646 |
MMX - 8to16sub 0.066 usec crc1=28064 crc2=16256 |
MMX - 8to16sub 0.108 usec crc32(1)=0xa1e07163 crc32(2)=0xd86c5d23 |
1647 |
MMX - 8to16sub2 0.111 usec crc=20384 |
MMX - 8to16sub2 0.164 usec crc32=0x99b6c4c7 |
1648 |
|
--- |
1649 |
|
MMXEXT - 8to16 -0.054 usec crc32=0x115814bb |
1650 |
|
MMXEXT - 16to8 0.010 usec crc32=0xee7ccbb4 |
1651 |
|
MMXEXT - 8to8 0.015 usec crc32=0xd37b3295 |
1652 |
|
MMXEXT - 16to8add 0.008 usec crc32=0xdd817bf4 |
1653 |
|
MMXEXT - 8to16sub 0.263 usec crc32(1)=0xa1e07163 crc32(2)=0xd86c5d23 |
1654 |
|
MMXEXT - 8to16sub2 0.178 usec crc32=0x99b6c4c7 |
1655 |
|
--- |
1656 |
|
3DNOW - 8to16 0.666 usec crc32=0x115814bb |
1657 |
|
3DNOW - 16to8 1.078 usec crc32=0xee7ccbb4 |
1658 |
|
3DNOW - 8to8 0.665 usec crc32=0xd37b3295 |
1659 |
|
3DNOW - 16to8add 1.365 usec crc32=0xdd817bf4 |
1660 |
|
3DNOW - 8to16sub 1.356 usec crc32(1)=0xa1e07163 crc32(2)=0xd86c5d23 |
1661 |
|
3DNOW - 8to16sub2 2.098 usec crc32=0x99b6c4c7 |
1662 |
|
--- |
1663 |
|
3DNOWE - 8to16 -0.024 usec crc32=0x115814bb |
1664 |
|
3DNOWE - 16to8 0.010 usec crc32=0xee7ccbb4 |
1665 |
|
3DNOWE - 8to8 0.014 usec crc32=0xd37b3295 |
1666 |
|
3DNOWE - 16to8add 0.016 usec crc32=0xdd817bf4 |
1667 |
|
3DNOWE - 8to16sub -0.000 usec crc32(1)=0xa1e07163 crc32(2)=0xd86c5d23 |
1668 |
|
3DNOWE - 8to16sub2 -0.031 usec crc32=0x99b6c4c7 |
1669 |
--- |
--- |
1670 |
|
|
1671 |
===== test quant ===== |
===== test quant ===== |
1672 |
PLAINC - quant4_intra 74.248 usec crc=29809 |
PLAINC - quant_mpeg_intra 98.631 usec crc32=0xfd6a21a4 |
1673 |
PLAINC - quant4_inter 70.850 usec crc=12574 |
PLAINC - quant_mpeg_inter 104.876 usec crc32=0xf6de7757 |
1674 |
PLAINC - dequant4_intra 40.628 usec crc=24052 |
PLAINC - dequant_mpeg_intra 50.285 usec crc32=0x2def7bc7 |
1675 |
PLAINC - dequant4_inter 45.691 usec crc=63847 |
PLAINC - dequant_mpeg_inter 58.316 usec crc32=0xd878c722 |
1676 |
PLAINC - quant_intra 43.357 usec crc=25662 |
PLAINC - quant_h263_intra 33.803 usec crc32=0x2eba9d43 |
1677 |
PLAINC - quant_inter 33.410 usec crc=23972 |
PLAINC - quant_h263_inter 45.411 usec crc32=0xbd315a7e |
1678 |
PLAINC - dequant_intra 36.384 usec crc=49900 |
PLAINC - dequant_h263_intra 39.302 usec crc32=0x9841212a |
1679 |
PLAINC - dequant_inter 48.930 usec crc=48899 |
PLAINC - dequant_h263_inter 44.124 usec crc32=0xe7df8fba |
1680 |
--- |
--- |
1681 |
MMX - quant4_intra 7.445 usec crc=3459 |
MMX - quant_mpeg_intra 4.273 usec crc32=0xdacabdb6 | ERROR |
1682 |
*** CRC ERROR! *** |
MMX - quant_mpeg_inter 3.576 usec crc32=0x72883ab6 | ERROR |
1683 |
MMX - quant4_inter 5.384 usec crc=51072 |
MMX - dequant_mpeg_intra 3.793 usec crc32=0x2def7bc7 |
1684 |
*** CRC ERROR! *** |
MMX - dequant_mpeg_inter 4.808 usec crc32=0xd878c722 |
1685 |
MMX - dequant4_intra 5.515 usec crc=24052 |
MMX - quant_h263_intra 2.881 usec crc32=0x2eba9d43 |
1686 |
MMX - dequant4_inter 7.745 usec crc=63847 |
MMX - quant_h263_inter 2.550 usec crc32=0xbd315a7e |
1687 |
MMX - quant_intra 4.661 usec crc=25662 |
MMX - dequant_h263_intra 2.974 usec crc32=0x9841212a |
1688 |
MMX - quant_inter 4.406 usec crc=23972 |
MMX - dequant_h263_inter 2.906 usec crc32=0xe7df8fba |
1689 |
MMX - dequant_intra 4.928 usec crc=49900 |
--- |
1690 |
MMX - dequant_inter 4.532 usec crc=48899 |
MMXEXT - quant_mpeg_intra 4.221 usec crc32=0xfd6a21a4 |
1691 |
|
MMXEXT - quant_mpeg_inter 4.339 usec crc32=0xf6de7757 |
1692 |
|
MMXEXT - dequant_mpeg_intra 3.802 usec crc32=0x2def7bc7 |
1693 |
|
MMXEXT - dequant_mpeg_inter 4.821 usec crc32=0xd878c722 |
1694 |
|
MMXEXT - quant_h263_intra 2.884 usec crc32=0x2eba9d43 |
1695 |
|
MMXEXT - quant_h263_inter 2.554 usec crc32=0xbd315a7e |
1696 |
|
MMXEXT - dequant_h263_intra 2.728 usec crc32=0x9841212a |
1697 |
|
MMXEXT - dequant_h263_inter 2.611 usec crc32=0xe7df8fba |
1698 |
|
--- |
1699 |
|
3DNOW - quant_mpeg_intra 98.512 usec crc32=0xfd6a21a4 |
1700 |
|
3DNOW - quant_mpeg_inter 104.873 usec crc32=0xf6de7757 |
1701 |
|
3DNOW - dequant_mpeg_intra 50.219 usec crc32=0x2def7bc7 |
1702 |
|
3DNOW - dequant_mpeg_inter 58.254 usec crc32=0xd878c722 |
1703 |
|
3DNOW - quant_h263_intra 33.778 usec crc32=0x2eba9d43 |
1704 |
|
3DNOW - quant_h263_inter 41.998 usec crc32=0xbd315a7e |
1705 |
|
3DNOW - dequant_h263_intra 39.344 usec crc32=0x9841212a |
1706 |
|
3DNOW - dequant_h263_inter 43.607 usec crc32=0xe7df8fba |
1707 |
|
--- |
1708 |
|
3DNOWE - quant_mpeg_intra 98.490 usec crc32=0xfd6a21a4 |
1709 |
|
3DNOWE - quant_mpeg_inter 104.889 usec crc32=0xf6de7757 |
1710 |
|
3DNOWE - dequant_mpeg_intra 3.277 usec crc32=0x2def7bc7 |
1711 |
|
3DNOWE - dequant_mpeg_inter 4.485 usec crc32=0xd878c722 |
1712 |
|
3DNOWE - quant_h263_intra 1.882 usec crc32=0x2eba9d43 |
1713 |
|
3DNOWE - quant_h263_inter 2.246 usec crc32=0xbd315a7e |
1714 |
|
3DNOWE - dequant_h263_intra 3.457 usec crc32=0x9841212a |
1715 |
|
3DNOWE - dequant_h263_inter 3.275 usec crc32=0xe7df8fba |
1716 |
--- |
--- |
1717 |
|
|
1718 |
===== test cbp ===== |
===== test cbp ===== |
1719 |
PLAINC - calc_cbp#1 0.371 usec cbp=0x15 |
PLAINC - calc_cbp#1 0.168 usec cbp=0x15 |
1720 |
PLAINC - calc_cbp#2 0.432 usec cbp=0x38 |
PLAINC - calc_cbp#2 0.168 usec cbp=0x38 |
1721 |
PLAINC - calc_cbp#3 0.339 usec cbp=0xf |
PLAINC - calc_cbp#3 0.157 usec cbp=0x0f |
1722 |
PLAINC - calc_cbp#4 0.506 usec cbp=0x5 |
PLAINC - calc_cbp#4 0.235 usec cbp=0x05 |
1723 |
--- |
--- |
1724 |
MMX - calc_cbp#1 0.136 usec cbp=0x15 |
MMX - calc_cbp#1 0.070 usec cbp=0x15 |
1725 |
MMX - calc_cbp#2 0.134 usec cbp=0x38 |
MMX - calc_cbp#2 0.062 usec cbp=0x38 |
1726 |
MMX - calc_cbp#3 0.138 usec cbp=0xf |
MMX - calc_cbp#3 0.062 usec cbp=0x0f |
1727 |
MMX - calc_cbp#4 0.135 usec cbp=0x5 |
MMX - calc_cbp#4 0.061 usec cbp=0x05 |
1728 |
--- |
--- |
1729 |
SSE2 - calc_cbp#1 0.136 usec cbp=0x15 |
MMXEXT - calc_cbp#1 0.062 usec cbp=0x15 |
1730 |
SSE2 - calc_cbp#2 0.133 usec cbp=0x38 |
MMXEXT - calc_cbp#2 0.061 usec cbp=0x38 |
1731 |
SSE2 - calc_cbp#3 0.133 usec cbp=0xf |
MMXEXT - calc_cbp#3 0.061 usec cbp=0x0f |
1732 |
SSE2 - calc_cbp#4 0.141 usec cbp=0x5 |
MMXEXT - calc_cbp#4 0.061 usec cbp=0x05 |
1733 |
|
--- |
1734 |
|
3DNOW - calc_cbp#1 0.168 usec cbp=0x15 |
1735 |
|
3DNOW - calc_cbp#2 0.168 usec cbp=0x38 |
1736 |
|
3DNOW - calc_cbp#3 0.157 usec cbp=0x0f |
1737 |
|
3DNOW - calc_cbp#4 0.238 usec cbp=0x05 |
1738 |
|
--- |
1739 |
|
3DNOWE - calc_cbp#1 0.049 usec cbp=0x15 |
1740 |
|
3DNOWE - calc_cbp#2 0.049 usec cbp=0x38 |
1741 |
|
3DNOWE - calc_cbp#3 0.049 usec cbp=0x0f |
1742 |
|
3DNOWE - calc_cbp#4 0.049 usec cbp=0x05 |
1743 |
--- |
--- |
1744 |
|
|
1745 |
*/ |
|
1746 |
|
NB: If a function isn't optimised for a specific set of intructions, |
1747 |
|
a C function is used instead. So don't panic if some functions |
1748 |
|
may appear to be slow. |
1749 |
|
|
1750 |
|
NB: MMX mpeg4 quantization is known to have very small errors (+/-1 magnitude) |
1751 |
|
for 1 or 2 coefficients a block. This is mainly caused by the fact the unit |
1752 |
|
test goes far behind the usual limits of real encoding. Please do not report |
1753 |
|
this error to the developers |
1754 |
|
|
1755 |
|
*********************************************************************/ |