19 |
* along with this program; if not, write to the Free Software |
* along with this program; if not, write to the Free Software |
20 |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
21 |
* |
* |
22 |
* $Id: xvid_bench.c,v 1.16 2005-04-18 08:31:42 Skal Exp $ |
* $Id: xvid_bench.c,v 1.17 2005-05-17 15:40:11 Skal Exp $ |
23 |
* |
* |
24 |
****************************************************************************/ |
****************************************************************************/ |
25 |
|
|
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 |
259 |
}; |
}; |
260 |
|
|
261 |
uint32_t |
uint32_t |
262 |
calc_crc(uint8_t *mem, int len, uint32_t initial) |
calc_crc(uint8_t *mem, int len, uint32_t crc) |
263 |
{ |
{ |
|
|
|
|
register unsigned int crc; |
|
|
|
|
|
crc = initial; |
|
|
|
|
264 |
while( len >= 8) { |
while( len >= 8) { |
265 |
DO8(mem, crc); |
DO8(mem, crc); |
266 |
len -= 8; |
len -= 8; |
271 |
len--; |
len--; |
272 |
} |
} |
273 |
|
|
274 |
return(crc); |
return crc; |
|
|
|
275 |
} |
} |
276 |
|
|
277 |
/********************************************************************* |
/********************************************************************* |
1314 |
* measure raw decoding speed |
* measure raw decoding speed |
1315 |
*********************************************************************/ |
*********************************************************************/ |
1316 |
|
|
1317 |
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) |
1318 |
{ |
{ |
1319 |
FILE *f = 0; |
FILE *f = 0; |
1320 |
void *dechandle = 0; |
void *dechandle = 0; |
1325 |
double t = 0.; |
double t = 0.; |
1326 |
int nb = 0; |
int nb = 0; |
1327 |
uint8_t *buf = 0; |
uint8_t *buf = 0; |
1328 |
uint8_t *rgb_out = 0; |
uint8_t *yuv_out = 0; |
1329 |
int buf_size, pos; |
int buf_size, pos; |
1330 |
uint32_t chksum = 0; |
uint32_t chksum = 0; |
1331 |
|
int bps = (width+31) & ~31; |
1332 |
|
|
1333 |
memset(&xinit, 0, sizeof(xinit)); |
memset(&xinit, 0, sizeof(xinit)); |
1334 |
xinit.cpu_flags = XVID_CPU_MMX | XVID_CPU_FORCE; |
xinit.cpu_flags = cpu_mask; |
1335 |
xinit.version = XVID_VERSION; |
xinit.version = XVID_VERSION; |
1336 |
xvid_global(NULL, 0, &xinit, NULL); |
xvid_global(NULL, 0, &xinit, NULL); |
1337 |
|
|
1341 |
xparam.version = XVID_VERSION; |
xparam.version = XVID_VERSION; |
1342 |
xerr = xvid_decore(NULL, XVID_DEC_CREATE, &xparam, NULL); |
xerr = xvid_decore(NULL, XVID_DEC_CREATE, &xparam, NULL); |
1343 |
if (xerr==XVID_ERR_FAIL) { |
if (xerr==XVID_ERR_FAIL) { |
1344 |
printf("can't init decoder (err=%d)\n", xerr); |
printf("ERROR: can't init decoder (err=%d)\n", xerr); |
1345 |
return; |
return; |
1346 |
} |
} |
1347 |
dechandle = xparam.handle; |
dechandle = xparam.handle; |
1349 |
|
|
1350 |
f = fopen(name, "rb"); |
f = fopen(name, "rb"); |
1351 |
if (f==0) { |
if (f==0) { |
1352 |
printf( "can't open file '%s'\n", name); |
printf( "ERROR: can't open file '%s'\n", name); |
1353 |
return; |
return; |
1354 |
} |
} |
1355 |
fseek(f, 0, SEEK_END); |
fseek(f, 0, SEEK_END); |
1356 |
buf_size = ftell(f); |
buf_size = ftell(f); |
1357 |
fseek(f, 0, SEEK_SET); |
fseek(f, 0, SEEK_SET); |
1358 |
if (buf_size<=0) { |
if (buf_size<=0) { |
1359 |
printf("error while stating file\n"); |
printf("ERROR: error while stating file\n"); |
1360 |
goto End; |
goto End; |
1361 |
} |
} |
|
else printf( "Input size: %d\n", buf_size); |
|
1362 |
|
|
1363 |
buf = malloc(buf_size); /* should be enuf' */ |
buf = malloc(buf_size); |
1364 |
rgb_out = calloc(4, width*height); /* <-room for _RGB24 */ |
yuv_out = calloc(1, bps*height*3/2 + 15); |
1365 |
if (buf==0 || rgb_out==0) { |
if (buf==0 || yuv_out==0) { |
1366 |
printf( "malloc failed!\n" ); |
printf( "ERROR: malloc failed!\n" ); |
1367 |
goto End; |
goto End; |
1368 |
} |
} |
1369 |
|
|
1370 |
if (fread(buf, buf_size, 1, f)!=1) { |
if (fread(buf, buf_size, 1, f)!=1) { |
1371 |
printf( "file-read failed\n" ); |
printf( "ERROR: file-read failed\n" ); |
1372 |
goto End; |
goto End; |
1373 |
} |
} |
1374 |
|
|
1376 |
pos = 0; |
pos = 0; |
1377 |
t = -gettime_usec(); |
t = -gettime_usec(); |
1378 |
while(1) { |
while(1) { |
1379 |
|
int y; |
1380 |
|
|
1381 |
memset(&xframe, 0, sizeof(xframe)); |
memset(&xframe, 0, sizeof(xframe)); |
1382 |
xframe.version = XVID_VERSION; |
xframe.version = XVID_VERSION; |
1383 |
xframe.bitstream = buf + pos; |
xframe.bitstream = buf + pos; |
1384 |
xframe.length = buf_size - pos; |
xframe.length = buf_size - pos; |
1385 |
xframe.output.plane[0] = rgb_out; |
xframe.output.plane[0] = (uint8_t*)(((size_t)yuv_out + 15) & ~15); |
1386 |
xframe.output.stride[0] = width; |
xframe.output.plane[1] = xframe.output.plane[0] + bps*height; |
1387 |
xframe.output.csp = XVID_CSP_BGR; |
xframe.output.plane[2] = xframe.output.plane[1] + bps/2; |
1388 |
|
xframe.output.stride[0] = bps; |
1389 |
|
xframe.output.stride[1] = bps; |
1390 |
|
xframe.output.stride[2] = bps; |
1391 |
|
xframe.output.csp = XVID_CSP_I420; |
1392 |
xerr = xvid_decore(dechandle, XVID_DEC_DECODE, &xframe, 0); |
xerr = xvid_decore(dechandle, XVID_DEC_DECODE, &xframe, 0); |
1393 |
|
if (xerr<0) { |
1394 |
|
printf("ERROR: decoding failed for frame #%d (err=%d)!\n", nb, xerr); |
1395 |
|
break; |
1396 |
|
} |
1397 |
|
else if (xerr==0) |
1398 |
|
break; |
1399 |
|
else if (verbose>0) printf("#%d %d\n", nb, xerr ); |
1400 |
|
|
1401 |
|
pos += xerr; |
1402 |
nb++; |
nb++; |
1403 |
pos += xframe.length; |
|
1404 |
if (with_chksum) { |
for(y=0; y<height/2; ++y) { |
1405 |
int k = width*height; |
chksum = calc_crc(xframe.output.plane[0] + (2*y+0)*bps, width, chksum); |
1406 |
uint32_t *ptr = (uint32_t *)rgb_out; |
chksum = calc_crc(xframe.output.plane[0] + (2*y+1)*bps, width, chksum); |
1407 |
while(k-->0) chksum += *ptr++; |
chksum = calc_crc(xframe.output.plane[1] + y*bps, width/2, chksum); |
1408 |
|
chksum = calc_crc(xframe.output.plane[2] + y*bps, width/2, chksum); |
1409 |
} |
} |
1410 |
if (pos==buf_size) |
if (pos==buf_size) |
1411 |
break; |
break; |
|
if (xerr==XVID_ERR_FAIL) { |
|
|
printf("decoding failed for frame #%d (err=%d)!\n", nb, xerr); |
|
|
break; |
|
|
} |
|
1412 |
} |
} |
1413 |
t += gettime_usec(); |
t += gettime_usec(); |
1414 |
|
if (ref_chksum==0) { |
1415 |
if (t>0.) |
if (t>0.) |
1416 |
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 ); |
1417 |
if (with_chksum) |
} |
1418 |
printf("checksum: 0x%.8x\n", chksum); |
else { |
1419 |
|
printf("FPS:%.1f Checksum: 0x%.8x Expected:0x%.8x | %s\n", |
1420 |
|
t>0. ? (float)(nb*1.e6f/t) : 0.f, chksum, ref_chksum, (chksum==ref_chksum) ? "OK" : "ERROR"); |
1421 |
|
} |
1422 |
|
|
1423 |
End: |
End: |
1424 |
if (rgb_out!=0) free(rgb_out); |
if (yuv_out!=0) free(yuv_out); |
1425 |
if (buf!=0) free(buf); |
if (buf!=0) free(buf); |
1426 |
if (dechandle!=0) { |
if (dechandle!=0) { |
1427 |
xerr= xvid_decore(dechandle, XVID_DEC_DESTROY, NULL, NULL); |
xerr= xvid_decore(dechandle, XVID_DEC_DESTROY, NULL, NULL); |
1428 |
if (xerr==XVID_ERR_FAIL) |
if (xerr==XVID_ERR_FAIL) |
1429 |
printf("destroy-decoder failed (err=%d)!\n", xerr); |
printf("ERROR: destroy-decoder failed (err=%d)!\n", xerr); |
1430 |
} |
} |
1431 |
if (f!=0) fclose(f); |
if (f!=0) fclose(f); |
1432 |
} |
} |
1593 |
* main |
* main |
1594 |
*********************************************************************/ |
*********************************************************************/ |
1595 |
|
|
1596 |
int main(int argc, char *argv[]) |
static void arg_missing(const char *opt) |
1597 |
|
{ |
1598 |
|
printf( "missing argument after option '%s'\n", opt); |
1599 |
|
exit(-1); |
1600 |
|
} |
1601 |
|
|
1602 |
|
int main(int argc, const char *argv[]) |
1603 |
|
{ |
1604 |
|
int c, what = 0; |
1605 |
|
int width, height; |
1606 |
|
uint32_t chksum = 0; |
1607 |
|
const char * test_bitstream = 0; |
1608 |
|
|
1609 |
|
cpu_mask = 0; // default => will use autodectect |
1610 |
|
for(c=1; c<argc; ++c) |
1611 |
{ |
{ |
1612 |
int what = 0; |
if (!strcmp(argv[c], "-v")) verbose++; |
1613 |
if (argc>1) what = atoi(argv[1]); |
else if (!strcmp(argv[c], "-c")) cpu_mask = 0 /* PLAIN_C */ | XVID_CPU_FORCE; |
1614 |
|
else if (!strcmp(argv[c], "-mmx")) cpu_mask = XVID_CPU_MMX | XVID_CPU_FORCE; |
1615 |
|
else if (!strcmp(argv[c], "-mmxext")) cpu_mask = XVID_CPU_MMXEXT | XVID_CPU_MMX | XVID_CPU_FORCE; |
1616 |
|
else if (!strcmp(argv[c], "-sse2")) cpu_mask = XVID_CPU_SSE2 | XVID_CPU_MMX | XVID_CPU_FORCE; |
1617 |
|
else if (!strcmp(argv[c], "-3dnow")) cpu_mask = XVID_CPU_3DNOW | XVID_CPU_FORCE; |
1618 |
|
else if (!strcmp(argv[c], "-3dnowe")) cpu_mask = XVID_CPU_3DNOW | XVID_CPU_3DNOWEXT | XVID_CPU_FORCE; |
1619 |
|
else if (!strcmp(argv[c], "-altivec")) cpu_mask = XVID_CPU_ALTIVEC | XVID_CPU_FORCE; |
1620 |
|
else if (!strcmp(argv[c], "-spd")) { |
1621 |
|
if (++c==argc) arg_missing( argv[argc-1] ); |
1622 |
|
speed_ref = atoi(argv[c]); |
1623 |
|
} |
1624 |
|
else if (argv[c][0]!='-') { |
1625 |
|
what = atoi(argv[c]); |
1626 |
|
if (what==9) { |
1627 |
|
if (c+4>argc) { |
1628 |
|
printf("usage: %s %d bitstream width height (checksum)\n", argv[0], what); |
1629 |
|
exit(-1); |
1630 |
|
} |
1631 |
|
test_bitstream = argv[++c]; |
1632 |
|
width = atoi(argv[++c]); |
1633 |
|
height = atoi(argv[++c]); |
1634 |
|
if (c+1<argc && argv[c+1][0]!='-') { |
1635 |
|
if (sscanf(argv[c+1], "0x%x", &chksum)!=1) { |
1636 |
|
printf( "can't read checksum value.\n" ); |
1637 |
|
exit(-1); |
1638 |
|
} |
1639 |
|
else c++; |
1640 |
|
} |
1641 |
|
// printf( "[%s] %dx%d (0x%.8x)\n", test_bitstream, width, height, chksum); |
1642 |
|
} |
1643 |
|
} |
1644 |
|
else { |
1645 |
|
printf( "unrecognized option '%s'\n", argv[c]); |
1646 |
|
exit(-1); |
1647 |
|
} |
1648 |
|
} |
1649 |
|
|
1650 |
|
|
1651 |
if (what==0 || what==1) test_dct(); |
if (what==0 || what==1) test_dct(); |
1652 |
if (what==0 || what==2) test_mb(); |
if (what==0 || what==2) test_mb(); |
1653 |
if (what==0 || what==3) test_sad(); |
if (what==0 || what==3) test_sad(); |
1666 |
} |
} |
1667 |
if (what==8) test_dct_saturation(-256, 255); |
if (what==8) test_dct_saturation(-256, 255); |
1668 |
|
|
1669 |
if (what==9) { |
if (test_bitstream) |
1670 |
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)); |
|
|
} |
|
|
|
|
1671 |
if (what==-1) { |
if (what==-1) { |
1672 |
test_dct_precision_diffs(); |
test_dct_precision_diffs(); |
1673 |
test_bugs1(); |
test_bugs1(); |