Mercurial > mplayer.hg
comparison libswscale/rgb2rgb_template.c @ 22995:70d7c6206f33
skip MMX code in rgb32to15 if the size of the input is smaller than the
size of the units the MMX code processes
| author | ivo |
|---|---|
| date | Wed, 18 Apr 2007 09:24:49 +0000 |
| parents | ac77d9ef8c83 |
| children | 2a60af5e78a7 |
comparison
equal
deleted
inserted
replaced
| 22994:ac77d9ef8c83 | 22995:70d7c6206f33 |
|---|---|
| 511 #if 1 //is faster only if multiplies are reasonable fast (FIXME figure out on which cpus this is faster, on Athlon its slightly faster) | 511 #if 1 //is faster only if multiplies are reasonable fast (FIXME figure out on which cpus this is faster, on Athlon its slightly faster) |
| 512 asm volatile( | 512 asm volatile( |
| 513 "movq %3, %%mm5 \n\t" | 513 "movq %3, %%mm5 \n\t" |
| 514 "movq %4, %%mm6 \n\t" | 514 "movq %4, %%mm6 \n\t" |
| 515 "movq %5, %%mm7 \n\t" | 515 "movq %5, %%mm7 \n\t" |
| 516 "jmp 2f \n\t" | |
| 516 ASMALIGN(4) | 517 ASMALIGN(4) |
| 517 "1: \n\t" | 518 "1: \n\t" |
| 518 PREFETCH" 32(%1) \n\t" | 519 PREFETCH" 32(%1) \n\t" |
| 519 "movd (%1), %%mm0 \n\t" | 520 "movd (%1), %%mm0 \n\t" |
| 520 "movd 4(%1), %%mm3 \n\t" | 521 "movd 4(%1), %%mm3 \n\t" |
| 534 "pslld $10, %%mm3 \n\t" | 535 "pslld $10, %%mm3 \n\t" |
| 535 "por %%mm3, %%mm0 \n\t" | 536 "por %%mm3, %%mm0 \n\t" |
| 536 MOVNTQ" %%mm0, (%0) \n\t" | 537 MOVNTQ" %%mm0, (%0) \n\t" |
| 537 "add $16, %1 \n\t" | 538 "add $16, %1 \n\t" |
| 538 "add $8, %0 \n\t" | 539 "add $8, %0 \n\t" |
| 540 "2: \n\t" | |
| 539 "cmp %2, %1 \n\t" | 541 "cmp %2, %1 \n\t" |
| 540 " jb 1b \n\t" | 542 " jb 1b \n\t" |
| 541 : "+r" (d), "+r"(s) | 543 : "+r" (d), "+r"(s) |
| 542 : "r" (mm_end), "m" (mask3215g), "m" (mask3216br), "m" (mul3215) | 544 : "r" (mm_end), "m" (mask3215g), "m" (mask3216br), "m" (mul3215) |
| 543 ); | 545 ); |