diff options
| author | Nils Wallménius <nils@rockbox.org> | 2010-07-28 23:25:45 +0000 |
|---|---|---|
| committer | Nils Wallménius <nils@rockbox.org> | 2010-07-28 23:25:45 +0000 |
| commit | 79ae416b9512ce28fec840a28e9b3594b310ec09 (patch) | |
| tree | 896fccf35d738b635f3d0ebd4560c92f60e00033 /apps/codecs/libwmapro/wmapro_math.h | |
| parent | 85f84060c4566d2a9e0867a1203751d77899547c (diff) | |
| download | rockbox-79ae416b9512ce28fec840a28e9b3594b310ec09.zip rockbox-79ae416b9512ce28fec840a28e9b3594b310ec09.tar.gz rockbox-79ae416b9512ce28fec840a28e9b3594b310ec09.tar.bz2 rockbox-79ae416b9512ce28fec840a28e9b3594b310ec09.tar.xz | |
libwmapro: coldfire asm for vector_fixmul_scalar, 1% speedup
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@27607 a1c6a512-1295-4272-9138-f99709370657
Diffstat (limited to 'apps/codecs/libwmapro/wmapro_math.h')
| -rw-r--r-- | apps/codecs/libwmapro/wmapro_math.h | 32 |
1 files changed, 32 insertions, 0 deletions
diff --git a/apps/codecs/libwmapro/wmapro_math.h b/apps/codecs/libwmapro/wmapro_math.h index 3672c01..06df737 100644 --- a/apps/codecs/libwmapro/wmapro_math.h +++ b/apps/codecs/libwmapro/wmapro_math.h @@ -238,6 +238,38 @@ static inline void vector_fixmul_window(int32_t *dst, const int32_t *src0, : [dst]"+r"(dst), [src]"+r"(src) \ : [mul]"r"(mul) \ : "r0", "r1", "r2", "r3", "r4", "r5", "memory"); +#elif defined (CPU_COLDFIRE) + #define VECT_MUL_SCALAR_KERNEL(dst, src, mul) \ + int32_t tmp; \ + asm volatile ( \ + "movem.l (%[src]), %%d0-%%d3 \n\t" \ + "mac.l %[mul], %%d0, %%acc0 \n\t" \ + "mac.l %[mul], %%d1, %%acc1 \n\t" \ + "mac.l %[mul], %%d2, %%acc2 \n\t" \ + "mac.l %[mul], %%d3, %%acc3 \n\t" \ + "move.l %%accext01, %[tmp] \n\t" \ + "movclr.l %%acc0, %%d0 \n\t" \ + "movclr.l %%acc1, %%d1 \n\t" \ + "lsl.l #7, %%d0 \n\t" \ + "move.b %[tmp], %%d0 \n\t" \ + "swap %[tmp] \n\t" \ + "lsl.l #7, %%d1 \n\t" \ + "move.b %[tmp], %%d1 \n\t" \ + "move.l %%accext23, %[tmp] \n\t" \ + "movclr.l %%acc2, %%d2 \n\t" \ + "movclr.l %%acc3, %%d3 \n\t" \ + "lsl.l #7, %%d2 \n\t" \ + "move.b %[tmp], %%d2 \n\t" \ + "swap %[tmp] \n\t" \ + "lsl.l #7, %%d3 \n\t" \ + "move.b %[tmp], %%d3 \n\t" \ + "movem.l %%d0-%%d3, (%[dst]) \n\t" \ + "lea.l (4*4, %[src]), %[src]\n\t" \ + "lea.l (4*4, %[dst]), %[dst]\n\t" \ + : [dst] "+a" (dst), [src] "+a" (src),\ + [tmp] "=d" (tmp) \ + : [mul] "r" (mul) \ + : "d0", "d1", "d2", "d3", "memory", "cc"); #else #define VECT_MUL_SCALAR_KERNEL(dst, src, mul) \ dst[i ] = fixmul24(src[i ], mul); \ |