diff options
| author | Michael Niedermayer <michaelni@gmx.at> | 2012-10-31 14:26:11 +0100 |
|---|---|---|
| committer | Michael Niedermayer <michaelni@gmx.at> | 2012-10-31 14:27:11 +0100 |
| commit | 84833b028bf3fa08d09939330989df8d8820cd59 (patch) | |
| tree | 532812ffae1097453064a82f86b10931fb55cc25 /libavcodec/x86/fmtconvert.asm | |
| parent | 28c0678eb764b8d10faf1a988d2f0e2dc79c96a5 (diff) | |
| parent | 5707af8d0b9c5a7e8a742a11798e406f99d78cea (diff) | |
Merge remote-tracking branch 'qatar/master'
* qatar/master:
aacdec: Drop some unused function arguments
x86: Fix assembly with NASM
x86: fmtconvert: Refactor cvtps2pi emulation through cpuflags
Conflicts:
libavcodec/aacdec.c
Merged-by: Michael Niedermayer <michaelni@gmx.at>
Diffstat (limited to 'libavcodec/x86/fmtconvert.asm')
| -rw-r--r-- | libavcodec/x86/fmtconvert.asm | 52 |
1 files changed, 26 insertions, 26 deletions
diff --git a/libavcodec/x86/fmtconvert.asm b/libavcodec/x86/fmtconvert.asm index f6b2c943e1..8313d261d9 100644 --- a/libavcodec/x86/fmtconvert.asm +++ b/libavcodec/x86/fmtconvert.asm @@ -23,6 +23,14 @@ SECTION_TEXT +%macro CVTPS2PI 2 +%if cpuflag(sse) + cvtps2pi %1, %2 +%elif cpuflag(3dnow) + pf2id %1, %2 +%endif +%endmacro + ;--------------------------------------------------------------------------------- ; void int32_to_float_fmul_scalar(float *dst, const int *src, float mul, int len); ;--------------------------------------------------------------------------------- @@ -88,10 +96,10 @@ cglobal float_to_int16, 3, 3, %1, dst, src, len packssdw m0, m1 mova [dstq+lenq], m0 %else - cvtps2pi m0, [srcq+2*lenq ] - cvtps2pi m1, [srcq+2*lenq+ 8] - cvtps2pi m2, [srcq+2*lenq+16] - cvtps2pi m3, [srcq+2*lenq+24] + CVTPS2PI m0, [srcq+2*lenq ] + CVTPS2PI m1, [srcq+2*lenq+ 8] + CVTPS2PI m2, [srcq+2*lenq+16] + CVTPS2PI m3, [srcq+2*lenq+24] packssdw m0, m1 packssdw m2, m3 mova [dstq+lenq ], m0 @@ -109,10 +117,8 @@ INIT_XMM sse2 FLOAT_TO_INT16 2 INIT_MMX sse FLOAT_TO_INT16 0 -%define cvtps2pi pf2id INIT_MMX 3dnow FLOAT_TO_INT16 0 -%undef cvtps2pi ;------------------------------------------------------------------------------ ; void ff_float_to_int16_step(int16_t *dst, const float *src, long len, long step); @@ -150,10 +156,10 @@ cglobal float_to_int16_step, 4, 7, %1, dst, src, len, step, step3, v1, v2 mov [dstq+step3q*2], v2w lea dstq, [dstq+stepq*8] %else - cvtps2pi m0, [srcq+2*lenq ] - cvtps2pi m1, [srcq+2*lenq+ 8] - cvtps2pi m2, [srcq+2*lenq+16] - cvtps2pi m3, [srcq+2*lenq+24] + CVTPS2PI m0, [srcq+2*lenq ] + CVTPS2PI m1, [srcq+2*lenq+ 8] + CVTPS2PI m2, [srcq+2*lenq+16] + CVTPS2PI m3, [srcq+2*lenq+24] packssdw m0, m1 packssdw m2, m3 movd v1d, m0 @@ -189,10 +195,8 @@ INIT_XMM sse2 FLOAT_TO_INT16_STEP 2 INIT_MMX sse FLOAT_TO_INT16_STEP 0 -%define cvtps2pi pf2id INIT_MMX 3dnow FLOAT_TO_INT16_STEP 0 -%undef cvtps2pi ;------------------------------------------------------------------------------- ; void ff_float_to_int16_interleave2(int16_t *dst, const float **src, long len); @@ -215,10 +219,10 @@ cglobal float_to_int16_interleave2, 3, 4, 2, dst, src0, src1, len punpcklwd m0, m1 mova [dstq+lenq], m0 %else - cvtps2pi m0, [src0q+lenq ] - cvtps2pi m1, [src0q+lenq+8] - cvtps2pi m2, [src1q+lenq ] - cvtps2pi m3, [src1q+lenq+8] + CVTPS2PI m0, [src0q+lenq ] + CVTPS2PI m1, [src0q+lenq+8] + CVTPS2PI m2, [src1q+lenq ] + CVTPS2PI m3, [src1q+lenq+8] packssdw m0, m1 packssdw m2, m3 mova m1, m0 @@ -236,9 +240,7 @@ cglobal float_to_int16_interleave2, 3, 4, 2, dst, src0, src1, len %endmacro INIT_MMX 3dnow -%define cvtps2pi pf2id FLOAT_TO_INT16_INTERLEAVE2 -%undef cvtps2pi INIT_MMX sse FLOAT_TO_INT16_INTERLEAVE2 INIT_XMM sse2 @@ -274,12 +276,12 @@ cglobal float_to_int16_interleave6, 2, 8, 0, dst, src, src1, src2, src3, src4, s sub src4q, srcq sub src5q, srcq .loop: - cvtps2pi mm0, [srcq] - cvtps2pi mm1, [srcq+src1q] - cvtps2pi mm2, [srcq+src2q] - cvtps2pi mm3, [srcq+src3q] - cvtps2pi mm4, [srcq+src4q] - cvtps2pi mm5, [srcq+src5q] + CVTPS2PI mm0, [srcq] + CVTPS2PI mm1, [srcq+src1q] + CVTPS2PI mm2, [srcq+src2q] + CVTPS2PI mm3, [srcq+src3q] + CVTPS2PI mm4, [srcq+src4q] + CVTPS2PI mm5, [srcq+src5q] packssdw mm0, mm3 packssdw mm1, mm4 packssdw mm2, mm5 @@ -306,13 +308,11 @@ INIT_MMX sse %define pswapd PSWAPD_SSE FLOAT_TO_INT16_INTERLEAVE6 INIT_MMX 3dnow -%define cvtps2pi pf2id %define pswapd PSWAPD_3DNOW FLOAT_TO_INT16_INTERLEAVE6 %undef pswapd INIT_MMX 3dnowext FLOAT_TO_INT16_INTERLEAVE6 -%undef cvtps2pi ;----------------------------------------------------------------------------- ; void ff_float_interleave6(float *dst, const float **src, unsigned int len); |