diff options
| author | Michael Niedermayer <michaelni@gmx.at> | 2013-01-23 14:04:50 +0100 |
|---|---|---|
| committer | Michael Niedermayer <michaelni@gmx.at> | 2013-01-23 14:04:50 +0100 |
| commit | 6e6e1708984e45881b9a5d4e26c3e7de852c54d5 (patch) | |
| tree | 5e04d38f8e152faf98921843ca5e4530cbdc46a4 /libavutil/x86/float_dsp.asm | |
| parent | b1b870fbd7185bffbe27c5918001b40a8ff8b920 (diff) | |
| parent | 42d324694883cdf1fff1612ac70fa403692a1ad4 (diff) | |
| download | ffmpeg-6e6e1708984e45881b9a5d4e26c3e7de852c54d5.tar.gz | |
Merge commit '42d324694883cdf1fff1612ac70fa403692a1ad4'
* commit '42d324694883cdf1fff1612ac70fa403692a1ad4':
floatdsp: move vector_fmul_reverse from dsputil to avfloatdsp.
Conflicts:
libavcodec/arm/dsputil_init_vfp.c
libavcodec/arm/dsputil_vfp.S
libavcodec/dsputil.c
libavcodec/ppc/float_altivec.c
libavcodec/x86/dsputil.asm
libavutil/x86/float_dsp.asm
Merged-by: Michael Niedermayer <michaelni@gmx.at>
Diffstat (limited to 'libavutil/x86/float_dsp.asm')
| -rw-r--r-- | libavutil/x86/float_dsp.asm | 39 |
1 files changed, 39 insertions, 0 deletions
diff --git a/libavutil/x86/float_dsp.asm b/libavutil/x86/float_dsp.asm index f69fc6b00a..3e5e91ad07 100644 --- a/libavutil/x86/float_dsp.asm +++ b/libavutil/x86/float_dsp.asm @@ -198,3 +198,42 @@ VECTOR_FMUL_ADD INIT_YMM avx VECTOR_FMUL_ADD %endif + +;----------------------------------------------------------------------------- +; void vector_fmul_reverse(float *dst, const float *src0, const float *src1, +; int len) +;----------------------------------------------------------------------------- +%macro VECTOR_FMUL_REVERSE 0 +cglobal vector_fmul_reverse, 4,4,2, dst, src0, src1, len + lea lenq, [lend*4 - 2*mmsize] +ALIGN 16 +.loop: +%if cpuflag(avx) + vmovaps xmm0, [src1q + 16] + vinsertf128 m0, m0, [src1q], 1 + vshufps m0, m0, m0, q0123 + vmovaps xmm1, [src1q + mmsize + 16] + vinsertf128 m1, m1, [src1q + mmsize], 1 + vshufps m1, m1, m1, q0123 +%else + mova m0, [src1q] + mova m1, [src1q + mmsize] + shufps m0, m0, q0123 + shufps m1, m1, q0123 +%endif + mulps m0, m0, [src0q + lenq + mmsize] + mulps m1, m1, [src0q + lenq] + mova [dstq + lenq + mmsize], m0 + mova [dstq + lenq], m1 + add src1q, 2*mmsize + sub lenq, 2*mmsize + jge .loop + REP_RET +%endmacro + +INIT_XMM sse +VECTOR_FMUL_REVERSE +%if HAVE_AVX_EXTERNAL +INIT_YMM avx +VECTOR_FMUL_REVERSE +%endif
\ No newline at end of file |