diff options
| author | Alex Converse <alex.converse@gmail.com> | 2010-01-22 23:07:58 +0000 |
|---|---|---|
| committer | Alex Converse <alex.converse@gmail.com> | 2010-01-22 23:07:58 +0000 |
| commit | 3deb53849e706b1ef932ff4f0c663f60275f8415 (patch) | |
| tree | 34fcb29f270b4bdb70f1ab4802d74c08695a91be /libavcodec/x86 | |
| parent | 57835fc1aeaf0b3bbc0816b123a931ea39de758b (diff) | |
Implement an sse version of scalarproduct_float().
Originally committed as revision 21386 to svn://svn.ffmpeg.org/ffmpeg/trunk
Diffstat (limited to 'libavcodec/x86')
| -rw-r--r-- | libavcodec/x86/dsputil_mmx.c | 5 | ||||
| -rw-r--r-- | libavcodec/x86/dsputil_yasm.asm | 24 |
2 files changed, 29 insertions, 0 deletions
diff --git a/libavcodec/x86/dsputil_mmx.c b/libavcodec/x86/dsputil_mmx.c index 0283fb41e5..8263ac59d4 100644 --- a/libavcodec/x86/dsputil_mmx.c +++ b/libavcodec/x86/dsputil_mmx.c @@ -2510,6 +2510,8 @@ void ff_snow_inner_add_yblock_mmx(const uint8_t *obmc, const int obmc_stride, ui int src_x, int src_y, int src_stride, slice_buffer * sb, int add, uint8_t * dst8); +float ff_scalarproduct_float_sse(const float *v1, const float *v2, int order); + void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx) { mm_flags = mm_support(); @@ -2965,6 +2967,9 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx) c->vector_clipf = vector_clipf_sse; c->float_to_int16 = float_to_int16_sse; c->float_to_int16_interleave = float_to_int16_interleave_sse; +#if HAVE_YASM + c->scalarproduct_float = ff_scalarproduct_float_sse; +#endif } if(mm_flags & FF_MM_3DNOW) c->vector_fmul_add = vector_fmul_add_3dnow; // faster than sse diff --git a/libavcodec/x86/dsputil_yasm.asm b/libavcodec/x86/dsputil_yasm.asm index 023fc4d3bd..e2478a4845 100644 --- a/libavcodec/x86/dsputil_yasm.asm +++ b/libavcodec/x86/dsputil_yasm.asm @@ -397,3 +397,27 @@ cglobal add_hfyu_left_prediction_sse4, 3,3,7, dst, src, w, left .unaligned: ADD_HFYU_LEFT_LOOP 0 + +; float ff_scalarproduct_float_sse(const float *v1, const float *v2, int len) +cglobal scalarproduct_float_sse, 3,3,2, v1, v2, offset + neg offsetq + shl offsetq, 2 + sub v1q, offsetq + sub v2q, offsetq + xorps xmm0, xmm0 + .loop: + movaps xmm1, [v1q+offsetq] + mulps xmm1, [v2q+offsetq] + addps xmm0, xmm1 + add offsetq, 16 + js .loop + movhlps xmm1, xmm0 + addps xmm0, xmm1 + movss xmm1, xmm0 + shufps xmm0, xmm0, 1 + addss xmm0, xmm1 +%ifndef ARCH_X86_64 + movd r0m, xmm0 + fld dword r0m +%endif + RET |