diff options
author | Justin Ruggles <justin.ruggles@gmail.com> | 2011-06-02 22:00:50 +0400 |
---|---|---|
committer | Justin Ruggles <justin.ruggles@gmail.com> | 2011-07-01 21:02:11 +0400 |
commit | 6054cd25b4d7dce97c4fa3cc6e4757ba1e59ab86 (patch) | |
tree | bb0f244fb2f37fcffede571532684b89ca43b590 /libavcodec/x86 | |
parent | 8a8d0ce208b77f506759185ff580fa61b5c41f70 (diff) |
ac3enc: add int32_t array clipping function to DSPUtil, including x86 versions.
Diffstat (limited to 'libavcodec/x86')
-rw-r--r-- | libavcodec/x86/dsputil_mmx.c | 23 | ||||
-rw-r--r-- | libavcodec/x86/dsputil_yasm.asm | 115 |
2 files changed, 138 insertions, 0 deletions
diff --git a/libavcodec/x86/dsputil_mmx.c b/libavcodec/x86/dsputil_mmx.c index 4d4bbc5b6f..03c094533f 100644 --- a/libavcodec/x86/dsputil_mmx.c +++ b/libavcodec/x86/dsputil_mmx.c @@ -2429,6 +2429,15 @@ int ff_add_hfyu_left_prediction_sse4(uint8_t *dst, const uint8_t *src, int w, i float ff_scalarproduct_float_sse(const float *v1, const float *v2, int order); +void ff_vector_clip_int32_mmx (int32_t *dst, const int32_t *src, int32_t min, + int32_t max, unsigned int len); +void ff_vector_clip_int32_sse2 (int32_t *dst, const int32_t *src, int32_t min, + int32_t max, unsigned int len); +void ff_vector_clip_int32_sse2_int(int32_t *dst, const int32_t *src, int32_t min, + int32_t max, unsigned int len); +void ff_vector_clip_int32_sse41 (int32_t *dst, const int32_t *src, int32_t min, + int32_t max, unsigned int len); + void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx) { int mm_flags = av_get_cpu_flags(); @@ -2570,6 +2579,8 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx) c->put_rv40_chroma_pixels_tab[0]= ff_put_rv40_chroma_mc8_mmx; c->put_rv40_chroma_pixels_tab[1]= ff_put_rv40_chroma_mc4_mmx; + + c->vector_clip_int32 = ff_vector_clip_int32_mmx; #endif if (mm_flags & AV_CPU_FLAG_MMX2) { @@ -2855,6 +2866,11 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx) #if HAVE_YASM c->scalarproduct_int16 = ff_scalarproduct_int16_sse2; c->scalarproduct_and_madd_int16 = ff_scalarproduct_and_madd_int16_sse2; + if (mm_flags & AV_CPU_FLAG_ATOM) { + c->vector_clip_int32 = ff_vector_clip_int32_sse2_int; + } else { + c->vector_clip_int32 = ff_vector_clip_int32_sse2; + } if (avctx->flags & CODEC_FLAG_BITEXACT) { c->apply_window_int16 = ff_apply_window_int16_sse2_ba; } else { @@ -2880,6 +2896,13 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx) } #endif } + + if (mm_flags & AV_CPU_FLAG_SSE4 && HAVE_SSE) { +#if HAVE_YASM + c->vector_clip_int32 = ff_vector_clip_int32_sse41; +#endif + } + #if HAVE_AVX && HAVE_YASM if (mm_flags & AV_CPU_FLAG_AVX) { if (bit_depth == 10) { diff --git a/libavcodec/x86/dsputil_yasm.asm b/libavcodec/x86/dsputil_yasm.asm index 8b19cc1441..4e1ec24a7a 100644 --- a/libavcodec/x86/dsputil_yasm.asm +++ b/libavcodec/x86/dsputil_yasm.asm @@ -1048,3 +1048,118 @@ emu_edge sse %ifdef ARCH_X86_32 emu_edge mmx %endif + +;----------------------------------------------------------------------------- +; void ff_vector_clip_int32(int32_t *dst, const int32_t *src, int32_t min, +; int32_t max, unsigned int len) +;----------------------------------------------------------------------------- + +%macro PMINSD_MMX 3 ; dst, src, tmp + mova %3, %2 + pcmpgtd %3, %1 + pxor %1, %2 + pand %1, %3 + pxor %1, %2 +%endmacro + +%macro PMAXSD_MMX 3 ; dst, src, tmp + mova %3, %1 + pcmpgtd %3, %2 + pand %1, %3 + pandn %3, %2 + por %1, %3 +%endmacro + +%macro CLIPD_MMX 3-4 ; src/dst, min, max, tmp + PMINSD_MMX %1, %3, %4 + PMAXSD_MMX %1, %2, %4 +%endmacro + +%macro CLIPD_SSE2 3-4 ; src/dst, min (float), max (float), unused + cvtdq2ps %1, %1 + minps %1, %3 + maxps %1, %2 + cvtps2dq %1, %1 +%endmacro + +%macro CLIPD_SSE41 3-4 ; src/dst, min, max, unused + pminsd %1, %3 + pmaxsd %1, %2 +%endmacro + +%macro SPLATD_MMX 1 + punpckldq %1, %1 +%endmacro + +%macro SPLATD_SSE2 1 + pshufd %1, %1, 0 +%endmacro + +%macro VECTOR_CLIP_INT32 4 +cglobal vector_clip_int32_%1, 5,5,%2, dst, src, min, max, len +%ifidn %1, sse2 + cvtsi2ss m4, minm + cvtsi2ss m5, maxm +%else + movd m4, minm + movd m5, maxm +%endif + SPLATD m4 + SPLATD m5 +.loop: +%assign %%i 1 +%rep %3 + mova m0, [srcq+mmsize*0*%%i] + mova m1, [srcq+mmsize*1*%%i] + mova m2, [srcq+mmsize*2*%%i] + mova m3, [srcq+mmsize*3*%%i] +%if %4 + mova m7, [srcq+mmsize*4*%%i] + mova m8, [srcq+mmsize*5*%%i] + mova m9, [srcq+mmsize*6*%%i] + mova m10, [srcq+mmsize*7*%%i] +%endif + CLIPD m0, m4, m5, m6 + CLIPD m1, m4, m5, m6 + CLIPD m2, m4, m5, m6 + CLIPD m3, m4, m5, m6 +%if %4 + CLIPD m7, m4, m5, m6 + CLIPD m8, m4, m5, m6 + CLIPD m9, m4, m5, m6 + CLIPD m10, m4, m5, m6 +%endif + mova [dstq+mmsize*0*%%i], m0 + mova [dstq+mmsize*1*%%i], m1 + mova [dstq+mmsize*2*%%i], m2 + mova [dstq+mmsize*3*%%i], m3 +%if %4 + mova [dstq+mmsize*4*%%i], m7 + mova [dstq+mmsize*5*%%i], m8 + mova [dstq+mmsize*6*%%i], m9 + mova [dstq+mmsize*7*%%i], m10 +%endif +%assign %%i %%i+1 +%endrep + add srcq, mmsize*4*(%3+%4) + add dstq, mmsize*4*(%3+%4) + sub lend, mmsize*(%3+%4) + jg .loop + REP_RET +%endmacro + +INIT_MMX +%define SPLATD SPLATD_MMX +%define CLIPD CLIPD_MMX +VECTOR_CLIP_INT32 mmx, 0, 1, 0 +INIT_XMM +%define SPLATD SPLATD_SSE2 +VECTOR_CLIP_INT32 sse2_int, 6, 1, 0 +%define CLIPD CLIPD_SSE2 +VECTOR_CLIP_INT32 sse2, 6, 2, 0 +%define CLIPD CLIPD_SSE41 +%ifdef m8 +VECTOR_CLIP_INT32 sse41, 11, 1, 1 +%else +VECTOR_CLIP_INT32 sse41, 6, 1, 0 +%endif |