Mercurial > libavcodec.hg
diff x86/h264dsp_mmx.c @ 11951:afee30fe8c26 libavcodec
16x16 and 8x8c x86 SIMD intra pred functions for VP8 and H.264
author | darkshikari |
---|---|
date | Fri, 25 Jun 2010 18:25:49 +0000 |
parents | 953d0c2d2c0a |
children | 65470eea7561 |
line wrap: on
line diff
--- a/x86/h264dsp_mmx.c Fri Jun 25 18:14:07 2010 +0000 +++ b/x86/h264dsp_mmx.c Fri Jun 25 18:25:49 2010 +0000 @@ -19,6 +19,7 @@ */ #include "dsputil_mmx.h" +#include "libavcodec/h264pred.h" DECLARE_ALIGNED(8, static const uint64_t, ff_pb_3_1 ) = 0x0103010301030103ULL; DECLARE_ALIGNED(8, static const uint64_t, ff_pb_7_3 ) = 0x0307030703070307ULL; @@ -2322,3 +2323,77 @@ H264_WEIGHT( 4, 4) H264_WEIGHT( 4, 2) +void ff_pred16x16_vertical_mmx (uint8_t *src, int stride); +void ff_pred16x16_vertical_sse (uint8_t *src, int stride); +void ff_pred16x16_horizontal_mmx (uint8_t *src, int stride); +void ff_pred16x16_horizontal_mmxext(uint8_t *src, int stride); +void ff_pred16x16_horizontal_ssse3 (uint8_t *src, int stride); +void ff_pred16x16_dc_mmx (uint8_t *src, int stride); +void ff_pred16x16_dc_mmxext (uint8_t *src, int stride); +void ff_pred16x16_dc_sse (uint8_t *src, int stride); +void ff_pred16x16_dc_sse2 (uint8_t *src, int stride); +void ff_pred16x16_dc_ssse3 (uint8_t *src, int stride); +void ff_pred16x16_tm_vp8_mmx (uint8_t *src, int stride); +void ff_pred16x16_tm_vp8_mmxext (uint8_t *src, int stride); +void ff_pred16x16_tm_vp8_sse2 (uint8_t *src, int stride); +void ff_pred8x8_dc_rv40_mmx (uint8_t *src, int stride); +void ff_pred8x8_dc_rv40_mmxext (uint8_t *src, int stride); +void ff_pred8x8_vertical_mmx (uint8_t *src, int stride); +void ff_pred8x8_horizontal_mmx (uint8_t *src, int stride); +void ff_pred8x8_horizontal_mmxext (uint8_t *src, int stride); +void ff_pred8x8_horizontal_ssse3 (uint8_t *src, int stride); +void ff_pred8x8_tm_vp8_mmx (uint8_t *src, int stride); +void ff_pred8x8_tm_vp8_mmxext (uint8_t *src, int stride); +void ff_pred8x8_tm_vp8_sse2 (uint8_t *src, int stride); +void ff_pred8x8_tm_vp8_ssse3 (uint8_t *src, int stride); + +void ff_h264_pred_init_x86(H264PredContext *h, int codec_id) +{ +#if HAVE_YASM + if (mm_flags & FF_MM_MMX) { + h->pred16x16[VERT_PRED8x8] = ff_pred16x16_vertical_mmx; + h->pred16x16[HOR_PRED8x8 ] = ff_pred16x16_horizontal_mmx; + h->pred16x16[DC_PRED8x8 ] = ff_pred16x16_dc_mmx; + h->pred8x8 [VERT_PRED8x8] = ff_pred8x8_vertical_mmx; + h->pred8x8 [HOR_PRED8x8 ] = ff_pred8x8_horizontal_mmx; + if (codec_id == CODEC_ID_VP8) { + h->pred16x16[PLANE_PRED8x8] = ff_pred16x16_tm_vp8_mmx; + h->pred8x8 [DC_PRED8x8 ] = ff_pred8x8_dc_rv40_mmx; + h->pred8x8 [PLANE_PRED8x8] = ff_pred8x8_tm_vp8_mmx; + } + } + + if (mm_flags & FF_MM_MMXEXT) { + h->pred16x16[HOR_PRED8x8 ] = ff_pred16x16_horizontal_mmxext; + h->pred16x16[DC_PRED8x8 ] = ff_pred16x16_dc_mmxext; + h->pred8x8 [HOR_PRED8x8 ] = ff_pred8x8_horizontal_mmxext; + if (codec_id == CODEC_ID_VP8) { + h->pred16x16[PLANE_PRED8x8] = ff_pred16x16_tm_vp8_mmxext; + h->pred8x8 [DC_PRED8x8 ] = ff_pred8x8_dc_rv40_mmxext; + h->pred8x8 [PLANE_PRED8x8] = ff_pred8x8_tm_vp8_mmxext; + } + } + + if (mm_flags & FF_MM_SSE) { + h->pred16x16[VERT_PRED8x8] = ff_pred16x16_vertical_sse; + h->pred16x16[DC_PRED8x8 ] = ff_pred16x16_dc_sse; + } + + if (mm_flags & FF_MM_SSE2) { + h->pred16x16[DC_PRED8x8 ] = ff_pred16x16_dc_sse2; + if (codec_id == CODEC_ID_VP8) { + h->pred16x16[PLANE_PRED8x8] = ff_pred16x16_tm_vp8_sse2; + h->pred8x8 [PLANE_PRED8x8] = ff_pred8x8_tm_vp8_sse2; + } + } + + if (mm_flags & FF_MM_SSSE3) { + h->pred16x16[HOR_PRED8x8 ] = ff_pred16x16_horizontal_ssse3; + h->pred16x16[DC_PRED8x8 ] = ff_pred16x16_dc_ssse3; + h->pred8x8 [HOR_PRED8x8 ] = ff_pred8x8_horizontal_ssse3; + if (codec_id == CODEC_ID_VP8) { + h->pred8x8 [PLANE_PRED8x8] = ff_pred8x8_tm_vp8_ssse3; + } + } +#endif +}