From 911b32f482985ab284a21522fa5a3c2baed421d5 Mon Sep 17 00:00:00 2001 From: Daniel Kang Date: Fri, 24 Dec 2010 22:43:07 +0000 Subject: [PATCH] Port pred4x4_down_left_mmxext (H.264 intra prediction) from x264 to FFmpeg. LGPL relicensing approved by original authors: Holger Lubitz , Jason Garrett-Glaser and Loren Merritt . Patch by Daniel Kang , as part of Google's GCI 2010. Originally committed as revision 26087 to svn://svn.ffmpeg.org/ffmpeg/trunk --- libavcodec/x86/h264_intrapred.asm | 28 ++++++++++++++++++++++++++++ libavcodec/x86/h264_intrapred_init.c | 3 +++ 2 files changed, 31 insertions(+) diff --git a/libavcodec/x86/h264_intrapred.asm b/libavcodec/x86/h264_intrapred.asm index 7d3ef95557..14a6038785 100644 --- a/libavcodec/x86/h264_intrapred.asm +++ b/libavcodec/x86/h264_intrapred.asm @@ -1104,3 +1104,31 @@ cglobal pred4x4_vertical_vp8_mmxext, 3,3 movd [r1+r2*1], m3 movd [r1+r2*2], m3 RET + +;----------------------------------------------------------------------------- +; void pred4x4_down_left_mmxext(uint8_t *src, const uint8_t *topright, int stride) +;----------------------------------------------------------------------------- + +INIT_MMX +cglobal pred4x4_down_left_mmxext, 3,3 + sub r0, r2 + movq m1, [r0] + punpckldq m1, [r1] + movq m2, m1 + movq m3, m1 + movq m4, m1 + psllq m1, 8 + pxor m2, m1 + psrlq m2, 8 + pxor m3, m2 + PRED4x4_LOWPASS m0, m1, m3, m4, m5 + lea r1, [r0+r2*2] + psrlq m0, 8 + movd [r0+r2*1], m0 + psrlq m0, 8 + movd [r0+r2*2], m0 + psrlq m0, 8 + movd [r1+r2*1], m0 + psrlq m0, 8 + movd [r1+r2*2], m0 + RET diff --git a/libavcodec/x86/h264_intrapred_init.c b/libavcodec/x86/h264_intrapred_init.c index 54b81426ce..10a6dd6703 100644 --- a/libavcodec/x86/h264_intrapred_init.c +++ b/libavcodec/x86/h264_intrapred_init.c @@ -58,6 +58,7 @@ 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_pred4x4_dc_mmxext (uint8_t *src, const uint8_t *topright, int stride); +void ff_pred4x4_down_left_mmxext (uint8_t *src, const uint8_t *topright, int stride); void ff_pred4x4_tm_vp8_mmx (uint8_t *src, const uint8_t *topright, int stride); void ff_pred4x4_tm_vp8_mmxext (uint8_t *src, const uint8_t *topright, int stride); void ff_pred4x4_tm_vp8_ssse3 (uint8_t *src, const uint8_t *topright, int stride); @@ -94,6 +95,8 @@ void ff_h264_pred_init_x86(H264PredContext *h, int codec_id) h->pred16x16[DC_PRED8x8 ] = ff_pred16x16_dc_mmxext; h->pred8x8 [HOR_PRED8x8 ] = ff_pred8x8_horizontal_mmxext; h->pred4x4 [DC_PRED ] = ff_pred4x4_dc_mmxext; + if (codec_id == CODEC_ID_VP8 || codec_id == CODEC_ID_H264) + h->pred4x4 [DIAG_DOWN_LEFT_PRED ] = ff_pred4x4_down_left_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;