From ac4a85f47642b8d84c369bdb49b8d58611f1ca5e Mon Sep 17 00:00:00 2001 From: Daniel Kang Date: Mon, 11 Jul 2011 17:26:43 -0400 Subject: H.264: Add more x86 assembly for 10-bit H.264 predict functions Mainly ported from 8-bit H.264 predict. Some code ported from x264. LGPL ok by author. Signed-off-by: Ronald S. Bultje --- libavcodec/x86/h264_intrapred_init.c | 23 +++++++++++++++++++++++ 1 file changed, 23 insertions(+) (limited to 'libavcodec/x86/h264_intrapred_init.c') diff --git a/libavcodec/x86/h264_intrapred_init.c b/libavcodec/x86/h264_intrapred_init.c index a0c5164db1..62e4c8796b 100644 --- a/libavcodec/x86/h264_intrapred_init.c +++ b/libavcodec/x86/h264_intrapred_init.c @@ -47,6 +47,7 @@ PRED8x8(dc, 10, mmxext) PRED8x8(dc, 10, sse2) PRED8x8(top_dc, 10, mmxext) PRED8x8(top_dc, 10, sse2) +PRED8x8(plane, 10, sse2) PRED8x8(vertical, 10, sse2) PRED8x8(horizontal, 10, sse2) @@ -55,6 +56,8 @@ void ff_pred8x8l_ ## TYPE ## _ ## DEPTH ## _ ## OPT (uint8_t *src, int has_tople PRED8x8L(dc, 10, sse2) PRED8x8L(dc, 10, ssse3) +PRED8x8L(128_dc, 10, mmxext) +PRED8x8L(128_dc, 10, sse2) PRED8x8L(top_dc, 10, sse2) PRED8x8L(top_dc, 10, ssse3) PRED8x8L(vertical, 10, sse2) @@ -73,6 +76,14 @@ PRED8x8L(horizontal_up, 10, ssse3) #define PRED16x16(TYPE, DEPTH, OPT)\ void ff_pred16x16_ ## TYPE ## _ ## DEPTH ## _ ## OPT (uint8_t *src, int stride); +PRED16x16(dc, 10, mmxext) +PRED16x16(dc, 10, sse2) +PRED16x16(top_dc, 10, mmxext) +PRED16x16(top_dc, 10, sse2) +PRED16x16(128_dc, 10, mmxext) +PRED16x16(128_dc, 10, sse2) +PRED16x16(left_dc, 10, mmxext) +PRED16x16(left_dc, 10, sse2) PRED16x16(vertical, 10, mmxext) PRED16x16(vertical, 10, sse2) PRED16x16(horizontal, 10, mmxext) @@ -289,6 +300,12 @@ void ff_h264_pred_init_x86(H264PredContext *h, int codec_id, const int bit_depth h->pred8x8[DC_PRED8x8 ] = ff_pred8x8_dc_10_mmxext; h->pred8x8[TOP_DC_PRED8x8 ] = ff_pred8x8_top_dc_10_mmxext; + h->pred8x8l[DC_128_PRED ] = ff_pred8x8l_128_dc_10_mmxext; + + h->pred16x16[DC_PRED8x8 ] = ff_pred16x16_dc_10_mmxext; + h->pred16x16[TOP_DC_PRED8x8 ] = ff_pred16x16_top_dc_10_mmxext; + h->pred16x16[DC_128_PRED8x8 ] = ff_pred16x16_128_dc_10_mmxext; + h->pred16x16[LEFT_DC_PRED8x8 ] = ff_pred16x16_left_dc_10_mmxext; h->pred16x16[VERT_PRED8x8 ] = ff_pred16x16_vertical_10_mmxext; h->pred16x16[HOR_PRED8x8 ] = ff_pred16x16_horizontal_10_mmxext; } @@ -301,18 +318,24 @@ void ff_h264_pred_init_x86(H264PredContext *h, int codec_id, const int bit_depth h->pred8x8[DC_PRED8x8 ] = ff_pred8x8_dc_10_sse2; h->pred8x8[TOP_DC_PRED8x8 ] = ff_pred8x8_top_dc_10_sse2; + h->pred8x8[PLANE_PRED8x8 ] = ff_pred8x8_plane_10_sse2; h->pred8x8[VERT_PRED8x8 ] = ff_pred8x8_vertical_10_sse2; h->pred8x8[HOR_PRED8x8 ] = ff_pred8x8_horizontal_10_sse2; h->pred8x8l[VERT_PRED ] = ff_pred8x8l_vertical_10_sse2; h->pred8x8l[HOR_PRED ] = ff_pred8x8l_horizontal_10_sse2; h->pred8x8l[DC_PRED ] = ff_pred8x8l_dc_10_sse2; + h->pred8x8l[DC_128_PRED ] = ff_pred8x8l_128_dc_10_sse2; h->pred8x8l[TOP_DC_PRED ] = ff_pred8x8l_top_dc_10_sse2; h->pred8x8l[DIAG_DOWN_LEFT_PRED ] = ff_pred8x8l_down_left_10_sse2; h->pred8x8l[DIAG_DOWN_RIGHT_PRED] = ff_pred8x8l_down_right_10_sse2; h->pred8x8l[VERT_RIGHT_PRED ] = ff_pred8x8l_vertical_right_10_sse2; h->pred8x8l[HOR_UP_PRED ] = ff_pred8x8l_horizontal_up_10_sse2; + h->pred16x16[DC_PRED8x8 ] = ff_pred16x16_dc_10_sse2; + h->pred16x16[TOP_DC_PRED8x8 ] = ff_pred16x16_top_dc_10_sse2; + h->pred16x16[DC_128_PRED8x8 ] = ff_pred16x16_128_dc_10_sse2; + h->pred16x16[LEFT_DC_PRED8x8 ] = ff_pred16x16_left_dc_10_sse2; h->pred16x16[VERT_PRED8x8 ] = ff_pred16x16_vertical_10_sse2; h->pred16x16[HOR_PRED8x8 ] = ff_pred16x16_horizontal_10_sse2; } -- cgit v1.2.3