2 * This file is part of FFmpeg.
4 * FFmpeg is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU Lesser General Public
6 * License as published by the Free Software Foundation; either
7 * version 2.1 of the License, or (at your option) any later version.
9 * FFmpeg is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12 * Lesser General Public License for more details.
14 * You should have received a copy of the GNU Lesser General Public
15 * License along with FFmpeg; if not, write to the Free Software
16 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
22 #include "libavutil/cpu.h"
23 #include "libavutil/x86/cpu.h"
24 #include "libavcodec/h264chroma.h"
26 void ff_put_h264_chroma_mc8_rnd_mmx (uint8_t *dst, uint8_t *src,
27 int stride, int h, int x, int y);
28 void ff_avg_h264_chroma_mc8_rnd_mmxext(uint8_t *dst, uint8_t *src,
29 int stride, int h, int x, int y);
30 void ff_avg_h264_chroma_mc8_rnd_3dnow(uint8_t *dst, uint8_t *src,
31 int stride, int h, int x, int y);
33 void ff_put_h264_chroma_mc4_mmx (uint8_t *dst, uint8_t *src,
34 int stride, int h, int x, int y);
35 void ff_avg_h264_chroma_mc4_mmxext (uint8_t *dst, uint8_t *src,
36 int stride, int h, int x, int y);
37 void ff_avg_h264_chroma_mc4_3dnow (uint8_t *dst, uint8_t *src,
38 int stride, int h, int x, int y);
40 void ff_put_h264_chroma_mc2_mmxext (uint8_t *dst, uint8_t *src,
41 int stride, int h, int x, int y);
42 void ff_avg_h264_chroma_mc2_mmxext (uint8_t *dst, uint8_t *src,
43 int stride, int h, int x, int y);
45 void ff_put_h264_chroma_mc8_rnd_ssse3(uint8_t *dst, uint8_t *src,
46 int stride, int h, int x, int y);
47 void ff_put_h264_chroma_mc4_ssse3 (uint8_t *dst, uint8_t *src,
48 int stride, int h, int x, int y);
50 void ff_avg_h264_chroma_mc8_rnd_ssse3(uint8_t *dst, uint8_t *src,
51 int stride, int h, int x, int y);
52 void ff_avg_h264_chroma_mc4_ssse3 (uint8_t *dst, uint8_t *src,
53 int stride, int h, int x, int y);
55 #define CHROMA_MC(OP, NUM, DEPTH, OPT) \
56 void ff_ ## OP ## _h264_chroma_mc ## NUM ## _ ## DEPTH ## _ ## OPT \
57 (uint8_t *dst, uint8_t *src, \
58 int stride, int h, int x, int y);
60 CHROMA_MC(put, 2, 10, mmxext)
61 CHROMA_MC(avg, 2, 10, mmxext)
62 CHROMA_MC(put, 4, 10, mmxext)
63 CHROMA_MC(avg, 4, 10, mmxext)
64 CHROMA_MC(put, 8, 10, sse2)
65 CHROMA_MC(avg, 8, 10, sse2)
66 CHROMA_MC(put, 8, 10, avx)
67 CHROMA_MC(avg, 8, 10, avx)
69 void ff_h264chroma_init_x86(H264ChromaContext *c, int bit_depth)
72 int high_bit_depth = bit_depth > 8;
73 int mm_flags = av_get_cpu_flags();
75 if (EXTERNAL_MMX(mm_flags) && !high_bit_depth) {
76 c->put_h264_chroma_pixels_tab[0] = ff_put_h264_chroma_mc8_rnd_mmx;
77 c->put_h264_chroma_pixels_tab[1] = ff_put_h264_chroma_mc4_mmx;
80 if (EXTERNAL_AMD3DNOW(mm_flags) && !high_bit_depth) {
81 c->avg_h264_chroma_pixels_tab[0] = ff_avg_h264_chroma_mc8_rnd_3dnow;
82 c->avg_h264_chroma_pixels_tab[1] = ff_avg_h264_chroma_mc4_3dnow;
85 if (EXTERNAL_MMXEXT(mm_flags) && !high_bit_depth) {
86 c->avg_h264_chroma_pixels_tab[0] = ff_avg_h264_chroma_mc8_rnd_mmxext;
87 c->avg_h264_chroma_pixels_tab[1] = ff_avg_h264_chroma_mc4_mmxext;
88 c->avg_h264_chroma_pixels_tab[2] = ff_avg_h264_chroma_mc2_mmxext;
89 c->put_h264_chroma_pixels_tab[2] = ff_put_h264_chroma_mc2_mmxext;
92 if (EXTERNAL_MMXEXT(mm_flags) && bit_depth == 10) {
93 c->put_h264_chroma_pixels_tab[2] = ff_put_h264_chroma_mc2_10_mmxext;
94 c->avg_h264_chroma_pixels_tab[2] = ff_avg_h264_chroma_mc2_10_mmxext;
95 c->put_h264_chroma_pixels_tab[1] = ff_put_h264_chroma_mc4_10_mmxext;
96 c->avg_h264_chroma_pixels_tab[1] = ff_avg_h264_chroma_mc4_10_mmxext;
99 if (EXTERNAL_SSE2(mm_flags) && bit_depth == 10) {
100 c->put_h264_chroma_pixels_tab[0] = ff_put_h264_chroma_mc8_10_sse2;
101 c->avg_h264_chroma_pixels_tab[0] = ff_avg_h264_chroma_mc8_10_sse2;
104 if (EXTERNAL_SSSE3(mm_flags) && !high_bit_depth) {
105 c->put_h264_chroma_pixels_tab[0] = ff_put_h264_chroma_mc8_rnd_ssse3;
106 c->avg_h264_chroma_pixels_tab[0] = ff_avg_h264_chroma_mc8_rnd_ssse3;
107 c->put_h264_chroma_pixels_tab[1] = ff_put_h264_chroma_mc4_ssse3;
108 c->avg_h264_chroma_pixels_tab[1] = ff_avg_h264_chroma_mc4_ssse3;
111 if (EXTERNAL_AVX(mm_flags) && bit_depth == 10) {
112 // AVX implies !cache64.
113 // TODO: Port cache(32|64) detection from x264.
114 c->put_h264_chroma_pixels_tab[0] = ff_put_h264_chroma_mc8_10_avx;
115 c->avg_h264_chroma_pixels_tab[0] = ff_avg_h264_chroma_mc8_10_avx;