FFmpeg: libavcodec/x86/vc1dsp

00001 /*
00002  * VC-1 and WMV3 - DSP functions MMX-optimized
00003  * Copyright (c) 2007 Christophe GISQUET <christophe.gisquet@free.fr>
00004  *
00005  * Permission is hereby granted, free of charge, to any person
00006  * obtaining a copy of this software and associated documentation
00007  * files (the "Software"), to deal in the Software without
00008  * restriction, including without limitation the rights to use,
00009  * copy, modify, merge, publish, distribute, sublicense, and/or sell
00010  * copies of the Software, and to permit persons to whom the
00011  * Software is furnished to do so, subject to the following
00012  * conditions:
00013  *
00014  * The above copyright notice and this permission notice shall be
00015  * included in all copies or substantial portions of the Software.
00016  *
00017  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
00018  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
00019  * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
00020  * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
00021  * HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
00022  * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
00023  * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
00024  * OTHER DEALINGS IN THE SOFTWARE.
00025  */
00026 
00027 #include "libavutil/cpu.h"
00028 #include "libavutil/x86/cpu.h"
00029 #include "libavcodec/vc1dsp.h"
00030 #include "vc1dsp.h"
00031 #include "config.h"
00032 
00033 #define LOOP_FILTER(EXT) \
00034 void ff_vc1_v_loop_filter4_ ## EXT(uint8_t *src, int stride, int pq); \
00035 void ff_vc1_h_loop_filter4_ ## EXT(uint8_t *src, int stride, int pq); \
00036 void ff_vc1_v_loop_filter8_ ## EXT(uint8_t *src, int stride, int pq); \
00037 void ff_vc1_h_loop_filter8_ ## EXT(uint8_t *src, int stride, int pq); \
00038 \
00039 static void vc1_v_loop_filter16_ ## EXT(uint8_t *src, int stride, int pq) \
00040 { \
00041     ff_vc1_v_loop_filter8_ ## EXT(src,   stride, pq); \
00042     ff_vc1_v_loop_filter8_ ## EXT(src+8, stride, pq); \
00043 } \
00044 \
00045 static void vc1_h_loop_filter16_ ## EXT(uint8_t *src, int stride, int pq) \
00046 { \
00047     ff_vc1_h_loop_filter8_ ## EXT(src,          stride, pq); \
00048     ff_vc1_h_loop_filter8_ ## EXT(src+8*stride, stride, pq); \
00049 }
00050 
00051 #if HAVE_YASM
00052 LOOP_FILTER(mmxext)
00053 LOOP_FILTER(sse2)
00054 LOOP_FILTER(ssse3)
00055 
00056 void ff_vc1_h_loop_filter8_sse4(uint8_t *src, int stride, int pq);
00057 
00058 static void vc1_h_loop_filter16_sse4(uint8_t *src, int stride, int pq)
00059 {
00060     ff_vc1_h_loop_filter8_sse4(src,          stride, pq);
00061     ff_vc1_h_loop_filter8_sse4(src+8*stride, stride, pq);
00062 }
00063 #endif /* HAVE_YASM */
00064 
00065 void ff_put_vc1_chroma_mc8_nornd_mmx  (uint8_t *dst, uint8_t *src,
00066                                        int stride, int h, int x, int y);
00067 void ff_avg_vc1_chroma_mc8_nornd_mmxext(uint8_t *dst, uint8_t *src,
00068                                         int stride, int h, int x, int y);
00069 void ff_avg_vc1_chroma_mc8_nornd_3dnow(uint8_t *dst, uint8_t *src,
00070                                        int stride, int h, int x, int y);
00071 void ff_put_vc1_chroma_mc8_nornd_ssse3(uint8_t *dst, uint8_t *src,
00072                                        int stride, int h, int x, int y);
00073 void ff_avg_vc1_chroma_mc8_nornd_ssse3(uint8_t *dst, uint8_t *src,
00074                                        int stride, int h, int x, int y);
00075 
00076 
00077 av_cold void ff_vc1dsp_init_x86(VC1DSPContext *dsp)
00078 {
00079     int mm_flags = av_get_cpu_flags();
00080 
00081     if (INLINE_MMX(mm_flags))
00082         ff_vc1dsp_init_mmx(dsp);
00083 
00084     if (INLINE_MMXEXT(mm_flags))
00085         ff_vc1dsp_init_mmxext(dsp);
00086 
00087 #define ASSIGN_LF(EXT) \
00088         dsp->vc1_v_loop_filter4  = ff_vc1_v_loop_filter4_ ## EXT; \
00089         dsp->vc1_h_loop_filter4  = ff_vc1_h_loop_filter4_ ## EXT; \
00090         dsp->vc1_v_loop_filter8  = ff_vc1_v_loop_filter8_ ## EXT; \
00091         dsp->vc1_h_loop_filter8  = ff_vc1_h_loop_filter8_ ## EXT; \
00092         dsp->vc1_v_loop_filter16 = vc1_v_loop_filter16_ ## EXT; \
00093         dsp->vc1_h_loop_filter16 = vc1_h_loop_filter16_ ## EXT
00094 
00095 #if HAVE_YASM
00096     if (mm_flags & AV_CPU_FLAG_MMX) {
00097         dsp->put_no_rnd_vc1_chroma_pixels_tab[0] = ff_put_vc1_chroma_mc8_nornd_mmx;
00098     }
00099 
00100     if (mm_flags & AV_CPU_FLAG_MMXEXT) {
00101         ASSIGN_LF(mmxext);
00102         dsp->avg_no_rnd_vc1_chroma_pixels_tab[0] = ff_avg_vc1_chroma_mc8_nornd_mmxext;
00103     } else if (mm_flags & AV_CPU_FLAG_3DNOW) {
00104         dsp->avg_no_rnd_vc1_chroma_pixels_tab[0] = ff_avg_vc1_chroma_mc8_nornd_3dnow;
00105     }
00106 
00107     if (mm_flags & AV_CPU_FLAG_SSE2) {
00108         dsp->vc1_v_loop_filter8  = ff_vc1_v_loop_filter8_sse2;
00109         dsp->vc1_h_loop_filter8  = ff_vc1_h_loop_filter8_sse2;
00110         dsp->vc1_v_loop_filter16 = vc1_v_loop_filter16_sse2;
00111         dsp->vc1_h_loop_filter16 = vc1_h_loop_filter16_sse2;
00112     }
00113     if (mm_flags & AV_CPU_FLAG_SSSE3) {
00114         ASSIGN_LF(ssse3);
00115         dsp->put_no_rnd_vc1_chroma_pixels_tab[0] = ff_put_vc1_chroma_mc8_nornd_ssse3;
00116         dsp->avg_no_rnd_vc1_chroma_pixels_tab[0] = ff_avg_vc1_chroma_mc8_nornd_ssse3;
00117     }
00118     if (mm_flags & AV_CPU_FLAG_SSE4) {
00119         dsp->vc1_h_loop_filter8  = ff_vc1_h_loop_filter8_sse4;
00120         dsp->vc1_h_loop_filter16 = vc1_h_loop_filter16_sse4;
00121     }
00122 #endif /* HAVE_YASM */
00123 }
libavcodec/x86/vc1dsp_init.c
