*
* Authors: Laurent Aimar <fenrir@via.ecp.fr>
* Loren Merritt <lorenm@u.washington.edu>
+ * Fiona Glaser <fiona@x264.com>
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software
- * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111, USA.
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA.
*****************************************************************************/
#include <stdlib.h>
#include <string.h>
#include "common/common.h"
-
-/* NASM functions */
-extern void x264_pixel_avg_16x16_sse2( uint8_t *, int, uint8_t *, int );
-extern void x264_pixel_avg_16x8_sse2( uint8_t *, int, uint8_t *, int );
-extern void x264_pixel_avg_16x16_mmxext( uint8_t *, int, uint8_t *, int );
-extern void x264_pixel_avg_16x8_mmxext( uint8_t *, int, uint8_t *, int );
-extern void x264_pixel_avg_8x16_mmxext( uint8_t *, int, uint8_t *, int );
-extern void x264_pixel_avg_8x8_mmxext( uint8_t *, int, uint8_t *, int );
-extern void x264_pixel_avg_8x4_mmxext( uint8_t *, int, uint8_t *, int );
-extern void x264_pixel_avg_4x8_mmxext( uint8_t *, int, uint8_t *, int );
-extern void x264_pixel_avg_4x4_mmxext( uint8_t *, int, uint8_t *, int );
-extern void x264_pixel_avg_4x2_mmxext( uint8_t *, int, uint8_t *, int );
+#include "mc.h"
+
+#define DECL_SUF( func, args )\
+ void func##_mmxext args;\
+ void func##_sse2 args;\
+ void func##_ssse3 args;
+
+DECL_SUF( x264_pixel_avg_16x16, ( uint8_t *, int, uint8_t *, int, uint8_t *, int, int ))
+DECL_SUF( x264_pixel_avg_16x8, ( uint8_t *, int, uint8_t *, int, uint8_t *, int, int ))
+DECL_SUF( x264_pixel_avg_8x16, ( uint8_t *, int, uint8_t *, int, uint8_t *, int, int ))
+DECL_SUF( x264_pixel_avg_8x8, ( uint8_t *, int, uint8_t *, int, uint8_t *, int, int ))
+DECL_SUF( x264_pixel_avg_8x4, ( uint8_t *, int, uint8_t *, int, uint8_t *, int, int ))
+DECL_SUF( x264_pixel_avg_4x8, ( uint8_t *, int, uint8_t *, int, uint8_t *, int, int ))
+DECL_SUF( x264_pixel_avg_4x4, ( uint8_t *, int, uint8_t *, int, uint8_t *, int, int ))
+DECL_SUF( x264_pixel_avg_4x2, ( uint8_t *, int, uint8_t *, int, uint8_t *, int, int ))
extern void x264_mc_copy_w4_mmx( uint8_t *, int, uint8_t *, int, int );
extern void x264_mc_copy_w8_mmx( uint8_t *, int, uint8_t *, int, int );
extern void x264_mc_copy_w16_mmx( uint8_t *, int, uint8_t *, int, int );
extern void x264_mc_copy_w16_sse2( uint8_t *, int, uint8_t *, int, int );
extern void x264_mc_copy_w16_sse3( uint8_t *, int, uint8_t *, int, int );
extern void x264_mc_copy_w16_aligned_sse2( uint8_t *, int, uint8_t *, int, int );
-extern void x264_pixel_avg_weight_4x4_mmxext( uint8_t *, int, uint8_t *, int, int );
-extern void x264_pixel_avg_weight_w8_mmxext( uint8_t *, int, uint8_t *, int, int, int );
-extern void x264_pixel_avg_weight_w16_mmxext( uint8_t *, int, uint8_t *, int, int, int );
extern void x264_prefetch_fenc_mmxext( uint8_t *, int, uint8_t *, int, int );
extern void x264_prefetch_ref_mmxext( uint8_t *, int, int );
extern void x264_mc_chroma_mmxext( uint8_t *src, int i_src_stride,
uint8_t *dst, int i_dst_stride,
int dx, int dy, int i_width, int i_height );
+extern void x264_mc_chroma_sse2( uint8_t *src, int i_src_stride,
+ uint8_t *dst, int i_dst_stride,
+ int dx, int dy, int i_width, int i_height );
+extern void x264_mc_chroma_ssse3( uint8_t *src, int i_src_stride,
+ uint8_t *dst, int i_dst_stride,
+ int dx, int dy, int i_width, int i_height );
extern void x264_plane_copy_mmxext( uint8_t *, int, uint8_t *, int, int w, int h);
extern void *x264_memcpy_aligned_mmx( void * dst, const void * src, size_t n );
extern void *x264_memcpy_aligned_sse2( void * dst, const void * src, size_t n );
extern void x264_memzero_aligned_mmx( void * dst, int n );
extern void x264_memzero_aligned_sse2( void * dst, int n );
+#define LOWRES(cpu) \
+extern void x264_frame_init_lowres_core_##cpu( uint8_t *src0, uint8_t *dst0, uint8_t *dsth, uint8_t *dstv, uint8_t *dstc,\
+ int src_stride, int dst_stride, int width, int height );
+LOWRES(mmxext)
+LOWRES(cache32_mmxext)
+LOWRES(sse2)
+LOWRES(ssse3)
#define PIXEL_AVG_W(width,cpu)\
extern void x264_pixel_avg2_w##width##_##cpu( uint8_t *, int, uint8_t *, int, uint8_t *, int );
PIXEL_AVG_WALL(cache64_mmxext)
PIXEL_AVG_WALL(cache64_sse2)
PIXEL_AVG_WALL(sse2)
-PIXEL_AVG_WALL(sse3)
-
-#define AVG_WEIGHT(W,H) \
-void x264_pixel_avg_weight_ ## W ## x ## H ## _mmxext( uint8_t *dst, int i_dst, uint8_t *src, int i_src, int i_weight_dst ) \
-{ \
- x264_pixel_avg_weight_w ## W ## _mmxext( dst, i_dst, src, i_src, i_weight_dst, H ); \
-}
-AVG_WEIGHT(16,16)
-AVG_WEIGHT(16,8)
-AVG_WEIGHT(8,16)
-AVG_WEIGHT(8,8)
-AVG_WEIGHT(8,4)
#define PIXEL_AVG_WTAB(instr, name1, name2, name3, name4, name5)\
static void (* const x264_pixel_avg_wtab_##instr[6])( uint8_t *, int, uint8_t *, int, uint8_t *, int ) =\
PIXEL_AVG_WTAB(mmxext, mmxext, mmxext, mmxext, mmxext, mmxext)
#ifdef ARCH_X86
PIXEL_AVG_WTAB(cache32_mmxext, mmxext, cache32_mmxext, cache32_mmxext, cache32_mmxext, cache32_mmxext)
-#endif
PIXEL_AVG_WTAB(cache64_mmxext, mmxext, cache64_mmxext, cache64_mmxext, cache64_mmxext, cache64_mmxext)
+#endif
PIXEL_AVG_WTAB(sse2, mmxext, mmxext, mmxext, sse2, sse2)
PIXEL_AVG_WTAB(cache64_sse2, mmxext, cache64_mmxext, cache64_sse2, cache64_sse2, cache64_sse2)
-PIXEL_AVG_WTAB(cache64_sse3, mmxext, cache64_mmxext, sse3, sse3, sse3)
#define MC_COPY_WTAB(instr, name1, name2, name3)\
static void (* const x264_mc_copy_wtab_##instr[5])( uint8_t *, int, uint8_t *, int, int ) =\
MC_COPY_WTAB(mmx,mmx,mmx,mmx)
MC_COPY_WTAB(sse2,mmx,mmx,sse2)
-MC_COPY_WTAB(sse3,mmx,mmx,sse3)
static const int hpel_ref0[16] = {0,1,1,1,0,1,1,1,2,3,3,3,0,1,1,1};
static const int hpel_ref1[16] = {0,0,0,0,2,2,3,2,2,2,3,2,2,2,3,2};
#define MC_LUMA(name,instr1,instr2)\
-void mc_luma_##name( uint8_t *dst, int i_dst_stride,\
+static void mc_luma_##name( uint8_t *dst, int i_dst_stride,\
uint8_t *src[4], int i_src_stride,\
int mvx, int mvy,\
int i_width, int i_height )\
#endif
MC_LUMA(sse2,sse2,sse2)
MC_LUMA(cache64_sse2,cache64_sse2,sse2)
-MC_LUMA(cache64_sse3,cache64_sse3,sse3)
#define GET_REF(name)\
-uint8_t *get_ref_##name( uint8_t *dst, int *i_dst_stride,\
+static uint8_t *get_ref_##name( uint8_t *dst, int *i_dst_stride,\
uint8_t *src[4], int i_src_stride,\
int mvx, int mvy,\
int i_width, int i_height )\
#endif
GET_REF(sse2)
GET_REF(cache64_sse2)
-GET_REF(cache64_sse3)
#define HPEL(align, cpu, cpuv, cpuc, cpuh)\
void x264_hpel_filter_v_##cpuv( uint8_t *dst, uint8_t *src, int16_t *buf, int stride, int width);\
void x264_hpel_filter_c_##cpuc( uint8_t *dst, int16_t *buf, int width );\
void x264_hpel_filter_h_##cpuh( uint8_t *dst, uint8_t *src, int width );\
void x264_sfence( void );\
-void x264_hpel_filter_##cpu( uint8_t *dsth, uint8_t *dstv, uint8_t *dstc, uint8_t *src,\
+static void x264_hpel_filter_##cpu( uint8_t *dsth, uint8_t *dstv, uint8_t *dstc, uint8_t *src,\
int stride, int width, int height )\
{\
int16_t *buf;\
HPEL(8, mmxext, mmxext, mmxext, mmxext)
HPEL(16, sse2_amd, mmxext, mmxext, sse2)
+#ifdef ARCH_X86_64
+void x264_hpel_filter_sse2( uint8_t *dsth, uint8_t *dstv, uint8_t *dstc, uint8_t *src, int stride, int width, int height );
+void x264_hpel_filter_ssse3( uint8_t *dsth, uint8_t *dstv, uint8_t *dstc, uint8_t *src, int stride, int width, int height );
+#else
HPEL(16, sse2, sse2, sse2, sse2)
-HPEL(16, ssse3, sse2, ssse3, sse2)
+HPEL(16, ssse3, sse2, ssse3, ssse3)
+#endif
void x264_mc_init_mmx( int cpu, x264_mc_functions_t *pf )
{
if( !(cpu&X264_CPU_MMX) )
return;
+ pf->copy_16x16_unaligned = x264_mc_copy_w16_mmx;
pf->copy[PIXEL_16x16] = x264_mc_copy_w16_mmx;
pf->copy[PIXEL_8x8] = x264_mc_copy_w8_mmx;
pf->copy[PIXEL_4x4] = x264_mc_copy_w4_mmx;
pf->avg[PIXEL_4x8] = x264_pixel_avg_4x8_mmxext;
pf->avg[PIXEL_4x4] = x264_pixel_avg_4x4_mmxext;
pf->avg[PIXEL_4x2] = x264_pixel_avg_4x2_mmxext;
-
- pf->avg_weight[PIXEL_16x16] = x264_pixel_avg_weight_16x16_mmxext;
- pf->avg_weight[PIXEL_16x8] = x264_pixel_avg_weight_16x8_mmxext;
- pf->avg_weight[PIXEL_8x16] = x264_pixel_avg_weight_8x16_mmxext;
- pf->avg_weight[PIXEL_8x8] = x264_pixel_avg_weight_8x8_mmxext;
- pf->avg_weight[PIXEL_8x4] = x264_pixel_avg_weight_8x4_mmxext;
- pf->avg_weight[PIXEL_4x4] = x264_pixel_avg_weight_4x4_mmxext;
- // avg_weight_4x8 is rare and 4x2 is not used
pf->plane_copy = x264_plane_copy_mmxext;
pf->hpel_filter = x264_hpel_filter_mmxext;
+ pf->frame_init_lowres_core = x264_frame_init_lowres_core_mmxext;
pf->prefetch_fenc = x264_prefetch_fenc_mmxext;
pf->prefetch_ref = x264_prefetch_ref_mmxext;
{
pf->mc_luma = mc_luma_cache32_mmxext;
pf->get_ref = get_ref_cache32_mmxext;
+ pf->frame_init_lowres_core = x264_frame_init_lowres_core_cache32_mmxext;
}
- else if( cpu&X264_CPU_CACHELINE_SPLIT )
+ else if( cpu&X264_CPU_CACHELINE_64 )
{
pf->mc_luma = mc_luma_cache64_mmxext;
pf->get_ref = get_ref_cache64_mmxext;
+ pf->frame_init_lowres_core = x264_frame_init_lowres_core_cache32_mmxext;
}
#endif
pf->memzero_aligned = x264_memzero_aligned_sse2;
pf->hpel_filter = x264_hpel_filter_sse2_amd;
- // disable on AMD processors since it is slower
- if( cpu&X264_CPU_3DNOW )
+ if( cpu&X264_CPU_SSE2_IS_SLOW )
return;
- pf->mc_luma = mc_luma_sse2;
- pf->get_ref = get_ref_sse2;
pf->copy[PIXEL_16x16] = x264_mc_copy_w16_aligned_sse2;
pf->avg[PIXEL_16x16] = x264_pixel_avg_16x16_sse2;
pf->avg[PIXEL_16x8] = x264_pixel_avg_16x8_sse2;
+ pf->avg[PIXEL_8x16] = x264_pixel_avg_8x16_sse2;
+ pf->avg[PIXEL_8x8] = x264_pixel_avg_8x8_sse2;
+ pf->avg[PIXEL_8x4] = x264_pixel_avg_8x4_sse2;
pf->hpel_filter = x264_hpel_filter_sse2;
+ pf->frame_init_lowres_core = x264_frame_init_lowres_core_sse2;
+ pf->mc_chroma = x264_mc_chroma_sse2;
- if( cpu&X264_CPU_CACHELINE_SPLIT )
+ if( cpu&X264_CPU_SSE2_IS_FAST )
{
- pf->mc_luma = mc_luma_cache64_sse2;
- pf->get_ref = get_ref_cache64_sse2;
- /* lddqu doesn't work on Core2 */
- if( (cpu&X264_CPU_SSE3) && !(cpu&X264_CPU_SSSE3) )
+ pf->mc_luma = mc_luma_sse2;
+ pf->get_ref = get_ref_sse2;
+ if( cpu&X264_CPU_CACHELINE_64 )
{
- pf->mc_luma = mc_luma_cache64_sse3;
- pf->get_ref = get_ref_cache64_sse3;
+ pf->mc_luma = mc_luma_cache64_sse2;
+ pf->get_ref = get_ref_cache64_sse2;
}
}
if( !(cpu&X264_CPU_SSSE3) )
return;
+ pf->avg[PIXEL_16x16] = x264_pixel_avg_16x16_ssse3;
+ pf->avg[PIXEL_16x8] = x264_pixel_avg_16x8_ssse3;
+ pf->avg[PIXEL_8x16] = x264_pixel_avg_8x16_ssse3;
+ pf->avg[PIXEL_8x8] = x264_pixel_avg_8x8_ssse3;
+ pf->avg[PIXEL_8x4] = x264_pixel_avg_8x4_ssse3;
+ pf->avg[PIXEL_4x8] = x264_pixel_avg_4x8_ssse3;
+ pf->avg[PIXEL_4x4] = x264_pixel_avg_4x4_ssse3;
+ pf->avg[PIXEL_4x2] = x264_pixel_avg_4x2_ssse3;
+
pf->hpel_filter = x264_hpel_filter_ssse3;
+ pf->frame_init_lowres_core = x264_frame_init_lowres_core_ssse3;
+ pf->mc_chroma = x264_mc_chroma_ssse3;
}
projects / x264 / blobdiff