/****************************************************************************
* x264_pixel_init:
****************************************************************************/
void x264_pixel_init( int cpu, x264_pixel_function_t *pixf )
{
memset( pixf, 0, sizeof(*pixf) );
#define INIT2( name, cpu ) \
pixf->name[PIXEL_16x16] = x264_pixel_##name##_16x16##cpu;\
pixf->name[PIXEL_16x8] = x264_pixel_##name##_16x8##cpu;
#define INIT4( name, cpu ) \
INIT2( name, cpu ) \
pixf->name[PIXEL_8x16] = x264_pixel_##name##_8x16##cpu;\
pixf->name[PIXEL_8x8] = x264_pixel_##name##_8x8##cpu;
#define INIT5( name, cpu ) \
INIT4( name, cpu ) \
pixf->name[PIXEL_8x4] = x264_pixel_##name##_8x4##cpu;
#define INIT7( name, cpu ) \
INIT5( name, cpu ) \
pixf->name[PIXEL_4x8] = x264_pixel_##name##_4x8##cpu;\
pixf->name[PIXEL_4x4] = x264_pixel_##name##_4x4##cpu;
#define INIT_ADS( cpu ) \
pixf->ads[PIXEL_16x16] = x264_pixel_ads4##cpu;\
pixf->ads[PIXEL_16x8] = x264_pixel_ads2##cpu;\
pixf->ads[PIXEL_8x8] = x264_pixel_ads1##cpu;
INIT7( sad, );
INIT7( sad_x3, );
INIT7( sad_x4, );
INIT7( ssd, );
INIT7( satd, );
INIT4( sa8d, );
INIT_ADS( );
pixf->ssim_4x4x2_core = ssim_4x4x2_core;
pixf->ssim_end4 = ssim_end4;
#ifdef HAVE_MMX
if( cpu&X264_CPU_MMX )
{
INIT7( ssd, _mmx );
}
if( cpu&X264_CPU_MMXEXT )
{
INIT7( sad, _mmxext );
INIT7( sad_x3, _mmxext );
INIT7( sad_x4, _mmxext );
INIT7( satd, _mmxext );
INIT_ADS( _mmxext );
#ifdef ARCH_X86
pixf->sa8d[PIXEL_16x16] = x264_pixel_sa8d_16x16_mmxext;
pixf->sa8d[PIXEL_8x8] = x264_pixel_sa8d_8x8_mmxext;
pixf->intra_sa8d_x3_8x8 = x264_intra_sa8d_x3_8x8_mmxext;
pixf->ssim_4x4x2_core = x264_pixel_ssim_4x4x2_core_mmxext;
if( cpu&X264_CPU_CACHELINE_SPLIT )
{
if( cpu&X264_CPU_CACHELINE_32 )
{
INIT5( sad, _cache32_mmxext );
INIT4( sad_x3, _cache32_mmxext );
INIT4( sad_x4, _cache32_mmxext );
}
else
{
INIT5( sad, _cache64_mmxext );
INIT4( sad_x3, _cache64_mmxext );
INIT4( sad_x4, _cache64_mmxext );
}
}
#else
if( cpu&X264_CPU_CACHELINE_SPLIT )
{
pixf->sad[PIXEL_8x16] = x264_pixel_sad_8x16_cache64_mmxext;
pixf->sad[PIXEL_8x8] = x264_pixel_sad_8x8_cache64_mmxext;
pixf->sad[PIXEL_8x4] = x264_pixel_sad_8x4_cache64_mmxext;
pixf->sad_x3[PIXEL_8x16] = x264_pixel_sad_x3_8x16_cache64_mmxext;
pixf->sad_x3[PIXEL_8x8] = x264_pixel_sad_x3_8x8_cache64_mmxext;
pixf->sad_x4[PIXEL_8x16] = x264_pixel_sad_x4_8x16_cache64_mmxext;
pixf->sad_x4[PIXEL_8x8] = x264_pixel_sad_x4_8x8_cache64_mmxext;
}
#endif
pixf->intra_satd_x3_16x16 = x264_intra_satd_x3_16x16_mmxext;
pixf->intra_satd_x3_8x8c = x264_intra_satd_x3_8x8c_mmxext;
pixf->intra_satd_x3_4x4 = x264_intra_satd_x3_4x4_mmxext;
}
// disable on AMD processors since it is slower
if( (cpu&X264_CPU_SSE2) && !(cpu&X264_CPU_3DNOW) )
{
INIT2( sad, _sse2 );
INIT2( sad_x3, _sse2 );
INIT2( sad_x4, _sse2 );
INIT5( satd, _sse2 );
INIT_ADS( _sse2 );
#ifdef ARCH_X86
if( cpu&X264_CPU_CACHELINE_SPLIT )
{
INIT2( sad, _cache64_sse2 );
INIT2( sad_x3, _cache64_sse2 );
INIT2( sad_x4, _cache64_sse2 );
}
#endif
}
// these are faster on both Intel and AMD
if( cpu&X264_CPU_SSE2 )
{
INIT2( ssd, _sse2 );
pixf->ssim_4x4x2_core = x264_pixel_ssim_4x4x2_core_sse2;
pixf->ssim_end4 = x264_pixel_ssim_end4_sse2;
#ifdef ARCH_X86_64
pixf->sa8d[PIXEL_16x16] = x264_pixel_sa8d_16x16_sse2;
pixf->sa8d[PIXEL_8x8] = x264_pixel_sa8d_8x8_sse2;
pixf->intra_sa8d_x3_8x8 = x264_intra_sa8d_x3_8x8_sse2;
#endif
}
#ifdef HAVE_SSE3
if( (cpu&X264_CPU_SSE3) && (cpu&X264_CPU_CACHELINE_SPLIT) )
{
INIT2( sad, _sse3 );
INIT2( sad_x3, _sse3 );
INIT2( sad_x4, _sse3 );
}
if( cpu&X264_CPU_SSSE3 )
{
INIT5( satd, _ssse3 );
INIT_ADS( _ssse3 );
#ifdef ARCH_X86_64
pixf->sa8d[PIXEL_16x16]= x264_pixel_sa8d_16x16_ssse3;
pixf->sa8d[PIXEL_8x8] = x264_pixel_sa8d_8x8_ssse3;
#endif
if( cpu&X264_CPU_CACHELINE_SPLIT )
{
INIT2( sad, _cache64_ssse3 );
INIT2( sad_x3, _cache64_ssse3 );
INIT2( sad_x4, _cache64_ssse3 );
}
}
#endif //HAVE_SSE3
#endif //HAVE_MMX
#ifdef ARCH_PPC
if( cpu&X264_CPU_ALTIVEC )
{
x264_pixel_altivec_init( pixf );
}
#endif
#ifdef ARCH_UltraSparc
INIT4( sad, _vis );
INIT4( sad_x3, _vis );
INIT4( sad_x4, _vis );
#endif
pixf->ads[PIXEL_8x16] =
pixf->ads[PIXEL_8x4] =
pixf->ads[PIXEL_4x8] = pixf->ads[PIXEL_16x8];
pixf->ads[PIXEL_4x4] = pixf->ads[PIXEL_8x8];
}
void Setup_Assembly_Primitives(EncoderPrimitives &p, int cpuMask)
{
#if HIGH_BIT_DEPTH
if (cpuMask & X265_CPU_SSE2) p.sa8d[0] = p.sa8d[0];
#else
if (cpuMask & X265_CPU_SSE2)
{
INIT8_NAME(sse_pp, ssd, _mmx);
INIT8(sad, _mmx2);
INIT8(sad_x3, _mmx2);
INIT8(sad_x4, _mmx2);
INIT8(satd, _mmx2);
p.satd[LUMA_8x32] = x265_pixel_satd_8x32_sse2;
p.satd[LUMA_12x16] = x265_pixel_satd_12x16_sse2;
p.satd[LUMA_16x4] = x265_pixel_satd_16x4_sse2;
p.satd[LUMA_16x12] = x265_pixel_satd_16x12_sse2;
p.satd[LUMA_16x32] = x265_pixel_satd_16x32_sse2;
p.satd[LUMA_16x64] = x265_pixel_satd_16x64_sse2;
p.satd[LUMA_32x8] = x265_pixel_satd_32x8_sse2;
p.satd[LUMA_32x16] = x265_pixel_satd_32x16_sse2;
p.satd[LUMA_32x24] = x265_pixel_satd_32x24_sse2;
p.sa8d[BLOCK_4x4] = x265_pixel_satd_4x4_mmx2;
p.frame_init_lowres_core = x265_frame_init_lowres_core_mmx2;
PIXEL_AVG(sse2);
PIXEL_AVG_W4(mmx2);
p.sad[LUMA_8x32] = x265_pixel_sad_8x32_sse2;
p.sad[LUMA_16x4] = x265_pixel_sad_16x4_sse2;
p.sad[LUMA_16x12] = x265_pixel_sad_16x12_sse2;
p.sad[LUMA_16x32] = x265_pixel_sad_16x32_sse2;
p.sad[LUMA_16x64] = x265_pixel_sad_16x64_sse2;
p.sad[LUMA_32x8] = x265_pixel_sad_32x8_sse2;
p.sad[LUMA_32x16] = x265_pixel_sad_32x16_sse2;
p.sad[LUMA_32x24] = x265_pixel_sad_32x24_sse2;
p.sad[LUMA_32x32] = x265_pixel_sad_32x32_sse2;
p.sad[LUMA_32x64] = x265_pixel_sad_32x64_sse2;
p.sad[LUMA_64x16] = x265_pixel_sad_64x16_sse2;
p.sad[LUMA_64x32] = x265_pixel_sad_64x32_sse2;
p.sad[LUMA_64x48] = x265_pixel_sad_64x48_sse2;
p.sad[LUMA_64x64] = x265_pixel_sad_64x64_sse2;
p.sad[LUMA_48x64] = x265_pixel_sad_48x64_sse2;
p.sad[LUMA_24x32] = x265_pixel_sad_24x32_sse2;
p.sad[LUMA_12x16] = x265_pixel_sad_12x16_sse2;
ASSGN_SSE(sse2);
INIT2(sad, _sse2);
INIT2(sad_x3, _sse2);
INIT2(sad_x4, _sse2);
INIT6(satd, _sse2);
HEVC_SATD(sse2);
CHROMA_BLOCKCOPY(_sse2);
LUMA_BLOCKCOPY(_sse2);
CHROMA_SS_FILTERS(_sse2);
LUMA_SS_FILTERS(_sse2);
// This function pointer initialization is temporary will be removed
// later with macro definitions. It is used to avoid linker errors
// until all partitions are coded and commit smaller patches, easier to
// review.
p.chroma_copy_sp[CHROMA_4x2] = x265_blockcopy_sp_4x2_sse2;
p.chroma_copy_sp[CHROMA_4x4] = x265_blockcopy_sp_4x4_sse2;
p.chroma_copy_sp[CHROMA_4x8] = x265_blockcopy_sp_4x8_sse2;
p.chroma_copy_sp[CHROMA_4x16] = x265_blockcopy_sp_4x16_sse2;
p.chroma_copy_sp[CHROMA_8x2] = x265_blockcopy_sp_8x2_sse2;
p.chroma_copy_sp[CHROMA_8x4] = x265_blockcopy_sp_8x4_sse2;
p.chroma_copy_sp[CHROMA_8x6] = x265_blockcopy_sp_8x6_sse2;
p.chroma_copy_sp[CHROMA_8x8] = x265_blockcopy_sp_8x8_sse2;
p.chroma_copy_sp[CHROMA_8x16] = x265_blockcopy_sp_8x16_sse2;
p.chroma_copy_sp[CHROMA_12x16] = x265_blockcopy_sp_12x16_sse2;
p.chroma_copy_sp[CHROMA_16x4] = x265_blockcopy_sp_16x4_sse2;
p.chroma_copy_sp[CHROMA_16x8] = x265_blockcopy_sp_16x8_sse2;
p.chroma_copy_sp[CHROMA_16x12] = x265_blockcopy_sp_16x12_sse2;
p.chroma_copy_sp[CHROMA_16x16] = x265_blockcopy_sp_16x16_sse2;
p.chroma_copy_sp[CHROMA_16x32] = x265_blockcopy_sp_16x32_sse2;
p.luma_copy_sp[LUMA_16x64] = x265_blockcopy_sp_16x64_sse2;
p.chroma_copy_sp[CHROMA_24x32] = x265_blockcopy_sp_24x32_sse2;
p.chroma_copy_sp[CHROMA_32x8] = x265_blockcopy_sp_32x8_sse2;
p.chroma_copy_sp[CHROMA_32x16] = x265_blockcopy_sp_32x16_sse2;
p.chroma_copy_sp[CHROMA_32x24] = x265_blockcopy_sp_32x24_sse2;
p.chroma_copy_sp[CHROMA_32x32] = x265_blockcopy_sp_32x32_sse2;
p.luma_copy_sp[LUMA_32x64] = x265_blockcopy_sp_32x64_sse2;
p.luma_copy_sp[LUMA_48x64] = x265_blockcopy_sp_48x64_sse2;
p.luma_copy_sp[LUMA_64x16] = x265_blockcopy_sp_64x16_sse2;
p.luma_copy_sp[LUMA_64x32] = x265_blockcopy_sp_64x32_sse2;
p.luma_copy_sp[LUMA_64x48] = x265_blockcopy_sp_64x48_sse2;
p.luma_copy_sp[LUMA_64x64] = x265_blockcopy_sp_64x64_sse2;
p.blockfill_s[BLOCK_4x4] = x265_blockfill_s_4x4_sse2;
p.blockfill_s[BLOCK_8x8] = x265_blockfill_s_8x8_sse2;
p.blockfill_s[BLOCK_16x16] = x265_blockfill_s_16x16_sse2;
p.blockfill_s[BLOCK_32x32] = x265_blockfill_s_32x32_sse2;
p.frame_init_lowres_core = x265_frame_init_lowres_core_sse2;
p.sa8d[BLOCK_8x8] = x265_pixel_sa8d_8x8_sse2;
p.sa8d[BLOCK_16x16] = x265_pixel_sa8d_16x16_sse2;
SA8D_INTER_FROM_BLOCK(sse2);
p.cvt32to16_shr = x265_cvt32to16_shr_sse2;
p.ipfilter_ss[FILTER_V_S_S_8] = x265_interp_8tap_v_ss_sse2;
p.calcrecon[BLOCK_4x4] = x265_calcRecons4_sse2;
p.calcrecon[BLOCK_8x8] = x265_calcRecons8_sse2;
p.calcresidual[BLOCK_4x4] = x265_getResidual4_sse2;
p.calcresidual[BLOCK_8x8] = x265_getResidual8_sse2;
}
if (cpuMask & X265_CPU_SSSE3)
{
p.frame_init_lowres_core = x265_frame_init_lowres_core_ssse3;
p.sa8d[BLOCK_8x8] = x265_pixel_sa8d_8x8_ssse3;
p.sa8d[BLOCK_16x16] = x265_pixel_sa8d_16x16_ssse3;
SA8D_INTER_FROM_BLOCK(ssse3);
p.sse_pp[LUMA_4x4] = x265_pixel_ssd_4x4_ssse3;
ASSGN_SSE(ssse3);
PIXEL_AVG(ssse3);
PIXEL_AVG_W4(ssse3);
p.scale1D_128to64 = x265_scale1D_128to64_ssse3;
p.sad_x4[LUMA_8x4] = x265_pixel_sad_x4_8x4_ssse3;
p.sad_x4[LUMA_8x8] = x265_pixel_sad_x4_8x8_ssse3;
p.sad_x3[LUMA_8x16] = x265_pixel_sad_x3_8x16_ssse3;
p.sad_x4[LUMA_8x16] = x265_pixel_sad_x4_8x16_ssse3;
p.sad_x3[LUMA_8x32] = x265_pixel_sad_x3_8x32_ssse3;
p.sad_x4[LUMA_8x32] = x265_pixel_sad_x4_8x32_ssse3;
p.sad_x3[LUMA_12x16] = x265_pixel_sad_x3_12x16_ssse3;
p.sad_x4[LUMA_12x16] = x265_pixel_sad_x4_12x16_ssse3;
p.sad_x3[LUMA_16x12] = x265_pixel_sad_x3_16x12_ssse3;
p.sad_x4[LUMA_16x12] = x265_pixel_sad_x4_16x12_ssse3;
p.sad_x3[LUMA_16x32] = x265_pixel_sad_x3_16x32_ssse3;
p.sad_x4[LUMA_16x32] = x265_pixel_sad_x4_16x32_ssse3;
p.sad_x3[LUMA_16x64] = x265_pixel_sad_x3_16x64_ssse3;
p.sad_x4[LUMA_16x64] = x265_pixel_sad_x4_16x64_ssse3;
p.sad_x3[LUMA_24x32] = x265_pixel_sad_x3_24x32_ssse3;
p.sad_x4[LUMA_24x32] = x265_pixel_sad_x4_24x32_ssse3;
p.sad_x3[LUMA_32x8] = x265_pixel_sad_x3_32x8_ssse3;
p.sad_x3[LUMA_32x16] = x265_pixel_sad_x3_32x16_ssse3;
p.sad_x3[LUMA_32x24] = x265_pixel_sad_x3_32x24_ssse3;
p.sad_x3[LUMA_32x32] = x265_pixel_sad_x3_32x32_ssse3;
p.sad_x3[LUMA_32x64] = x265_pixel_sad_x3_32x64_ssse3;
p.sad_x4[LUMA_32x8] = x265_pixel_sad_x4_32x8_ssse3;
p.sad_x4[LUMA_32x16] = x265_pixel_sad_x4_32x16_ssse3;
p.sad_x4[LUMA_32x24] = x265_pixel_sad_x4_32x24_ssse3;
p.sad_x4[LUMA_32x32] = x265_pixel_sad_x4_32x32_ssse3;
p.sad_x4[LUMA_32x64] = x265_pixel_sad_x4_32x64_ssse3;
p.sad_x3[LUMA_48x64] = x265_pixel_sad_x3_48x64_ssse3;
p.sad_x4[LUMA_48x64] = x265_pixel_sad_x4_48x64_ssse3;
p.sad_x3[LUMA_64x16] = x265_pixel_sad_x3_64x16_ssse3;
p.sad_x3[LUMA_64x32] = x265_pixel_sad_x3_64x32_ssse3;
p.sad_x3[LUMA_64x48] = x265_pixel_sad_x3_64x48_ssse3;
p.sad_x3[LUMA_64x64] = x265_pixel_sad_x3_64x64_ssse3;
p.sad_x4[LUMA_64x16] = x265_pixel_sad_x4_64x16_ssse3;
p.sad_x4[LUMA_64x32] = x265_pixel_sad_x4_64x32_ssse3;
p.sad_x4[LUMA_64x48] = x265_pixel_sad_x4_64x48_ssse3;
p.sad_x4[LUMA_64x64] = x265_pixel_sad_x4_64x64_ssse3;
p.luma_hvpp[LUMA_8x8] = x265_interp_8tap_hv_pp_8x8_ssse3;
p.luma_p2s = x265_luma_p2s_ssse3;
p.chroma_p2s = x265_chroma_p2s_ssse3;
CHROMA_SP_FILTERS(_ssse3);
LUMA_SP_FILTERS(_ssse3);
}
if (cpuMask & X265_CPU_SSE4)
{
p.satd[LUMA_4x16] = x265_pixel_satd_4x16_sse4;
p.satd[LUMA_12x16] = x265_pixel_satd_12x16_sse4;
p.satd[LUMA_32x8] = x265_pixel_satd_32x8_sse4;
p.satd[LUMA_32x16] = x265_pixel_satd_32x16_sse4;
p.satd[LUMA_32x24] = x265_pixel_satd_32x24_sse4;
p.sa8d[BLOCK_8x8] = x265_pixel_sa8d_8x8_sse4;
p.sa8d[BLOCK_16x16] = x265_pixel_sa8d_16x16_sse4;
SA8D_INTER_FROM_BLOCK(sse4);
CHROMA_FILTERS(_sse4);
LUMA_FILTERS(_sse4);
HEVC_SATD(sse4);
p.chroma_copy_sp[CHROMA_2x4] = x265_blockcopy_sp_2x4_sse4;
p.chroma_copy_sp[CHROMA_2x8] = x265_blockcopy_sp_2x8_sse4;
p.chroma_copy_sp[CHROMA_6x8] = x265_blockcopy_sp_6x8_sse4;
p.chroma_vsp[CHROMA_2x4] = x265_interp_4tap_vert_sp_2x4_sse4;
p.chroma_vsp[CHROMA_2x8] = x265_interp_4tap_vert_sp_2x8_sse4;
p.chroma_vsp[CHROMA_6x8] = x265_interp_4tap_vert_sp_6x8_sse4;
p.calcrecon[BLOCK_16x16] = x265_calcRecons16_sse4;
p.calcrecon[BLOCK_32x32] = x265_calcRecons32_sse4;
p.calcresidual[BLOCK_16x16] = x265_getResidual16_sse4;
p.calcresidual[BLOCK_32x32] = x265_getResidual32_sse4;
}
if (cpuMask & X265_CPU_AVX)
{
p.frame_init_lowres_core = x265_frame_init_lowres_core_avx;
p.satd[LUMA_4x16] = x265_pixel_satd_4x16_avx;
p.satd[LUMA_12x16] = x265_pixel_satd_12x16_avx;
p.satd[LUMA_32x8] = x265_pixel_satd_32x8_avx;
p.satd[LUMA_32x16] = x265_pixel_satd_32x16_avx;
p.satd[LUMA_32x24] = x265_pixel_satd_32x24_avx;
p.sa8d[BLOCK_8x8] = x265_pixel_sa8d_8x8_avx;
p.sa8d[BLOCK_16x16] = x265_pixel_sa8d_16x16_avx;
SA8D_INTER_FROM_BLOCK(avx);
ASSGN_SSE(avx);
HEVC_SATD(avx);
p.sad_x3[LUMA_12x16] = x265_pixel_sad_x3_12x16_avx;
p.sad_x4[LUMA_12x16] = x265_pixel_sad_x4_12x16_avx;
p.sad_x3[LUMA_16x4] = x265_pixel_sad_x3_16x4_avx;
p.sad_x4[LUMA_16x4] = x265_pixel_sad_x4_16x4_avx;
p.sad_x3[LUMA_16x12] = x265_pixel_sad_x3_16x12_avx;
p.sad_x4[LUMA_16x12] = x265_pixel_sad_x4_16x12_avx;
p.sad_x3[LUMA_16x32] = x265_pixel_sad_x3_16x32_avx;
p.sad_x4[LUMA_16x32] = x265_pixel_sad_x4_16x32_avx;
p.sad_x3[LUMA_16x64] = x265_pixel_sad_x3_16x64_avx;
p.sad_x4[LUMA_16x64] = x265_pixel_sad_x4_16x64_avx;
p.sad_x3[LUMA_24x32] = x265_pixel_sad_x3_24x32_avx;
p.sad_x4[LUMA_24x32] = x265_pixel_sad_x4_24x32_avx;
p.sad_x3[LUMA_32x8] = x265_pixel_sad_x3_32x8_avx;
p.sad_x3[LUMA_32x16] = x265_pixel_sad_x3_32x16_avx;
p.sad_x3[LUMA_32x24] = x265_pixel_sad_x3_32x24_avx;
p.sad_x3[LUMA_32x32] = x265_pixel_sad_x3_32x32_avx;
p.sad_x3[LUMA_32x64] = x265_pixel_sad_x3_32x64_avx;
p.sad_x4[LUMA_32x8] = x265_pixel_sad_x4_32x8_avx;
p.sad_x4[LUMA_32x16] = x265_pixel_sad_x4_32x16_avx;
p.sad_x4[LUMA_32x24] = x265_pixel_sad_x4_32x24_avx;
p.sad_x4[LUMA_32x32] = x265_pixel_sad_x4_32x32_avx;
p.sad_x4[LUMA_32x64] = x265_pixel_sad_x4_32x64_avx;
p.sad_x3[LUMA_48x64] = x265_pixel_sad_x3_48x64_avx;
p.sad_x4[LUMA_48x64] = x265_pixel_sad_x4_48x64_avx;
p.sad_x3[LUMA_64x16] = x265_pixel_sad_x3_64x16_avx;
p.sad_x3[LUMA_64x32] = x265_pixel_sad_x3_64x32_avx;
p.sad_x3[LUMA_64x48] = x265_pixel_sad_x3_64x48_avx;
p.sad_x3[LUMA_64x64] = x265_pixel_sad_x3_64x64_avx;
p.sad_x4[LUMA_64x16] = x265_pixel_sad_x4_64x16_avx;
p.sad_x4[LUMA_64x32] = x265_pixel_sad_x4_64x32_avx;
p.sad_x4[LUMA_64x48] = x265_pixel_sad_x4_64x48_avx;
p.sad_x4[LUMA_64x64] = x265_pixel_sad_x4_64x64_avx;
}
if (cpuMask & X265_CPU_XOP)
{
p.frame_init_lowres_core = x265_frame_init_lowres_core_xop;
p.sa8d[BLOCK_8x8] = x265_pixel_sa8d_8x8_xop;
p.sa8d[BLOCK_16x16] = x265_pixel_sa8d_16x16_xop;
SA8D_INTER_FROM_BLOCK(xop);
INIT7(satd, _xop);
INIT5_NAME(sse_pp, ssd, _xop);
HEVC_SATD(xop);
}
if (cpuMask & X265_CPU_AVX2)
{
INIT2(sad_x4, _avx2);
INIT4(satd, _avx2);
INIT2_NAME(sse_pp, ssd, _avx2);
p.sa8d[BLOCK_8x8] = x265_pixel_sa8d_8x8_avx2;
SA8D_INTER_FROM_BLOCK8(avx2);
p.satd[LUMA_32x32] = cmp<32, 32, 16, 16, x265_pixel_satd_16x16_avx2>;
p.satd[LUMA_24x32] = cmp<24, 32, 8, 16, x265_pixel_satd_8x16_avx2>;
p.satd[LUMA_64x64] = cmp<64, 64, 16, 16, x265_pixel_satd_16x16_avx2>;
p.satd[LUMA_64x32] = cmp<64, 32, 16, 16, x265_pixel_satd_16x16_avx2>;
p.satd[LUMA_32x64] = cmp<32, 64, 16, 16, x265_pixel_satd_16x16_avx2>;
p.satd[LUMA_64x48] = cmp<64, 48, 16, 16, x265_pixel_satd_16x16_avx2>;
p.satd[LUMA_48x64] = cmp<48, 64, 16, 16, x265_pixel_satd_16x16_avx2>;
p.satd[LUMA_64x16] = cmp<64, 16, 16, 16, x265_pixel_satd_16x16_avx2>;
p.sad_x4[LUMA_16x12] = x265_pixel_sad_x4_16x12_avx2;
p.sad_x4[LUMA_16x32] = x265_pixel_sad_x4_16x32_avx2;
}
#endif // if HIGH_BIT_DEPTH
}
/* bbbbbbbb ffffffff ffffffff ffffffff ffffffff ffffffff eeeeffff 0eeeeeee */
#define INIT8(e, f, b) (b), (f), (f), (f), (f), (f), (f)>>4 | ((e)&0x0FU)<<4, (e)>>4
/* bbbbbbbb ffffffff ffffffff ffffffff f...f 1fffffff eeeeeeee 0eeeeeee 00000000 00000000 */
#define INIT12(e, f, b) (b), (f), (f), (f), (f), (f), (f), 1U<<7 | (f)>>1, (e)&0xFFU, (e)>>8
_Fparam _Flt = {
(int)SIGF, /* FLT_MANT_DIG */
(int)((SIGF-1) * LOG2), /* FLT_DIG */
(int)(1-BIASF+1), /* FLT_MIN_EXP */
(int)((-BIASF+1) * LOG2), /* FLT_MIN_10_EXP */
(int)(BIASF+1), /* FLT_MAX_EXP */
(int)((BIASF+1) * LOG2), /* FLT_MAX_10_EXP */
{ INIT4(MAXEF-1, 0xFF, 0xFF) }, /* FLT_MAX */
{ INIT4(BIASF-(SIGF-1), 0x00, 0x00) }, /* FLT_EPSILON */
{ INIT4(0x01, 0x00, 0x00) }, /* FLT_MIN */
};
_Fparam _Dbl = {
(int)SIGD, /* DBL_MANT_DIG */
(int)((SIGD-1) * LOG2), /* DBL_DIG */
(int)(1-BIASD+1), /* DBL_MIN_EXP */
(int)((-BIASD+1) * LOG2), /* DBL_MIN_10_EXP */
(int)(BIASD+1), /* DBL_MAX_EXP */
(int)((BIASD+1) * LOG2), /* DBL_MAX_10_EXP */
{ INIT8(MAXED-1, 0xFF, 0xFF) }, /* DBL_MAX */
{ INIT8(BIASD-(SIGD-1), 0x00, 0x00) }, /* DBL_EPSILON */
{ INIT8(0x001, 0x00, 0x00) }, /* DBL_MIN */
};
/****************************************************************************
* x264_pixel_init:
****************************************************************************/
void x264_pixel_init( int cpu, x264_pixel_function_t *pixf )
{
memset( pixf, 0, sizeof(*pixf) );
#define INIT2( name, cpu ) \
pixf->name[PIXEL_16x16] = x264_pixel_##name##_16x16##cpu;\
pixf->name[PIXEL_16x8] = x264_pixel_##name##_16x8##cpu;
#define INIT4( name, cpu ) \
INIT2( name, cpu ) \
pixf->name[PIXEL_8x16] = x264_pixel_##name##_8x16##cpu;\
pixf->name[PIXEL_8x8] = x264_pixel_##name##_8x8##cpu;
#define INIT5( name, cpu ) \
INIT4( name, cpu ) \
pixf->name[PIXEL_8x4] = x264_pixel_##name##_8x4##cpu;
#define INIT7( name, cpu ) \
INIT5( name, cpu ) \
pixf->name[PIXEL_4x8] = x264_pixel_##name##_4x8##cpu;\
pixf->name[PIXEL_4x4] = x264_pixel_##name##_4x4##cpu;
#define INIT_ADS( cpu ) \
pixf->ads[PIXEL_16x16] = x264_pixel_ads4##cpu;\
pixf->ads[PIXEL_16x8] = x264_pixel_ads2##cpu;\
pixf->ads[PIXEL_8x8] = x264_pixel_ads1##cpu;
INIT7( sad, );
INIT7( sad_x3, );
INIT7( sad_x4, );
INIT7( ssd, );
INIT7( satd, );
INIT7( satd_x3, );
INIT7( satd_x4, );
INIT4( sa8d, );
INIT_ADS( );
pixf->var[PIXEL_16x16] = x264_pixel_var_16x16;
pixf->var[PIXEL_8x8] = x264_pixel_var_8x8;
pixf->ssim_4x4x2_core = ssim_4x4x2_core;
pixf->ssim_end4 = ssim_end4;
#ifdef HAVE_MMX
if( cpu&X264_CPU_MMX )
{
INIT7( ssd, _mmx );
}
if( cpu&X264_CPU_MMXEXT )
{
INIT7( sad, _mmxext );
INIT7( sad_x3, _mmxext );
INIT7( sad_x4, _mmxext );
INIT7( satd, _mmxext );
INIT7( satd_x3, _mmxext );
INIT7( satd_x4, _mmxext );
INIT_ADS( _mmxext );
pixf->var[PIXEL_16x16] = x264_pixel_var_16x16_mmxext;
pixf->var[PIXEL_8x8] = x264_pixel_var_8x8_mmxext;
#ifdef ARCH_X86
pixf->sa8d[PIXEL_16x16] = x264_pixel_sa8d_16x16_mmxext;
pixf->sa8d[PIXEL_8x8] = x264_pixel_sa8d_8x8_mmxext;
pixf->intra_sa8d_x3_8x8 = x264_intra_sa8d_x3_8x8_mmxext;
pixf->ssim_4x4x2_core = x264_pixel_ssim_4x4x2_core_mmxext;
if( cpu&X264_CPU_CACHELINE_32 )
{
INIT5( sad, _cache32_mmxext );
INIT4( sad_x3, _cache32_mmxext );
INIT4( sad_x4, _cache32_mmxext );
}
else if( cpu&X264_CPU_CACHELINE_64 )
{
INIT5( sad, _cache64_mmxext );
INIT4( sad_x3, _cache64_mmxext );
INIT4( sad_x4, _cache64_mmxext );
}
#else
if( cpu&X264_CPU_CACHELINE_64 )
{
pixf->sad[PIXEL_8x16] = x264_pixel_sad_8x16_cache64_mmxext;
pixf->sad[PIXEL_8x8] = x264_pixel_sad_8x8_cache64_mmxext;
pixf->sad[PIXEL_8x4] = x264_pixel_sad_8x4_cache64_mmxext;
pixf->sad_x3[PIXEL_8x16] = x264_pixel_sad_x3_8x16_cache64_mmxext;
pixf->sad_x3[PIXEL_8x8] = x264_pixel_sad_x3_8x8_cache64_mmxext;
pixf->sad_x4[PIXEL_8x16] = x264_pixel_sad_x4_8x16_cache64_mmxext;
pixf->sad_x4[PIXEL_8x8] = x264_pixel_sad_x4_8x8_cache64_mmxext;
}
#endif
pixf->intra_satd_x3_16x16 = x264_intra_satd_x3_16x16_mmxext;
pixf->intra_satd_x3_8x8c = x264_intra_satd_x3_8x8c_mmxext;
pixf->intra_satd_x3_4x4 = x264_intra_satd_x3_4x4_mmxext;
}
if( (cpu&X264_CPU_SSE2) && !(cpu&X264_CPU_SSE2_IS_SLOW) )
{
INIT2( sad, _sse2 );
INIT2( sad_x3, _sse2 );
INIT2( sad_x4, _sse2 );
INIT_ADS( _sse2 );
pixf->var[PIXEL_8x8] = x264_pixel_var_8x8_sse2;
#ifdef ARCH_X86
if( cpu&X264_CPU_CACHELINE_64 )
{
INIT2( sad, _cache64_sse2 );
INIT2( sad_x3, _cache64_sse2 );
INIT2( sad_x4, _cache64_sse2 );
}
#endif
}
if( cpu&X264_CPU_SSE2 )
{
INIT5( ssd, _sse2 );
INIT5( satd, _sse2 );
INIT5( satd_x3, _sse2 );
INIT5( satd_x4, _sse2 );
pixf->var[PIXEL_16x16] = x264_pixel_var_16x16_sse2;
pixf->ssim_4x4x2_core = x264_pixel_ssim_4x4x2_core_sse2;
pixf->ssim_end4 = x264_pixel_ssim_end4_sse2;
pixf->sa8d[PIXEL_16x16] = x264_pixel_sa8d_16x16_sse2;
pixf->sa8d[PIXEL_8x8] = x264_pixel_sa8d_8x8_sse2;
#ifdef ARCH_X86_64
pixf->intra_sa8d_x3_8x8 = x264_intra_sa8d_x3_8x8_sse2;
#endif
}
if( (cpu&X264_CPU_SSE3) && (cpu&X264_CPU_CACHELINE_64) )
{
INIT2( sad, _sse3 );
INIT2( sad_x3, _sse3 );
INIT2( sad_x4, _sse3 );
}
if( cpu&X264_CPU_SSSE3 )
{
INIT7( satd, _ssse3 );
INIT7( satd_x3, _ssse3 );
INIT7( satd_x4, _ssse3 );
INIT_ADS( _ssse3 );
pixf->sa8d[PIXEL_16x16]= x264_pixel_sa8d_16x16_ssse3;
pixf->sa8d[PIXEL_8x8] = x264_pixel_sa8d_8x8_ssse3;
pixf->intra_satd_x3_16x16 = x264_intra_satd_x3_16x16_ssse3;
pixf->intra_satd_x3_8x8c = x264_intra_satd_x3_8x8c_ssse3;
pixf->intra_satd_x3_4x4 = x264_intra_satd_x3_4x4_ssse3;
#ifdef ARCH_X86_64
pixf->intra_sa8d_x3_8x8 = x264_intra_sa8d_x3_8x8_ssse3;
#endif
if( cpu&X264_CPU_CACHELINE_64 )
{
INIT2( sad, _cache64_ssse3 );
INIT2( sad_x3, _cache64_ssse3 );
INIT2( sad_x4, _cache64_ssse3 );
}
if( cpu&X264_CPU_PHADD_IS_FAST )
{
INIT5( satd, _ssse3_phadd );
INIT5( satd_x3, _ssse3_phadd );
INIT5( satd_x4, _ssse3_phadd );
}
}
#endif //HAVE_MMX
#ifdef ARCH_PPC
if( cpu&X264_CPU_ALTIVEC )
{
x264_pixel_altivec_init( pixf );
}
#endif
#ifdef ARCH_UltraSparc
INIT4( sad, _vis );
INIT4( sad_x3, _vis );
INIT4( sad_x4, _vis );
#endif
pixf->ads[PIXEL_8x16] =
pixf->ads[PIXEL_8x4] =
pixf->ads[PIXEL_4x8] = pixf->ads[PIXEL_16x8];
pixf->ads[PIXEL_4x4] = pixf->ads[PIXEL_8x8];
}
