/*****************************************************************************
* predict-c.c: intra prediction
*****************************************************************************
- * Copyright (C) 2003-2013 x264 project
+ * Copyright (C) 2003-2015 x264 project
*
* Authors: Laurent Aimar <fenrir@via.ecp.fr>
* Loren Merritt <lorenm@u.washington.edu>
PREDICT_16x16_DC( mmx2 )
PREDICT_16x16_DC( sse2 )
+PREDICT_16x16_DC( avx2 )
#define PREDICT_16x16_DC_LEFT(name)\
static void x264_predict_16x16_dc_left_##name( pixel *src )\
PREDICT_16x16_DC_LEFT( mmx2 )
PREDICT_16x16_DC_LEFT( sse2 )
+PREDICT_16x16_DC_LEFT( avx2 )
#define PREDICT_P_SUM(j,i)\
H += i * ( src[j+i - FDEC_STRIDE ] - src[j-i - FDEC_STRIDE ] );\
- V += i * ( src[(j+i)*FDEC_STRIDE -1] - src[(j-i)*FDEC_STRIDE -1] );\
+ V += i * ( src[(j+i)*FDEC_STRIDE -1] - src[(j-i)*FDEC_STRIDE -1] );
+#if HAVE_X86_INLINE_ASM
+#if HIGH_BIT_DEPTH
ALIGNED_16( static const int16_t pw_12345678[8] ) = {1,2,3,4,5,6,7,8};
ALIGNED_16( static const int16_t pw_m87654321[8] ) = {-8,-7,-6,-5,-4,-3,-2,-1};
ALIGNED_16( static const int16_t pw_m32101234[8] ) = {-3,-2,-1,0,1,2,3,4};
+#else // !HIGH_BIT_DEPTH
ALIGNED_8( static const int8_t pb_12345678[8] ) = {1,2,3,4,5,6,7,8};
ALIGNED_8( static const int8_t pb_m87654321[8] ) = {-8,-7,-6,-5,-4,-3,-2,-1};
ALIGNED_8( static const int8_t pb_m32101234[8] ) = {-3,-2,-1,0,1,2,3,4};
+#endif // HIGH_BIT_DEPTH
+#endif // HAVE_X86_INLINE_ASM
-#if !HIGH_BIT_DEPTH
-#define PREDICT_16x16_P(name)\
-static void x264_predict_16x16_p_##name( pixel *src )\
-{\
- int a, b, c;\
+#define PREDICT_16x16_P_CORE\
int H = 0;\
int V = 0;\
- int i00;\
- PREDICT_P_SUM(7,1) \
- PREDICT_P_SUM(7,2) \
- PREDICT_P_SUM(7,3) \
- PREDICT_P_SUM(7,4) \
- PREDICT_P_SUM(7,5) \
- PREDICT_P_SUM(7,6) \
- PREDICT_P_SUM(7,7) \
- PREDICT_P_SUM(7,8) \
- a = 16 * ( src[15*FDEC_STRIDE -1] + src[15 - FDEC_STRIDE] );\
- b = ( 5 * H + 32 ) >> 6;\
- c = ( 5 * V + 32 ) >> 6;\
- i00 = a - b * 7 - c * 7 + 16;\
- x264_predict_16x16_p_core_##name( src, i00, b, c );\
-}
-#ifndef ARCH_X86_64
-PREDICT_16x16_P( mmx2 )
-#endif
-PREDICT_16x16_P( sse2 )
-PREDICT_16x16_P( avx )
-#endif //!HIGH_BIT_DEPTH
-
-#define PREDICT_8x16C_P_CORE \
- int H = 0, V = 0;\
- for( int i = 0; i < 4; i++ )\
- H += ( i + 1 ) * ( src[4 + i - FDEC_STRIDE] - src[2 - i - FDEC_STRIDE] );\
- for( int i = 0; i < 8; i++ )\
- V += ( i + 1 ) * ( src[-1 + (i+8)*FDEC_STRIDE] - src[-1 + (6-i)*FDEC_STRIDE] );\
- int a = 16 * ( src[-1 + 15*FDEC_STRIDE] + src[7 - FDEC_STRIDE] );\
- int b = ( 17 * H + 16 ) >> 5;\
- int c = ( 5 * V + 32 ) >> 6;
+ PREDICT_P_SUM(7,1)\
+ PREDICT_P_SUM(7,2)\
+ PREDICT_P_SUM(7,3)\
+ PREDICT_P_SUM(7,4)\
+ PREDICT_P_SUM(7,5)\
+ PREDICT_P_SUM(7,6)\
+ PREDICT_P_SUM(7,7)\
+ PREDICT_P_SUM(7,8)
-#if HIGH_BIT_DEPTH
-#define PREDICT_8x16_P(name)\
-static void x264_predict_8x16c_p_##name( uint16_t *src )\
-{\
- PREDICT_8x16C_P_CORE \
- x264_predict_8x16c_p_core_##name( src, a, b, c );\
-}
+#define PREDICT_16x16_P_END(name)\
+ int a = 16 * ( src[15*FDEC_STRIDE -1] + src[15 - FDEC_STRIDE] );\
+ int b = ( 5 * H + 32 ) >> 6;\
+ int c = ( 5 * V + 32 ) >> 6;\
+ int i00 = a - b * 7 - c * 7 + 16;\
+ /* b*15 + c*15 can overflow: it's easier to just branch away in this rare case
+ * than to try to consider it in the asm. */\
+ if( BIT_DEPTH > 8 && (i00 > 0x7fff || abs(b) > 1092 || abs(c) > 1092) )\
+ x264_predict_16x16_p_c( src );\
+ else\
+ x264_predict_16x16_p_core_##name( src, i00, b, c );
-PREDICT_8x16_P(sse2)
-PREDICT_8x16_P(avx)
-#else
-#define PREDICT_8x16_P(name)\
-static void x264_predict_8x16c_p_##name( uint8_t *src )\
+#define PREDICT_16x16_P(name, name2)\
+static void x264_predict_16x16_p_##name( pixel *src )\
{\
- PREDICT_8x16C_P_CORE \
- int i00 = a -3*b -7*c + 16;\
- x264_predict_8x16c_p_core_##name( src, i00, b, c );\
+ PREDICT_16x16_P_CORE\
+ PREDICT_16x16_P_END(name2)\
}
-#ifndef ARCH_X86_64
-PREDICT_8x16_P(mmx2)
-#endif
-PREDICT_8x16_P(sse2)
-PREDICT_8x16_P(avx)
-#endif
#if HAVE_X86_INLINE_ASM
#if HIGH_BIT_DEPTH
-static void x264_predict_16x16_p_sse2( uint16_t *src )
-#else
-static void x264_predict_16x16_p_ssse3( uint8_t *src )
-#endif
-{
- int a, b, c, i00;
- int H, V;
-#if HIGH_BIT_DEPTH
- asm (
- "movdqu %1, %%xmm1 \n"
- "movdqa %2, %%xmm0 \n"
- "pmaddwd %3, %%xmm0 \n"
- "pmaddwd %4, %%xmm1 \n"
- "paddd %%xmm1, %%xmm0 \n"
- "movhlps %%xmm0, %%xmm1 \n"
- "paddd %%xmm1, %%xmm0 \n"
- "pshuflw $14, %%xmm0, %%xmm1 \n"
- "paddd %%xmm1, %%xmm0 \n"
- "movd %%xmm0, %0 \n"
- :"=r"(H)
- :"m"(src[-FDEC_STRIDE-1]), "m"(src[-FDEC_STRIDE+8]),
- "m"(*pw_12345678), "m"(*pw_m87654321)
+#define PREDICT_16x16_P_ASM\
+ asm (\
+ "movdqu %1, %%xmm1 \n"\
+ "movdqa %2, %%xmm0 \n"\
+ "pmaddwd %3, %%xmm0 \n"\
+ "pmaddwd %4, %%xmm1 \n"\
+ "paddd %%xmm1, %%xmm0 \n"\
+ "movhlps %%xmm0, %%xmm1 \n"\
+ "paddd %%xmm1, %%xmm0 \n"\
+ "pshuflw $14, %%xmm0, %%xmm1 \n"\
+ "paddd %%xmm1, %%xmm0 \n"\
+ "movd %%xmm0, %0 \n"\
+ :"=r"(H)\
+ :"m"(src[-FDEC_STRIDE-1]), "m"(src[-FDEC_STRIDE+8]),\
+ "m"(*pw_12345678), "m"(*pw_m87654321)\
);
-#else
- asm (
- "movq %1, %%mm1 \n"
- "movq %2, %%mm0 \n"
- "palignr $7, %3, %%mm1 \n"
- "pmaddubsw %4, %%mm0 \n"
- "pmaddubsw %5, %%mm1 \n"
- "paddw %%mm1, %%mm0 \n"
- "pshufw $14, %%mm0, %%mm1 \n"
- "paddw %%mm1, %%mm0 \n"
- "pshufw $1, %%mm0, %%mm1 \n"
- "paddw %%mm1, %%mm0 \n"
- "movd %%mm0, %0 \n"
- "movswl %w0, %0 \n"
- :"=r"(H)
- :"m"(src[-FDEC_STRIDE]), "m"(src[-FDEC_STRIDE+8]),
- "m"(src[-FDEC_STRIDE-8]), "m"(*pb_12345678), "m"(*pb_m87654321)
+#else // !HIGH_BIT_DEPTH
+#define PREDICT_16x16_P_ASM\
+ asm (\
+ "movq %1, %%mm1 \n"\
+ "movq %2, %%mm0 \n"\
+ "palignr $7, %3, %%mm1 \n"\
+ "pmaddubsw %4, %%mm0 \n"\
+ "pmaddubsw %5, %%mm1 \n"\
+ "paddw %%mm1, %%mm0 \n"\
+ "pshufw $14, %%mm0, %%mm1 \n"\
+ "paddw %%mm1, %%mm0 \n"\
+ "pshufw $1, %%mm0, %%mm1 \n"\
+ "paddw %%mm1, %%mm0 \n"\
+ "movd %%mm0, %0 \n"\
+ "movswl %w0, %0 \n"\
+ :"=r"(H)\
+ :"m"(src[-FDEC_STRIDE]), "m"(src[-FDEC_STRIDE+8]),\
+ "m"(src[-FDEC_STRIDE-8]), "m"(*pb_12345678), "m"(*pb_m87654321)\
);
-#endif
- V = 8 * ( src[15*FDEC_STRIDE-1] - src[-1*FDEC_STRIDE-1] )
- + 7 * ( src[14*FDEC_STRIDE-1] - src[ 0*FDEC_STRIDE-1] )
- + 6 * ( src[13*FDEC_STRIDE-1] - src[ 1*FDEC_STRIDE-1] )
- + 5 * ( src[12*FDEC_STRIDE-1] - src[ 2*FDEC_STRIDE-1] )
- + 4 * ( src[11*FDEC_STRIDE-1] - src[ 3*FDEC_STRIDE-1] )
- + 3 * ( src[10*FDEC_STRIDE-1] - src[ 4*FDEC_STRIDE-1] )
- + 2 * ( src[ 9*FDEC_STRIDE-1] - src[ 5*FDEC_STRIDE-1] )
+#endif // HIGH_BIT_DEPTH
+
+#define PREDICT_16x16_P_CORE_INLINE\
+ int H, V;\
+ PREDICT_16x16_P_ASM\
+ V = 8 * ( src[15*FDEC_STRIDE-1] - src[-1*FDEC_STRIDE-1] )\
+ + 7 * ( src[14*FDEC_STRIDE-1] - src[ 0*FDEC_STRIDE-1] )\
+ + 6 * ( src[13*FDEC_STRIDE-1] - src[ 1*FDEC_STRIDE-1] )\
+ + 5 * ( src[12*FDEC_STRIDE-1] - src[ 2*FDEC_STRIDE-1] )\
+ + 4 * ( src[11*FDEC_STRIDE-1] - src[ 3*FDEC_STRIDE-1] )\
+ + 3 * ( src[10*FDEC_STRIDE-1] - src[ 4*FDEC_STRIDE-1] )\
+ + 2 * ( src[ 9*FDEC_STRIDE-1] - src[ 5*FDEC_STRIDE-1] )\
+ 1 * ( src[ 8*FDEC_STRIDE-1] - src[ 6*FDEC_STRIDE-1] );
- a = 16 * ( src[15*FDEC_STRIDE -1] + src[15 - FDEC_STRIDE] );
- b = ( 5 * H + 32 ) >> 6;
- c = ( 5 * V + 32 ) >> 6;
- i00 = a - b * 7 - c * 7 + 16;
- /* b*15 + c*15 can overflow: it's easier to just branch away in this rare case
- * than to try to consider it in the asm. */
- if( BIT_DEPTH > 8 && (i00 > 0x7fff || abs(b) > 1092 || abs(c) > 1092) )
- x264_predict_16x16_p_c( src );
- else
- x264_predict_16x16_p_core_sse2( src, i00, b, c );
+
+#define PREDICT_16x16_P_INLINE(name, name2)\
+static void x264_predict_16x16_p_##name( pixel *src )\
+{\
+ PREDICT_16x16_P_CORE_INLINE\
+ PREDICT_16x16_P_END(name2)\
}
-#endif
+#else // !HAVE_X86_INLINE_ASM
+#define PREDICT_16x16_P_INLINE(name, name2) PREDICT_16x16_P(name, name2)
+#endif // HAVE_X86_INLINE_ASM
-#if !HIGH_BIT_DEPTH
+#if HIGH_BIT_DEPTH
+PREDICT_16x16_P_INLINE( sse2, sse2 )
+#else // !HIGH_BIT_DEPTH
+#if !ARCH_X86_64
+PREDICT_16x16_P( mmx2, mmx2 )
+#endif // !ARCH_X86_64
+PREDICT_16x16_P( sse2, sse2 )
+#if HAVE_X86_INLINE_ASM
+PREDICT_16x16_P_INLINE( ssse3, sse2 )
+#endif // HAVE_X86_INLINE_ASM
+PREDICT_16x16_P_INLINE( avx, avx )
+#endif // HIGH_BIT_DEPTH
+PREDICT_16x16_P_INLINE( avx2, avx2 )
+
+#define PREDICT_8x16C_P_CORE\
+ int H = 0, V = 0;\
+ for( int i = 0; i < 4; i++ )\
+ H += ( i + 1 ) * ( src[4 + i - FDEC_STRIDE] - src[2 - i - FDEC_STRIDE] );\
+ for( int i = 0; i < 8; i++ )\
+ V += ( i + 1 ) * ( src[-1 + (i+8)*FDEC_STRIDE] - src[-1 + (6-i)*FDEC_STRIDE] );
+
+#if HIGH_BIT_DEPTH
+#define PREDICT_8x16C_P_END(name)\
+ int a = 16 * ( src[-1 + 15*FDEC_STRIDE] + src[7 - FDEC_STRIDE] );\
+ int b = ( 17 * H + 16 ) >> 5;\
+ int c = ( 5 * V + 32 ) >> 6;\
+ x264_predict_8x16c_p_core_##name( src, a, b, c );
+#else // !HIGH_BIT_DEPTH
+#define PREDICT_8x16C_P_END(name)\
+ int a = 16 * ( src[-1 + 15*FDEC_STRIDE] + src[7 - FDEC_STRIDE] );\
+ int b = ( 17 * H + 16 ) >> 5;\
+ int c = ( 5 * V + 32 ) >> 6;\
+ int i00 = a -3*b -7*c + 16;\
+ x264_predict_8x16c_p_core_##name( src, i00, b, c );
+#endif // HIGH_BIT_DEPTH
-#define PREDICT_8x8_P(name)\
-static void x264_predict_8x8c_p_##name( uint8_t *src )\
+#define PREDICT_8x16C_P(name)\
+static void x264_predict_8x16c_p_##name( pixel *src )\
{\
- int a, b, c;\
+ PREDICT_8x16C_P_CORE\
+ PREDICT_8x16C_P_END(name)\
+}
+
+#if !ARCH_X86_64 && !HIGH_BIT_DEPTH
+PREDICT_8x16C_P( mmx2 )
+#endif // !ARCH_X86_64 && !HIGH_BIT_DEPTH
+PREDICT_8x16C_P( sse2 )
+PREDICT_8x16C_P( avx )
+PREDICT_8x16C_P( avx2 )
+
+#define PREDICT_8x8C_P_CORE\
int H = 0;\
int V = 0;\
- int i00;\
PREDICT_P_SUM(3,1)\
PREDICT_P_SUM(3,2)\
PREDICT_P_SUM(3,3)\
- PREDICT_P_SUM(3,4)\
- a = 16 * ( src[7*FDEC_STRIDE -1] + src[7 - FDEC_STRIDE] );\
- b = ( 17 * H + 16 ) >> 5;\
- c = ( 17 * V + 16 ) >> 5;\
- i00 = a -3*b -3*c + 16;\
- x264_predict_8x8c_p_core_##name( src, i00, b, c );\
-}
-#ifndef ARCH_X86_64
-PREDICT_8x8_P( mmx2 )
-#endif
-PREDICT_8x8_P( sse2 )
-
-#endif //!HIGH_BIT_DEPTH
-
-#if HAVE_X86_INLINE_ASM
+ PREDICT_P_SUM(3,4)
-#define PREDICT_8x8C_P_CORE\
- V = 1 * ( src[4*FDEC_STRIDE -1] - src[ 2*FDEC_STRIDE -1] )\
- + 2 * ( src[5*FDEC_STRIDE -1] - src[ 1*FDEC_STRIDE -1] )\
- + 3 * ( src[6*FDEC_STRIDE -1] - src[ 0*FDEC_STRIDE -1] )\
- + 4 * ( src[7*FDEC_STRIDE -1] - src[-1*FDEC_STRIDE -1] );\
- H += -4 * src[-1*FDEC_STRIDE -1];\
+#if HIGH_BIT_DEPTH
+#define PREDICT_8x8C_P_END(name)\
+ int a = 16 * ( src[7*FDEC_STRIDE -1] + src[7 - FDEC_STRIDE] );\
+ int b = ( 17 * H + 16 ) >> 5;\
+ int c = ( 17 * V + 16 ) >> 5;\
+ x264_predict_8x8c_p_core_##name( src, a, b, c );
+#else // !HIGH_BIT_DEPTH
+#define PREDICT_8x8C_P_END(name)\
int a = 16 * ( src[7*FDEC_STRIDE -1] + src[7 - FDEC_STRIDE] );\
int b = ( 17 * H + 16 ) >> 5;\
- int c = ( 17 * V + 16 ) >> 5;
+ int c = ( 17 * V + 16 ) >> 5;\
+ int i00 = a -3*b -3*c + 16;\
+ x264_predict_8x8c_p_core_##name( src, i00, b, c );
+#endif // HIGH_BIT_DEPTH
-#if HIGH_BIT_DEPTH
-#define PREDICT_8x8_P2(cpu1, cpu2)\
-static void x264_predict_8x8c_p_ ## cpu1( pixel *src )\
+#define PREDICT_8x8C_P(name, name2)\
+static void x264_predict_8x8c_p_##name( pixel *src )\
{\
- int H, V;\
+ PREDICT_8x8C_P_CORE\
+ PREDICT_8x8C_P_END(name2)\
+}
+
+#if HAVE_X86_INLINE_ASM
+#if HIGH_BIT_DEPTH
+#define PREDICT_8x8C_P_ASM\
asm (\
"movdqa %1, %%xmm0 \n"\
"pmaddwd %2, %%xmm0 \n"\
"movd %%xmm0, %0 \n"\
:"=r"(H)\
:"m"(src[-FDEC_STRIDE]), "m"(*pw_m32101234)\
- );\
- PREDICT_8x8C_P_CORE\
- x264_predict_8x8c_p_core_ ## cpu2( src, a, b, c );\
-}
-
-PREDICT_8x8_P2(sse2, sse2)
-PREDICT_8x8_P2( avx, avx)
-
-#else //!HIGH_BIT_DEPTH
-#define PREDICT_8x8_P2(cpu1, cpu2)\
-static void x264_predict_8x8c_p_ ## cpu1( pixel *src )\
-{\
- int H, V;\
+ );
+#else // !HIGH_BIT_DEPTH
+#define PREDICT_8x8C_P_ASM\
asm (\
"movq %1, %%mm0 \n"\
"pmaddubsw %2, %%mm0 \n"\
"movswl %w0, %0 \n"\
:"=r"(H)\
:"m"(src[-FDEC_STRIDE]), "m"(*pb_m32101234)\
- );\
- PREDICT_8x8C_P_CORE\
- int i00 = a -3*b -3*c + 16;\
- x264_predict_8x8c_p_core_ ## cpu2( src, i00, b, c );\
+ );
+#endif // HIGH_BIT_DEPTH
+
+#define PREDICT_8x8C_P_CORE_INLINE\
+ int H, V;\
+ PREDICT_8x8C_P_ASM\
+ V = 1 * ( src[4*FDEC_STRIDE -1] - src[ 2*FDEC_STRIDE -1] )\
+ + 2 * ( src[5*FDEC_STRIDE -1] - src[ 1*FDEC_STRIDE -1] )\
+ + 3 * ( src[6*FDEC_STRIDE -1] - src[ 0*FDEC_STRIDE -1] )\
+ + 4 * ( src[7*FDEC_STRIDE -1] - src[-1*FDEC_STRIDE -1] );\
+ H += -4 * src[-1*FDEC_STRIDE -1];
+
+#define PREDICT_8x8C_P_INLINE(name, name2)\
+static void x264_predict_8x8c_p_##name( pixel *src )\
+{\
+ PREDICT_8x8C_P_CORE_INLINE\
+ PREDICT_8x8C_P_END(name2)\
}
+#else // !HAVE_X86_INLINE_ASM
+#define PREDICT_8x8C_P_INLINE(name, name2) PREDICT_8x8C_P(name, name2)
+#endif // HAVE_X86_INLINE_ASM
-PREDICT_8x8_P2(ssse3, sse2)
-PREDICT_8x8_P2( avx, avx)
-#endif
-#endif
+#if HIGH_BIT_DEPTH
+PREDICT_8x8C_P_INLINE( sse2, sse2 )
+#else //!HIGH_BIT_DEPTH
+#if !ARCH_X86_64
+PREDICT_8x8C_P( mmx2, mmx2 )
+#endif // !ARCH_X86_64
+PREDICT_8x8C_P( sse2, sse2 )
+#if HAVE_X86_INLINE_ASM
+PREDICT_8x8C_P_INLINE( ssse3, sse2 )
+#endif // HAVE_X86_INLINE_ASM
+#endif // HIGH_BIT_DEPTH
+PREDICT_8x8C_P_INLINE( avx, avx )
+PREDICT_8x8C_P_INLINE( avx2, avx2 )
#if ARCH_X86_64 && !HIGH_BIT_DEPTH
static void x264_predict_8x8c_dc_left( uint8_t *src )
M64( src ) = dc1;
src += FDEC_STRIDE;
}
-
}
#endif // ARCH_X86_64 && !HIGH_BIT_DEPTH
pf[I_PRED_16x16_DC_TOP] = x264_predict_16x16_dc_top_sse2;
pf[I_PRED_16x16_DC_LEFT] = x264_predict_16x16_dc_left_sse2;
pf[I_PRED_16x16_H] = x264_predict_16x16_h_sse2;
-#if HAVE_X86_INLINE_ASM
pf[I_PRED_16x16_P] = x264_predict_16x16_p_sse2;
-#endif
if( !(cpu&X264_CPU_AVX) )
return;
pf[I_PRED_16x16_V] = x264_predict_16x16_v_avx;
return;
pf[I_PRED_16x16_P] = x264_predict_16x16_p_avx;
#endif // HIGH_BIT_DEPTH
+
+ if( cpu&X264_CPU_AVX2 )
+ {
+ pf[I_PRED_16x16_P] = x264_predict_16x16_p_avx2;
+ pf[I_PRED_16x16_DC] = x264_predict_16x16_dc_avx2;
+ pf[I_PRED_16x16_DC_TOP] = x264_predict_16x16_dc_top_avx2;
+ pf[I_PRED_16x16_DC_LEFT] = x264_predict_16x16_dc_left_avx2;
+ }
}
void x264_predict_8x8c_init_mmx( int cpu, x264_predict_t pf[7] )
pf[I_PRED_CHROMA_DC] = x264_predict_8x8c_dc_sse2;
pf[I_PRED_CHROMA_DC_TOP] = x264_predict_8x8c_dc_top_sse2;
pf[I_PRED_CHROMA_H] = x264_predict_8x8c_h_sse2;
-#if HAVE_X86_INLINE_ASM
pf[I_PRED_CHROMA_P] = x264_predict_8x8c_p_sse2;
if( !(cpu&X264_CPU_AVX) )
return;
if( !(cpu&X264_CPU_AVX2) )
return;
pf[I_PRED_CHROMA_H] = x264_predict_8x8c_h_avx2;
-#endif
#else
#if ARCH_X86_64
pf[I_PRED_CHROMA_DC_LEFT] = x264_predict_8x8c_dc_left;
pf[I_PRED_CHROMA_H] = x264_predict_8x8c_h_ssse3;
#if HAVE_X86_INLINE_ASM
pf[I_PRED_CHROMA_P] = x264_predict_8x8c_p_ssse3;
+#endif
if( !(cpu&X264_CPU_AVX) )
return;
pf[I_PRED_CHROMA_P] = x264_predict_8x8c_p_avx;
-#endif
#endif // HIGH_BIT_DEPTH
+
+ if( cpu&X264_CPU_AVX2 )
+ {
+ pf[I_PRED_CHROMA_P] = x264_predict_8x8c_p_avx2;
+ }
}
void x264_predict_8x16c_init_mmx( int cpu, x264_predict_t pf[7] )
pf[I_PRED_CHROMA_DC_TOP] = x264_predict_8x16c_dc_top_mmx2;
pf[I_PRED_CHROMA_DC] = x264_predict_8x16c_dc_mmx2;
pf[I_PRED_CHROMA_H] = x264_predict_8x16c_h_mmx2;
-#ifndef ARCH_X86_64
+#if !ARCH_X86_64
pf[I_PRED_CHROMA_P] = x264_predict_8x16c_p_mmx2;
#endif
if( !(cpu&X264_CPU_SSE2) )
return;
pf[I_PRED_CHROMA_P] = x264_predict_8x16c_p_avx;
#endif // HIGH_BIT_DEPTH
+
+ if( cpu&X264_CPU_AVX2 )
+ {
+ pf[I_PRED_CHROMA_P] = x264_predict_8x16c_p_avx2;
+ }
}
void x264_predict_8x8_init_mmx( int cpu, x264_predict8x8_t pf[12], x264_predict_8x8_filter_t *predict_8x8_filter )
projects / x264 / blobdiff