diff --git a/source/Lib/CommonLib/CommonDef.h b/source/Lib/CommonLib/CommonDef.h
index a9244427b39c930a470f36806028e13d02f5956d..66bca8c8439c8b781e195b0fad6993c78eb419c5 100644
--- a/source/Lib/CommonLib/CommonDef.h
+++ b/source/Lib/CommonLib/CommonDef.h
@@ -802,7 +802,11 @@ static const int COM16_C806_TRANS_PREC = 0;
#else
static const int NTAPS_LUMA = 8; ///< Number of taps for luma
#endif
+#if JVET_Z0117_CHROMA_IF
+static const int NTAPS_CHROMA = 6; ///< Number of taps for chroma
+#else
static const int NTAPS_CHROMA = 4; ///< Number of taps for chroma
+#endif
#if LUMA_ADAPTIVE_DEBLOCKING_FILTER_QP_OFFSET
static const int MAX_LADF_INTERVALS = 5; /// max number of luma adaptive deblocking filter qp offset intervals
#endif
diff --git a/source/Lib/CommonLib/InterpolationFilter.cpp b/source/Lib/CommonLib/InterpolationFilter.cpp
index d0eebf531d0245b93da3408c60e78a2e2d2e784c..abb80caf38de09b181616700b56f1a2331e3a891 100644
--- a/source/Lib/CommonLib/InterpolationFilter.cpp
+++ b/source/Lib/CommonLib/InterpolationFilter.cpp
@@ -344,7 +344,47 @@ const TFilterCoeff InterpolationFilter::m_lumaAltHpelIFilter[NTAPS_LUMA] = { 0,
#endif
#if IF_12TAP
+#if JVET_Z0117_CHROMA_IF
const TFilterCoeff InterpolationFilter::m_chromaFilter[CHROMA_INTERPOLATION_FILTER_SUB_SAMPLE_POSITIONS][NTAPS_CHROMA] =
+{
+ {0, 0, 256, 0, 0, 0},
+ {1, -6, 256, 7, -2, 0},
+ {2, -11, 253, 15, -4, 1},
+ {3, -16, 251, 23, -6, 1},
+ {4, -21, 248, 33, -10, 2},
+ {5, -25, 244, 42, -12, 2},
+ {7, -30, 239, 53, -17, 4},
+ {7, -32, 234, 62, -19, 4},
+ {8, -35, 227, 73, -22, 5},
+ {9, -38, 220, 84, -26, 7},
+ {10, -40, 213, 95, -29, 7},
+ {10, -41, 204, 106, -31, 8},
+ {10, -42, 196, 117, -34, 9},
+ {10, -41, 187, 127, -35, 8},
+ {11, -42, 177, 138, -38, 10},
+ {10, -41, 168, 148, -39, 10},
+ {10, -40, 158, 158, -40, 10},
+ {10, -39, 148, 168, -41, 10},
+ {10, -38, 138, 177, -42, 11},
+ {8, -35, 127, 187, -41, 10},
+ {9, -34, 117, 196, -42, 10},
+ {8, -31, 106, 204, -41, 10},
+ {7, -29, 95, 213, -40, 10},
+ {7, -26, 84, 220, -38, 9},
+ {5, -22, 73, 227, -35, 8},
+ {4, -19, 62, 234, -32, 7},
+ {4, -17, 53, 239, -30, 7},
+ {2, -12, 42, 244, -25, 5},
+ {2, -10, 33, 248, -21, 4},
+ {1, -6, 23, 251, -16, 3},
+ {1, -4, 15, 253, -11, 2},
+ {0, -2, 7, 256, -6, 1},
+};
+
+const TFilterCoeff InterpolationFilter::m_chromaFilter4[CHROMA_INTERPOLATION_FILTER_SUB_SAMPLE_POSITIONS][4] =
+#else
+const TFilterCoeff InterpolationFilter::m_chromaFilter[CHROMA_INTERPOLATION_FILTER_SUB_SAMPLE_POSITIONS][NTAPS_CHROMA] =
+#endif
{
{ 0 * 4, 64 * 4, 0 * 4, 0 * 4 },
{ -1 * 4, 63 * 4, 2 * 4, 0 * 4 },
@@ -917,6 +957,22 @@ InterpolationFilter::InterpolationFilter()
m_filterHor[1][1][0] = filter<8, false, true, false>;
m_filterHor[1][1][1] = filter<8, false, true, true>;
+#if JVET_Z0117_CHROMA_IF
+ m_filterHor[2][0][0] = filter<6, false, false, false>;
+ m_filterHor[2][0][1] = filter<6, false, false, true>;
+ m_filterHor[2][1][0] = filter<6, false, true, false>;
+ m_filterHor[2][1][1] = filter<6, false, true, true>;
+
+ m_filterHor[3][0][0] = filter<4, false, false, false>;
+ m_filterHor[3][0][1] = filter<4, false, false, true>;
+ m_filterHor[3][1][0] = filter<4, false, true, false>;
+ m_filterHor[3][1][1] = filter<4, false, true, true>;
+
+ m_filterHor[4][0][0] = filter<2, false, false, false>;
+ m_filterHor[4][0][1] = filter<2, false, false, true>;
+ m_filterHor[4][1][0] = filter<2, false, true, false>;
+ m_filterHor[4][1][1] = filter<2, false, true, true>;
+#else
m_filterHor[2][0][0] = filter<4, false, false, false>;
m_filterHor[2][0][1] = filter<4, false, false, true>;
m_filterHor[2][1][0] = filter<4, false, true, false>;
@@ -926,6 +982,7 @@ InterpolationFilter::InterpolationFilter()
m_filterHor[3][0][1] = filter<2, false, false, true>;
m_filterHor[3][1][0] = filter<2, false, true, false>;
m_filterHor[3][1][1] = filter<2, false, true, true>;
+#endif
m_filterVer[0][0][0] = filter<12, true, false, false>;
m_filterVer[0][0][1] = filter<12, true, false, true>;
@@ -937,6 +994,22 @@ InterpolationFilter::InterpolationFilter()
m_filterVer[1][1][0] = filter<8, true, true, false>;
m_filterVer[1][1][1] = filter<8, true, true, true>;
+#if JVET_Z0117_CHROMA_IF
+ m_filterVer[2][0][0] = filter<6, true, false, false>;
+ m_filterVer[2][0][1] = filter<6, true, false, true>;
+ m_filterVer[2][1][0] = filter<6, true, true, false>;
+ m_filterVer[2][1][1] = filter<6, true, true, true>;
+
+ m_filterVer[3][0][0] = filter<4, true, false, false>;
+ m_filterVer[3][0][1] = filter<4, true, false, true>;
+ m_filterVer[3][1][0] = filter<4, true, true, false>;
+ m_filterVer[3][1][1] = filter<4, true, true, true>;
+
+ m_filterVer[4][0][0] = filter<2, true, false, false>;
+ m_filterVer[4][0][1] = filter<2, true, false, true>;
+ m_filterVer[4][1][0] = filter<2, true, true, false>;
+ m_filterVer[4][1][1] = filter<2, true, true, true>;
+#else
m_filterVer[2][0][0] = filter<4, true, false, false>;
m_filterVer[2][0][1] = filter<4, true, false, true>;
m_filterVer[2][1][0] = filter<4, true, true, false>;
@@ -947,6 +1020,7 @@ InterpolationFilter::InterpolationFilter()
m_filterVer[3][1][0] = filter<2, true, true, false>;
m_filterVer[3][1][1] = filter<2, true, true, true>;
#endif
+#endif
#else
#if !QC_SIF_SIMD
m_filterHor[0][0][0] = filter<8, false, false, false>;
@@ -1374,6 +1448,20 @@ void InterpolationFilter::filterHor(const ClpRng& clpRng, Pel const *src, int sr
{
m_filterHor[1][1][isLast](clpRng, src, srcStride, dst, dstStride, width, height, coeff, biMCForDMVR);
}
+#if JVET_Z0117_CHROMA_IF
+ else if( N == 6 )
+ {
+ m_filterHor[2][1][isLast](clpRng, src, srcStride, dst, dstStride, width, height, coeff, biMCForDMVR);
+ }
+ else if (N == 4)
+ {
+ m_filterHor[3][1][isLast](clpRng, src, srcStride, dst, dstStride, width, height, coeff, biMCForDMVR);
+ }
+ else if (N == 2)
+ {
+ m_filterHor[4][1][isLast](clpRng, src, srcStride, dst, dstStride, width, height, coeff, biMCForDMVR);
+ }
+#else
else if (N == 4)
{
m_filterHor[2][1][isLast](clpRng, src, srcStride, dst, dstStride, width, height, coeff, biMCForDMVR);
@@ -1382,11 +1470,26 @@ void InterpolationFilter::filterHor(const ClpRng& clpRng, Pel const *src, int sr
{
m_filterHor[3][1][isLast](clpRng, src, srcStride, dst, dstStride, width, height, coeff, biMCForDMVR);
}
+#endif
#else
if( N == 8 )
{
m_filterHor[0][1][isLast](clpRng, src, srcStride, dst, dstStride, width, height, coeff, biMCForDMVR);
}
+#if JVET_Z0117_CHROMA_IF
+ else if( N == 6 )
+ {
+ m_filterHor[1][1][isLast](clpRng, src, srcStride, dst, dstStride, width, height, coeff, biMCForDMVR);
+ }
+ else if (N == 4)
+ {
+ m_filterHor[2][1][isLast](clpRng, src, srcStride, dst, dstStride, width, height, coeff, biMCForDMVR);
+ }
+ else if (N == 2)
+ {
+ m_filterHor[3][1][isLast](clpRng, src, srcStride, dst, dstStride, width, height, coeff, biMCForDMVR);
+ }
+#else
else if( N == 4 )
{
m_filterHor[1][1][isLast](clpRng, src, srcStride, dst, dstStride, width, height, coeff, biMCForDMVR);
@@ -1395,6 +1498,7 @@ void InterpolationFilter::filterHor(const ClpRng& clpRng, Pel const *src, int sr
{
m_filterHor[2][1][isLast](clpRng, src, srcStride, dst, dstStride, width, height, coeff, biMCForDMVR);
}
+#endif
#endif
else
{
@@ -1430,6 +1534,20 @@ void InterpolationFilter::filterVer(const ClpRng& clpRng, Pel const *src, int sr
{
m_filterVer[1][isFirst][isLast](clpRng, src, srcStride, dst, dstStride, width, height, coeff, biMCForDMVR);
}
+#if JVET_Z0117_CHROMA_IF
+ else if (N == 6)
+ {
+ m_filterVer[2][isFirst][isLast](clpRng, src, srcStride, dst, dstStride, width, height, coeff, biMCForDMVR);
+ }
+ else if (N == 4)
+ {
+ m_filterVer[3][isFirst][isLast](clpRng, src, srcStride, dst, dstStride, width, height, coeff, biMCForDMVR);
+ }
+ else if (N == 2)
+ {
+ m_filterVer[4][isFirst][isLast](clpRng, src, srcStride, dst, dstStride, width, height, coeff, biMCForDMVR);
+ }
+#else
else if (N == 4)
{
m_filterVer[2][isFirst][isLast](clpRng, src, srcStride, dst, dstStride, width, height, coeff, biMCForDMVR);
@@ -1438,11 +1556,26 @@ void InterpolationFilter::filterVer(const ClpRng& clpRng, Pel const *src, int sr
{
m_filterVer[3][isFirst][isLast](clpRng, src, srcStride, dst, dstStride, width, height, coeff, biMCForDMVR);
}
+#endif
#else
if( N == 8 )
{
m_filterVer[0][isFirst][isLast]( clpRng, src, srcStride, dst, dstStride, width, height, coeff, biMCForDMVR);
}
+#if JVET_Z0117_CHROMA_IF
+ else if (N == 6)
+ {
+ m_filterVer[1][isFirst][isLast](clpRng, src, srcStride, dst, dstStride, width, height, coeff, biMCForDMVR);
+ }
+ else if (N == 4)
+ {
+ m_filterVer[2][isFirst][isLast](clpRng, src, srcStride, dst, dstStride, width, height, coeff, biMCForDMVR);
+ }
+ else if (N == 2)
+ {
+ m_filterVer[3][isFirst][isLast](clpRng, src, srcStride, dst, dstStride, width, height, coeff, biMCForDMVR);
+ }
+#else
else if( N == 4 )
{
m_filterVer[1][isFirst][isLast]( clpRng, src, srcStride, dst, dstStride, width, height, coeff, biMCForDMVR);
@@ -1451,6 +1584,7 @@ void InterpolationFilter::filterVer(const ClpRng& clpRng, Pel const *src, int sr
{
m_filterVer[2][isFirst][isLast]( clpRng, src, srcStride, dst, dstStride, width, height, coeff, biMCForDMVR);
}
+#endif
#endif
else
{
diff --git a/source/Lib/CommonLib/InterpolationFilter.h b/source/Lib/CommonLib/InterpolationFilter.h
index 0faecf4b68f1c6206646eccf43c28d35706d5c2f..f287c2a885111c86df2885ac7dabfe1924e7b82d 100644
--- a/source/Lib/CommonLib/InterpolationFilter.h
+++ b/source/Lib/CommonLib/InterpolationFilter.h
@@ -69,6 +69,9 @@ class InterpolationFilter
#endif
public:
static const TFilterCoeff m_chromaFilter[CHROMA_INTERPOLATION_FILTER_SUB_SAMPLE_POSITIONS][NTAPS_CHROMA]; ///< Chroma filter taps
+#if JVET_Z0117_CHROMA_IF
+ static const TFilterCoeff m_chromaFilter4[CHROMA_INTERPOLATION_FILTER_SUB_SAMPLE_POSITIONS][4]; ///< Chroma filter taps
+#endif
static const TFilterCoeff m_chromaFilterRPR1[CHROMA_INTERPOLATION_FILTER_SUB_SAMPLE_POSITIONS][NTAPS_CHROMA]; ///< Chroma filter taps 1.5x
static const TFilterCoeff m_chromaFilterRPR2[CHROMA_INTERPOLATION_FILTER_SUB_SAMPLE_POSITIONS][NTAPS_CHROMA]; ///< Chroma filter taps 2x
#if IF_12TAP
@@ -146,11 +149,21 @@ public:
void(*m_filter4x4[2])(const int16_t* src, int srcStride, int16_t *dst, int dstStride, int shiftH, int offsetH, int shiftV, int offsetV, int16_t const *coeffH, int16_t const *coeffV, int ibdimin, int ibdimax); //kolya
#endif
+#if JVET_Z0117_CHROMA_IF
+ void(*m_filterHor[5][2][2])(const ClpRng& clpRng, Pel const *src, int srcStride, Pel *dst, int dstStride, int width, int height, TFilterCoeff const *coeff, bool biMCForDMVR);
+ void(*m_filterVer[5][2][2])(const ClpRng& clpRng, Pel const *src, int srcStride, Pel *dst, int dstStride, int width, int height, TFilterCoeff const *coeff, bool biMCForDMVR);
+#else
+ void(*m_filterHor[4][2][2])(const ClpRng& clpRng, Pel const *src, int srcStride, Pel *dst, int dstStride, int width, int height, TFilterCoeff const *coeff, bool biMCForDMVR);
+ void(*m_filterVer[4][2][2])(const ClpRng& clpRng, Pel const *src, int srcStride, Pel *dst, int dstStride, int width, int height, TFilterCoeff const *coeff, bool biMCForDMVR);
+#endif
+#else
+#if JVET_Z0117_CHROMA_IF
void(*m_filterHor[4][2][2])(const ClpRng& clpRng, Pel const *src, int srcStride, Pel *dst, int dstStride, int width, int height, TFilterCoeff const *coeff, bool biMCForDMVR);
void(*m_filterVer[4][2][2])(const ClpRng& clpRng, Pel const *src, int srcStride, Pel *dst, int dstStride, int width, int height, TFilterCoeff const *coeff, bool biMCForDMVR);
#else
void(*m_filterHor[3][2][2])(const ClpRng& clpRng, Pel const *src, int srcStride, Pel *dst, int dstStride, int width, int height, TFilterCoeff const *coeff, bool biMCForDMVR);
void(*m_filterVer[3][2][2])(const ClpRng& clpRng, Pel const *src, int srcStride, Pel *dst, int dstStride, int width, int height, TFilterCoeff const *coeff, bool biMCForDMVR);
+#endif
#endif
void( *m_filterCopy[2][2] ) ( const ClpRng& clpRng, Pel const *src, int srcStride, Pel *dst, int dstStride, int width, int height, bool biMCForDMVR);
void( *m_weightedGeoBlk )(const PredictionUnit &pu, const uint32_t width, const uint32_t height, const ComponentID compIdx, const uint8_t splitDir, PelUnitBuf& predDst, PelUnitBuf& predSrc0, PelUnitBuf& predSrc1);
@@ -181,7 +194,11 @@ public:
static TFilterCoeff const * const getIntraLumaFilterTableExt(const int deltaFract) { return m_lumaIntraFilterExt[deltaFract]; };
#endif
#endif
+#if JVET_Z0117_CHROMA_IF
+ static TFilterCoeff const * const getChromaFilterTable(const int deltaFract) { return m_chromaFilter4[deltaFract]; };
+#else
static TFilterCoeff const * const getChromaFilterTable(const int deltaFract) { return m_chromaFilter[deltaFract]; };
+#endif
#if JVET_W0123_TIMD_FUSION
static TFilterCoeff const * const getExtIntraCubicFilter(const int deltaFract) { return g_aiExtIntraCubicFilter[deltaFract]; };
static TFilterCoeff const * const getExtIntraGaussFilter(const int deltaFract) { return g_aiExtIntraGaussFilter[deltaFract]; };
diff --git a/source/Lib/CommonLib/TypeDef.h b/source/Lib/CommonLib/TypeDef.h
index fdb94667f7bc0185271c7a47a184610120f546b8..21ec59f0546be63ef6d364aae45f5d7cb5356f0e 100644
--- a/source/Lib/CommonLib/TypeDef.h
+++ b/source/Lib/CommonLib/TypeDef.h
@@ -155,6 +155,7 @@
#define JVET_Z0127_SPS_MHP_MAX_MRG_CAND 1
#endif
#define IF_12TAP 1 // 12-tap IF
+#define JVET_Z0117_CHROMA_IF 1 //JVET_Z0117: 6-tap interpolation filter for chroma MC
#define ENABLE_OBMC 1 // Enable Overlapped Block Motion Compensation
#if JVET_X0049_BDMVR_SW_OPT
diff --git a/source/Lib/CommonLib/x86/InterpolationFilterX86.h b/source/Lib/CommonLib/x86/InterpolationFilterX86.h
index b3b76f563d0c07f6c084434fddb9c50a7ba6b39c..3ff1d56b57923c4484ae9b08cddb142556b8e7a6 100644
--- a/source/Lib/CommonLib/x86/InterpolationFilterX86.h
+++ b/source/Lib/CommonLib/x86/InterpolationFilterX86.h
@@ -519,6 +519,9 @@ static void simdInterpolateHorM4( const int16_t* src, int srcStride, int16_t *ds
__m128i vibdimax = _mm_set1_epi16( clpRng.max );
__m128i vcoeffh = _mm_lddqu_si128( ( __m128i const * )coeff );
+#if JVET_Z0117_CHROMA_IF
+ __m128i vcoeffh1;
+#endif
__m128i vzero, vshufc0, vshufc1;
__m128i vsum;
@@ -528,6 +531,28 @@ static void simdInterpolateHorM4( const int16_t* src, int srcStride, int16_t *ds
vshufc0 = _mm_set_epi8( 0x9, 0x8, 0x7, 0x6, 0x5, 0x4, 0x3, 0x2, 0x7, 0x6, 0x5, 0x4, 0x3, 0x2, 0x1, 0x0 );
vshufc1 = _mm_set_epi8( 0xd, 0xc, 0xb, 0xa, 0x9, 0x8, 0x7, 0x6, 0xb, 0xa, 0x9, 0x8, 0x7, 0x6, 0x5, 0x4 );
}
+#if JVET_Z0117_CHROMA_IF
+ int16_t c0[8];
+ int16_t c1[8];
+ __m128i vshufc2;
+ if(N == 6)
+ {
+ vshufc2 = _mm_set_epi8( 0x9, 0x8, 0x7, 0x6, 0x7, 0x6, 0x5, 0x4, 0x5, 0x4, 0x3, 0x2, 0x3, 0x2, 0x1, 0x0 );
+ for(int i = 0; i < 8; i += 4)
+ {
+ c0[i] = coeff[0];
+ c0[i+1] = coeff[1];
+ c0[i+2] = coeff[2];
+ c0[i+3] = coeff[3];
+ c1[i] = coeff[4];
+ c1[i+1] = coeff[5];
+ c1[i+2] = coeff[4];
+ c1[i+3] = coeff[5];
+ }
+ vcoeffh = _mm_lddqu_si128( ( __m128i const * )c0 );
+ vcoeffh1 = _mm_lddqu_si128( ( __m128i const * )c1 );
+ }
+#endif
vzero = _mm_setzero_si128();
@@ -549,6 +574,21 @@ static void simdInterpolateHorM4( const int16_t* src, int srcStride, int16_t *ds
}
vsum = _mm_hadd_epi32( vtmp[0], vtmp[1] );
}
+#if JVET_Z0117_CHROMA_IF
+ else if(N == 6)
+ {
+ __m128i vsrc0 = _mm_lddqu_si128( ( __m128i const * )&src[col] );
+ __m128i vsrc1 = _mm_lddqu_si128( ( __m128i const * )&src[col+4] );
+ __m128i temp1 = _mm_shuffle_epi8( vsrc0, vshufc0 );
+ __m128i temp2 = _mm_shuffle_epi8( vsrc0, vshufc1 );
+ __m128i temp3 = _mm_shuffle_epi8( vsrc1, vshufc2 );
+ temp1 = _mm_madd_epi16( temp1, vcoeffh );
+ temp2 = _mm_madd_epi16( temp2, vcoeffh );
+ temp3 = _mm_madd_epi16( temp3, vcoeffh1 );
+ temp1 = _mm_hadd_epi32(temp1, temp2);
+ vsum = _mm_add_epi32(temp1, temp3);
+ }
+#endif
else
{
__m128i vtmp0, vtmp1;
@@ -578,6 +618,68 @@ static void simdInterpolateHorM4( const int16_t* src, int srcStride, int16_t *ds
}
}
+#if JVET_Z0117_CHROMA_IF
+template<X86_VEXT vext, int N, bool shiftBack>
+static void simdInterpolateHorM4_sym( const int16_t* src, int srcStride, int16_t *dst, int dstStride, int width, int height, int shift, int offset, const ClpRng& clpRng, int16_t const *coeff )
+{
+ _mm_prefetch( (const char*)src + srcStride, _MM_HINT_T0 );
+ __m128i voffset = _mm_set1_epi32( offset );
+ __m128i vibdimin = _mm_set1_epi16( clpRng.min );
+ __m128i vibdimax = _mm_set1_epi16( clpRng.max );
+ __m128i vcoeffh = _mm_lddqu_si128( ( __m128i const * )coeff );
+ __m128i vcoeffh1;
+ __m128i vzero, vshufc0, vshufc1, vshufc2, vshufc3, vshufc4;
+ __m128i vsum;
+ vcoeffh = _mm_shuffle_epi32( vcoeffh, 0x44 );
+ vzero = _mm_setzero_si128();
+ vshufc0 = _mm_set_epi8( 0x9, 0x8, 0x7, 0x6, 0x5, 0x4, 0x3, 0x2, 0x7, 0x6, 0x5, 0x4, 0x3, 0x2, 0x1, 0x0 );
+ vshufc1 = _mm_set_epi8( 0xd, 0xc, 0xb, 0xa, 0x9, 0x8, 0x7, 0x6, 0xb, 0xa, 0x9, 0x8, 0x7, 0x6, 0x5, 0x4);
+ vshufc2 = _mm_set_epi8( 0x7, 0x6, 0x5, 0x4, 0x3, 0x2, 0x1, 0x0, 0x9, 0x8, 0x7, 0x6, 0x5, 0x4, 0x3, 0x2);
+ vshufc3 = _mm_set_epi8( 0xf, 0xe, 0x7, 0x6, 0xd, 0xc, 0x5, 0x4, 0xb, 0xa, 0x3, 0x2, 0x9, 0x8, 0x1, 0x0 );
+ vshufc4 = _mm_set_epi8( 0xf, 0xe, 0xd, 0xc, 0xb, 0xa, 0x9, 0x8, 0xd, 0xc, 0xb, 0xa, 0x9, 0x8, 0x7, 0x6);
+ int16_t c0[8];
+ for(int i = 0; i < 8; i += 4)
+ {
+ c0[i] = coeff[0];
+ c0[i+1] = coeff[1];
+ c0[i+2] = coeff[0];
+ c0[i+3] = coeff[1];
+ }
+ vcoeffh = _mm_lddqu_si128( ( __m128i const * )c0 );
+ vcoeffh1 = _mm_set1_epi32( coeff[2]);
+ vzero = _mm_setzero_si128();
+ for( int row = 0; row < height; row++ )
+ {
+ _mm_prefetch( (const char*)src + 2 * srcStride, _MM_HINT_T0 );
+ for( int col = 0; col < width; col += 4 )
+ {
+ __m128i vsrc0 = _mm_lddqu_si128( ( __m128i const * )&src[col] );
+ __m128i vsrc1 = _mm_lddqu_si128( ( __m128i const * )&src[col+4] );
+ __m128i temp1 = _mm_shuffle_epi8( vsrc0, vshufc0 );
+ __m128i temp3 = _mm_shuffle_epi8( vsrc1, vshufc2 );
+ __m128i temp2 = _mm_shuffle_epi8( vsrc0, vshufc1 );
+ __m128i temp4 = _mm_shuffle_epi8( vsrc0, vshufc4) ;
+ temp1 = _mm_add_epi16(temp1, temp3);
+ temp2 = _mm_add_epi16(temp2, temp4);
+ temp1 = _mm_shuffle_epi8(temp1, vshufc3);
+ temp1 = _mm_madd_epi16( temp1, vcoeffh );
+ temp2 = _mm_cvtepi16_epi32(temp2);
+ temp2 = _mm_mullo_epi32(temp2, vcoeffh1);
+ vsum = _mm_add_epi32(temp1, temp2);
+ vsum = _mm_add_epi32( vsum, voffset );
+ vsum = _mm_srai_epi32( vsum, shift );
+ vsum = _mm_packs_epi32( vsum, vzero );
+ if( shiftBack )
+ {
+ vsum = _mm_min_epi16( vibdimax, _mm_max_epi16( vibdimin, vsum ) );
+ }
+ _mm_storel_epi64( ( __m128i * )&dst[col], vsum );
+ }
+ src += srcStride;
+ dst += dstStride;
+ }
+}
+#endif
// SIMD interpolation horizontal, block width modulo 8
template<X86_VEXT vext, int N, bool shiftBack>
@@ -595,13 +697,37 @@ static void simdInterpolateHorM8( const int16_t* src, int srcStride, int16_t *ds
__m128i vshufc0, vshufc1;
__m128i vsum, vsuma, vsumb;
-
+#if JVET_Z0117_CHROMA_IF
+ __m128i vcoeffh1;
+#endif
if( N != 8 ){
vcoeffh = _mm_shuffle_epi32( vcoeffh, 0x44 );
vshufc0 = _mm_set_epi8( 0x9, 0x8, 0x7, 0x6, 0x5, 0x4, 0x3, 0x2, 0x7, 0x6, 0x5, 0x4, 0x3, 0x2, 0x1, 0x0 );
vshufc1 = _mm_set_epi8( 0xd, 0xc, 0xb, 0xa, 0x9, 0x8, 0x7, 0x6, 0xb, 0xa, 0x9, 0x8, 0x7, 0x6, 0x5, 0x4 );
}
+#if JVET_Z0117_CHROMA_IF
+ int16_t c0[8];
+ int16_t c1[8];
+ __m128i vshufc2;
+ if(N == 6)
+ {
+ vshufc2 = _mm_set_epi8( 0x9, 0x8, 0x7, 0x6, 0x7, 0x6, 0x5, 0x4, 0x5, 0x4, 0x3, 0x2, 0x3, 0x2, 0x1, 0x0 );
+ for(int i = 0; i < 8; i += 4)
+ {
+ c0[i] = coeff[0];
+ c0[i+1] = coeff[1];
+ c0[i+2] = coeff[2];
+ c0[i+3] = coeff[3];
+ c1[i] = coeff[4];
+ c1[i+1] = coeff[5];
+ c1[i+2] = coeff[4];
+ c1[i+3] = coeff[5];
+ }
+ vcoeffh = _mm_lddqu_si128( ( __m128i const * )c0 );
+ vcoeffh1 = _mm_lddqu_si128( ( __m128i const * )c1 );
+ }
+#endif
for( int row = 0; row < height; row++ )
{
_mm_prefetch( (const char*)src + 2 * srcStride, _MM_HINT_T0 );
@@ -622,6 +748,27 @@ static void simdInterpolateHorM8( const int16_t* src, int srcStride, int16_t *ds
vsuma = _mm_hadd_epi32( vtmp[0], vtmp[1] );
vsumb = _mm_hadd_epi32( vtmp[2], vtmp[3] );
}
+#if JVET_Z0117_CHROMA_IF
+ else if(N == 6)
+ {
+ __m128i vtmp[2];
+ for( int i = 0; i < 8; i += 4 )
+ {
+ __m128i vsrc0 = _mm_lddqu_si128( ( __m128i const * )&src[col + i] );
+ __m128i vsrc1 = _mm_lddqu_si128( ( __m128i const * )&src[col + i + 4] );
+ __m128i temp1 = _mm_shuffle_epi8( vsrc0, vshufc0 );
+ __m128i temp2 = _mm_shuffle_epi8( vsrc0, vshufc1 );
+ __m128i temp3 = _mm_shuffle_epi8( vsrc1, vshufc2 );
+ temp1 = _mm_madd_epi16( temp1, vcoeffh );
+ temp2 = _mm_madd_epi16( temp2, vcoeffh );
+ temp3 = _mm_madd_epi16( temp3, vcoeffh1 );
+ temp1 = _mm_hadd_epi32(temp1, temp2);
+ vtmp[i / 4] = _mm_add_epi32(temp1, temp3);
+ }
+ vsuma = vtmp[0];
+ vsumb = vtmp[1];
+ }
+#endif
else
{
__m128i vtmp00, vtmp01, vtmp10, vtmp11;
@@ -1010,11 +1157,40 @@ static void simdInterpolateHorM8_AVX2( const int16_t* src, int srcStride, int16_
__m256i vshuf1 = _mm256_set_epi8( 0xd, 0xc, 0xb, 0xa, 0xb, 0xa, 0x9, 0x8, 0x9, 0x8, 0x7, 0x6, 0x7, 0x6, 0x5, 0x4,
0xd, 0xc, 0xb, 0xa, 0xb, 0xa, 0x9, 0x8, 0x9, 0x8, 0x7, 0x6, 0x7, 0x6, 0x5, 0x4 );
+#if JVET_Z0117_CHROMA_IF
+ __m256i vcoeff[6];
+ int16_t c0[8];
+ int16_t c1[8];
+ if(N == 6)
+ {
+ for(int i = 0; i < 8; i += 4)
+ {
+ c0[i] = coeff[0];
+ c0[i+1] = coeff[1];
+ c0[i+2] = coeff[2];
+ c0[i+3] = coeff[3];
+ c1[i] = coeff[4];
+ c1[i+1] = coeff[5];
+ c1[i+2] = coeff[4];
+ c1[i+3] = coeff[5];
+ }
+ __m128i vcoeff0 = _mm_lddqu_si128( ( __m128i const * )c0 );
+ __m128i vcoeff1 = _mm_lddqu_si128( ( __m128i const * )c1 );
+ vcoeff[0] = _mm256_insertf128_si256(_mm256_castsi128_si256(vcoeff0), (vcoeff0), 1);
+ vcoeff[1] = _mm256_insertf128_si256(_mm256_castsi128_si256(vcoeff1), (vcoeff1), 1);
+ }
+ else
+ {
+#else
__m256i vcoeff[N/2];
+#endif
for( int i=0; i<N; i+=2 )
{
vcoeff[i/2] = _mm256_unpacklo_epi16( _mm256_set1_epi16( coeff[i] ), _mm256_set1_epi16( coeff[i+1] ) );
}
+#if JVET_Z0117_CHROMA_IF
+ }
+#endif
// __m256i vw0, vw1, voffsetW, vwp;
__m256i vsum;
@@ -1043,6 +1219,34 @@ static void simdInterpolateHorM8_AVX2( const int16_t* src, int srcStride, int16_
vsum = _mm256_add_epi32( vsum, _mm256_add_epi32( _mm256_madd_epi16( vsrca0, vcoeff[2*i] ), _mm256_madd_epi16( vsrca1, vcoeff[2*i+1] ) ) );
}
}
+#if JVET_Z0117_CHROMA_IF
+ else if( N==6 )
+ {
+ __m128i vshuf2 = _mm_set_epi8(0x9, 0x8, 0x7, 0x6, 0x5, 0x4, 0x3, 0x2,0x7, 0x6, 0x5, 0x4, 0x3, 0x2, 0x1, 0x0 );
+ __m128i vshuf4 = _mm_set_epi8( 0x9, 0x8, 0x7, 0x6,
+ 0x7, 0x6, 0x5, 0x4,
+ 0x5, 0x4, 0x3, 0x2,
+ 0x3, 0x2, 0x1, 0x0 );
+ __m128i vshuf3 = _mm_set_epi8(0xd, 0xc, 0xb, 0xa, 0x9, 0x8, 0x7, 0x6,0xb, 0xa, 0x9, 0x8, 0x7, 0x6, 0x5, 0x4);
+ __m128i t0 = _mm_lddqu_si128( ( const __m128i* )&src[col] );
+ __m128i t2 = _mm_lddqu_si128( ( const __m128i* )&src[col+8]);
+ __m128i t1 = _mm_alignr_epi8(t2, t0, 8);
+ __m128i b0 = _mm_shuffle_epi8(t1, vshuf4);
+ __m128i b1 = _mm_shuffle_epi8(t2, vshuf4);
+ __m128i c0 = _mm_shuffle_epi8(t0, vshuf2);
+ __m128i c1 = _mm_shuffle_epi8(t1, vshuf2);
+ __m128i d0 = _mm_shuffle_epi8(t0, vshuf3);
+ __m128i d1 = _mm_shuffle_epi8(t1, vshuf3);
+ __m256i vsrca0 = _mm256_insertf128_si256(_mm256_castsi128_si256(c0), (c1), 1);
+ __m256i vsrca1 = _mm256_insertf128_si256(_mm256_castsi128_si256(d0), (d1), 1);
+ __m256i vsrca2 = _mm256_insertf128_si256(_mm256_castsi128_si256(b0), (b1), 1);
+ vsrca0 = _mm256_madd_epi16(vsrca0, vcoeff[0]);
+ vsrca1 = _mm256_madd_epi16(vsrca1, vcoeff[0]);
+ vsrca2 = _mm256_madd_epi16(vsrca2, vcoeff[1]);
+ vsrca0 = _mm256_hadd_epi32(vsrca0, vsrca1);
+ vsum = _mm256_add_epi32(vsrca0, vsrca2);
+ }
+#endif
else
{
__m256i vtmp02, vtmp13;
@@ -1078,6 +1282,85 @@ static void simdInterpolateHorM8_AVX2( const int16_t* src, int srcStride, int16_
}
#endif
}
+#if JVET_Z0117_CHROMA_IF
+template<X86_VEXT vext, int N, bool shiftBack>
+static void simdInterpolateHorM8_sym_AVX2( const int16_t* src, int srcStride, int16_t *dst, int dstStride, int width, int height, int shift, int offset, const ClpRng& clpRng, int16_t const *coeff )
+{
+#ifdef USE_AVX2
+ const int filterSpan =/* bChromaIntl ? 2* ( N-1 ) : */( N-1 );
+ _mm_prefetch( (const char*)( src+srcStride ), _MM_HINT_T0 );
+ _mm_prefetch( (const char*)( src+( width>>1 )+srcStride ), _MM_HINT_T0 );
+ _mm_prefetch( (const char*)( src+width+filterSpan+srcStride ), _MM_HINT_T0 );
+ __m256i voffset = _mm256_set1_epi32( offset );
+ __m128i vibdimin = _mm_set1_epi16( clpRng.min );
+ __m128i vibdimax = _mm_set1_epi16( clpRng.max );
+ __m256i vcoeff[2];
+ int16_t c0[8];
+ if(N == 6)
+ {
+ for(int i = 0; i < 8; i += 4)
+ {
+ c0[i] = coeff[0];
+ c0[i+1] = coeff[1];
+ c0[i+2] = coeff[0];
+ c0[i+3] = coeff[1];
+
+ }
+ __m128i vcoeff0 = _mm_lddqu_si128( ( __m128i const * )c0 );
+ vcoeff[0] = _mm256_insertf128_si256(_mm256_castsi128_si256(vcoeff0), (vcoeff0), 1);
+ vcoeff[1] = _mm256_set1_epi32(coeff[2]);
+
+ }
+
+ __m256i vsum;
+ for( int row = 0; row < height; row++ )
+ {
+ _mm_prefetch( (const char*)( src+2*srcStride ), _MM_HINT_T0 );
+ _mm_prefetch( (const char*)( src+( width>>1 )+2*srcStride ), _MM_HINT_T0 );
+ _mm_prefetch( (const char*)( src+width+filterSpan + 2*srcStride ), _MM_HINT_T0 );
+ for( int col = 0; col < width; col+=8 )
+ {
+ if( N==6 )
+ {
+ __m128i t0 = _mm_lddqu_si128( ( const __m128i* )&src[col] );
+ __m128i t6 = _mm_lddqu_si128( ( const __m128i* )&src[col+8]);
+ __m128i t1 = _mm_alignr_epi8(t6, t0, 2);
+ __m128i t2 = _mm_alignr_epi8(t6, t0, 4);
+ __m128i t3 = _mm_alignr_epi8(t6, t0, 6);
+ __m128i t4 = _mm_alignr_epi8(t6, t0, 8);
+ __m128i t5 = _mm_alignr_epi8(t6, t0, 10);
+ t0 = _mm_add_epi16(t5, t0);
+ t1 = _mm_add_epi16(t4, t1);
+ t2 = _mm_add_epi16(t3, t2);
+ t3 = _mm_unpacklo_epi16(t0, t1);
+ t4 = _mm_unpackhi_epi16(t0, t1);
+ __m256i vsrca0 = _mm256_insertf128_si256(_mm256_castsi128_si256(t3), (t4), 1);
+ vsrca0 = _mm256_madd_epi16(vsrca0, vcoeff[0]);
+ __m256i vsrca1 = _mm256_cvtepi16_epi32(t2);
+ vsrca1 = _mm256_mullo_epi32(vsrca1, vcoeff[1]);
+ vsum = _mm256_add_epi32(vsrca0, vsrca1 );
+ }
+
+ vsum = _mm256_add_epi32( vsum, voffset );
+ vsum = _mm256_srai_epi32( vsum, shift );
+
+ __m128i vsump = _mm256_castsi256_si128( _mm256_permute4x64_epi64( _mm256_packs_epi32( vsum, vsum ), 0x88 ) );
+ if( shiftBack )
+ { //clip
+ vsump = _mm_min_epi16( vibdimax, _mm_max_epi16( vibdimin, vsump ) );
+ }
+#if JEM_UNALIGNED_DST
+ _mm_storeu_si128( ( __m128i * )&dst[col], vsump );
+#else
+ _mm_store_si128( ( __m128i * )&dst[col], vsump );
+#endif
+ }
+ src += srcStride;
+ dst += dstStride;
+ }
+#endif
+}
+#endif
#if MCIF_SIMD_NEW
template<X86_VEXT vext, int N, bool shiftBack>
static void simdInterpolateHorM16_AVX2(const int16_t* src, int srcStride, int16_t *dst, int dstStride, int width, int height, int shift, int offset, const ClpRng& clpRng, int16_t const *coeff)
@@ -1106,13 +1389,44 @@ static void simdInterpolateHorM16_AVX2(const int16_t* src, int srcStride, int16_
0xd, 0xc, 0xb, 0xa, 0xb, 0xa, 0x9, 0x8, 0x9, 0x8, 0x7, 0x6, 0x7, 0x6, 0x5, 0x4);
#if __INTEL_COMPILER
__m256i vcoeff[4];
+#else
+#if JVET_Z0117_CHROMA_IF
+ __m256i vcoeff[6];
#else
__m256i vcoeff[N / 2];
+#endif
+#endif
+#if JVET_Z0117_CHROMA_IF
+ int16_t c0[8];
+ int16_t c1[8];
+ if(N == 6)
+ {
+ for(int i = 0; i < 8; i += 4)
+ {
+ c0[i] = coeff[0];
+ c0[i+1] = coeff[1];
+ c0[i+2] = coeff[2];
+ c0[i+3] = coeff[3];
+ c1[i] = coeff[4];
+ c1[i+1] = coeff[5];
+ c1[i+2] = coeff[4];
+ c1[i+3] = coeff[5];
+ }
+ __m128i vcoeff0 = _mm_lddqu_si128( ( __m128i const * )c0 );
+ __m128i vcoeff1 = _mm_lddqu_si128( ( __m128i const * )c1 );
+ vcoeff[0] = _mm256_insertf128_si256(_mm256_castsi128_si256(vcoeff0), (vcoeff0), 1);
+ vcoeff[1] = _mm256_insertf128_si256(_mm256_castsi128_si256(vcoeff1), (vcoeff1), 1);
+ }
+ else
+ {
#endif
for (int i = 0; i < N; i += 2)
{
vcoeff[i / 2] = _mm256_unpacklo_epi16(_mm256_set1_epi16(coeff[i]), _mm256_set1_epi16(coeff[i + 1]));
}
+#if JVET_Z0117_CHROMA_IF
+ }
+#endif
for (int row = 0; row < height; row++)
{
@@ -1144,6 +1458,53 @@ static void simdInterpolateHorM16_AVX2(const int16_t* src, int srcStride, int16_
vsuma = _mm256_add_epi32(vsuma, vsrc1);
vsumb = _mm256_add_epi32(vsumb, vsrc0);
}
+#if JVET_Z0117_CHROMA_IF
+ else if( N == 6 )
+ {
+ __m128i vshuf2 = _mm_set_epi8(0x9, 0x8, 0x7, 0x6, 0x5, 0x4, 0x3, 0x2,0x7, 0x6, 0x5, 0x4, 0x3, 0x2, 0x1, 0x0 );
+ __m128i vshuf4 = _mm_set_epi8( 0x9, 0x8, 0x7, 0x6,
+ 0x7, 0x6, 0x5, 0x4,
+ 0x5, 0x4, 0x3, 0x2,
+ 0x3, 0x2, 0x1, 0x0 );
+ __m128i vshuf3 = _mm_set_epi8(0xd, 0xc, 0xb, 0xa, 0x9, 0x8, 0x7, 0x6,0xb, 0xa, 0x9, 0x8, 0x7, 0x6, 0x5, 0x4);
+ vsuma = vsumb = vzero;
+ __m128i t0 = _mm_lddqu_si128( ( const __m128i* )&src[col] );
+ __m128i t2 = _mm_lddqu_si128( ( const __m128i* )&src[col+8]);
+ __m128i t4 = _mm_lddqu_si128( ( const __m128i* )&src[col+16]);
+ __m128i t1, t3, b0, b1, b2, b3, c0, c1, c2, c3, d0, d1, d2, d3;
+ __m256i vsrca0, vsrca1, vsrca2;
+ t1 = _mm_alignr_epi8(t2, t0, 8);
+ t3 = _mm_alignr_epi8(t4, t2, 8);
+ b0 = _mm_shuffle_epi8(t1, vshuf4);
+ b1 = _mm_shuffle_epi8(t2, vshuf4);
+ c0 = _mm_shuffle_epi8(t0, vshuf2);
+ c1 = _mm_shuffle_epi8(t1, vshuf2);
+ d0 = _mm_shuffle_epi8(t0, vshuf3);
+ d1 = _mm_shuffle_epi8(t1, vshuf3);
+ b2 = _mm_shuffle_epi8(t3, vshuf4);
+ b3 = _mm_shuffle_epi8(t4, vshuf4);
+ c2 = _mm_shuffle_epi8(t2, vshuf2);
+ c3 = _mm_shuffle_epi8(t3, vshuf2);
+ d2 = _mm_shuffle_epi8(t2, vshuf3);
+ d3 = _mm_shuffle_epi8(t3, vshuf3);
+ vsrca0 = _mm256_insertf128_si256(_mm256_castsi128_si256(c0), (c2), 1);
+ vsrca1 = _mm256_insertf128_si256(_mm256_castsi128_si256(d0), (d2), 1);
+ vsrca2 = _mm256_insertf128_si256(_mm256_castsi128_si256(b0), (b2), 1);
+ vsrca0 = _mm256_madd_epi16(vsrca0, vcoeff[0]);
+ vsrca1 = _mm256_madd_epi16(vsrca1, vcoeff[0]);
+ vsrca2 = _mm256_madd_epi16(vsrca2, vcoeff[1]);
+ vsrca0 = _mm256_hadd_epi32(vsrca0, vsrca1);
+ vsuma = _mm256_add_epi32(vsrca0, vsrca2);
+ vsrca0 = _mm256_insertf128_si256(_mm256_castsi128_si256(c1), (c3), 1);
+ vsrca1 = _mm256_insertf128_si256(_mm256_castsi128_si256(d1), (d3), 1);
+ vsrca2 = _mm256_insertf128_si256(_mm256_castsi128_si256(b1), (b3), 1);
+ vsrca0 = _mm256_madd_epi16(vsrca0, vcoeff[0]);
+ vsrca1 = _mm256_madd_epi16(vsrca1, vcoeff[0]);
+ vsrca2 = _mm256_madd_epi16(vsrca2, vcoeff[1]);
+ vsrca0 = _mm256_hadd_epi32(vsrca0, vsrca1);
+ vsumb = _mm256_add_epi32(vsrca0, vsrca2);
+ }
+#endif
else
{
__m256i vsrc0 = _mm256_loadu_si256((const __m256i *)&src[col]);
@@ -1190,6 +1551,114 @@ static void simdInterpolateHorM16_AVX2(const int16_t* src, int srcStride, int16_
_mm256_zeroupper();
#endif
}
+#if JVET_Z0117_CHROMA_IF
+template<X86_VEXT vext, int N, bool shiftBack>
+static void simdInterpolateHorM16_sym_AVX2(const int16_t* src, int srcStride, int16_t *dst, int dstStride, int width, int height, int shift, int offset, const ClpRng& clpRng, int16_t const *coeff)
+{
+#ifdef USE_AVX2
+ const int filterSpan = (N - 1);
+ _mm_prefetch((const char*)(src + srcStride), _MM_HINT_T0);
+ _mm_prefetch((const char*)(src + (width >> 1) + srcStride), _MM_HINT_T0);
+ _mm_prefetch((const char*)(src + width + filterSpan + srcStride), _MM_HINT_T0);
+ _mm_prefetch((const char*)(src + 2 * srcStride), _MM_HINT_T0);
+ _mm_prefetch((const char*)(src + (width >> 1) + 2 * srcStride), _MM_HINT_T0);
+ _mm_prefetch((const char*)(src + width + filterSpan + 2 * srcStride), _MM_HINT_T0);
+ _mm_prefetch((const char*)(src + 3 * srcStride), _MM_HINT_T0);
+ _mm_prefetch((const char*)(src + (width >> 1) + 3 * srcStride), _MM_HINT_T0);
+ _mm_prefetch((const char*)(src + width + filterSpan + 3 * srcStride), _MM_HINT_T0);
+ __m256i voffset = _mm256_set1_epi32(offset);
+ __m256i vibdimin = _mm256_set1_epi16(clpRng.min);
+ __m256i vibdimax = _mm256_set1_epi16(clpRng.max);
+ __m256i vzero = _mm256_setzero_si256();
+ __m256i vsum, vsuma, vsumb;
+#if __INTEL_COMPILER
+ __m256i vcoeff[4];
+#else
+ __m256i vcoeff[6];
+#endif
+ int16_t c0[8];
+ for(int i = 0; i < 8; i += 4)
+ {
+ c0[i] = coeff[0];
+ c0[i+1] = coeff[1];
+ c0[i+2] = coeff[0];
+ c0[i+3] = coeff[1];
+
+ }
+ __m128i vcoeff0 = _mm_lddqu_si128( ( __m128i const * )c0 );
+ vcoeff[0] = _mm256_insertf128_si256(_mm256_castsi128_si256(vcoeff0), (vcoeff0), 1);
+ vcoeff[1] = _mm256_set1_epi32(coeff[2]);
+ for (int row = 0; row < height; row++)
+ {
+ _mm_prefetch((const char*)(src + 2 * srcStride), _MM_HINT_T0);
+ _mm_prefetch((const char*)(src + (width >> 1) + 2 * srcStride), _MM_HINT_T0);
+ _mm_prefetch((const char*)(src + width + filterSpan + 2 * srcStride), _MM_HINT_T0);
+ for (int col = 0; col < width; col += 16)
+ {
+ vsuma = vsumb = vzero;
+ __m128i t0 = _mm_lddqu_si128( ( const __m128i* )&src[col] );
+ __m128i t6 = _mm_lddqu_si128( ( const __m128i* )&src[col+8]);
+ __m128i t7 = _mm_lddqu_si128( ( const __m128i* )&src[col+16]);
+ __m128i t1 = _mm_alignr_epi8(t6, t0, 2);
+ __m128i t2 = _mm_alignr_epi8(t6, t0, 4);
+ __m128i t3 = _mm_alignr_epi8(t6, t0, 6);
+ __m128i t4 = _mm_alignr_epi8(t6, t0, 8);
+ __m128i t5 = _mm_alignr_epi8(t6, t0, 10);
+ t0 = _mm_add_epi16(t5, t0);
+ t1 = _mm_add_epi16(t4, t1);
+ t2 = _mm_add_epi16(t3, t2);
+ t3 = _mm_unpacklo_epi16(t0, t1);
+ t4 = _mm_unpackhi_epi16(t0, t1);
+ __m256i vsrca0 = _mm256_insertf128_si256(_mm256_castsi128_si256(t3), (t4), 1);
+ vsrca0 = _mm256_madd_epi16(vsrca0, vcoeff[0]);
+ __m256i vsrca1 = _mm256_cvtepi16_epi32(t2);
+ vsrca1 = _mm256_mullo_epi32(vsrca1, vcoeff[1]);
+ __m256i vsumatmp = _mm256_add_epi32(vsrca0, vsrca1 );
+
+ t1 = _mm_alignr_epi8(t7, t6, 2);
+ t2 = _mm_alignr_epi8(t7, t6, 4);
+ t3 = _mm_alignr_epi8(t7, t6, 6);
+ t4 = _mm_alignr_epi8(t7, t6, 8);
+ t5 = _mm_alignr_epi8(t7, t6, 10);
+ t0 = _mm_add_epi16(t5, t6);
+ t1 = _mm_add_epi16(t4, t1);
+ t2 = _mm_add_epi16(t3, t2);
+ t3 = _mm_unpacklo_epi16(t0, t1);
+ t4 = _mm_unpackhi_epi16(t0, t1);
+ vsrca0 = _mm256_insertf128_si256(_mm256_castsi128_si256(t3), (t4), 1);
+ vsrca0 = _mm256_madd_epi16(vsrca0, vcoeff[0]);
+
+ vsrca1 = _mm256_cvtepi16_epi32(t2);
+ vsrca1 = _mm256_mullo_epi32(vsrca1, vcoeff[1]);
+ __m256i vsumbtmp = _mm256_add_epi32(vsrca0, vsrca1 );
+ vsuma = _mm256_permute2f128_si256(vsumatmp, vsumbtmp, 0x20);
+ vsumb = _mm256_permute2f128_si256(vsumatmp, vsumbtmp, 0x31);
+
+ vsuma = _mm256_add_epi32(vsuma, voffset);
+ vsumb = _mm256_add_epi32(vsumb, voffset);
+ vsuma = _mm256_srai_epi32(vsuma, shift);
+ vsumb = _mm256_srai_epi32(vsumb, shift);
+ vsum = _mm256_packs_epi32(vsuma, vsumb);
+
+ if (shiftBack)
+ { //clip
+ vsum = _mm256_min_epi16(vibdimax, _mm256_max_epi16(vibdimin, vsum));
+ }
+#if JEM_UNALIGNED_DST
+ _mm256_storeu_si256((__m256i *)&dst[col], vsum);
+#else
+ _mm256_store_si256((__m256i *)&dst[col], vsum);
+#endif
+ }
+ src += srcStride;
+ dst += dstStride;
+ }
+#endif
+#if USE_AVX2
+ _mm256_zeroupper();
+#endif
+}
+#endif
#endif
#if !MCIF_SIMD_NEW
template<X86_VEXT vext, int N, bool shiftBack>
@@ -1455,6 +1924,82 @@ static void simdInterpolateVerNonM4_12tap(const int16_t *src, int srcStride, int
#endif
#endif
+#if JVET_Z0117_CHROMA_IF
+template<X86_VEXT vext, int N, bool shiftBack>
+static void simdInterpolateVerM4_sym( const int16_t *src, int srcStride, int16_t *dst, int dstStride, int width, int height, int shift, int offset, const ClpRng& clpRng, int16_t const *coeff )
+{
+#if MCIF_SIMD_NEW
+ _mm_prefetch((const char *)&src[0 * srcStride], _MM_HINT_T0);
+ _mm_prefetch((const char *)&src[1 * srcStride], _MM_HINT_T0);
+ if (N >= 2)
+ {
+ _mm_prefetch((const char *)&src[2 * srcStride], _MM_HINT_T0);
+ _mm_prefetch((const char *)&src[3 * srcStride], _MM_HINT_T0);
+ }
+ if (N >= 6)
+ {
+ _mm_prefetch((const char *)&src[4 * srcStride], _MM_HINT_T0);
+ _mm_prefetch((const char *)&src[5 * srcStride], _MM_HINT_T0);
+ }
+ if (N >= 8)
+ {
+ _mm_prefetch((const char *)&src[6 * srcStride], _MM_HINT_T0);
+ _mm_prefetch((const char *)&src[7 * srcStride], _MM_HINT_T0);
+ }
+#endif
+ const int16_t *srcOrig = src;
+ int16_t *dstOrig = dst;
+ __m128i vcoeff[2], vsrc[6];
+ __m128i vzero = _mm_setzero_si128();
+ __m128i voffset = _mm_set1_epi32( offset );
+ __m128i vibdimin = _mm_set1_epi16( clpRng.min );
+ __m128i vibdimax = _mm_set1_epi16( clpRng.max );
+ __m128i vsum;
+ vcoeff[0] = _mm_unpacklo_epi16( _mm_set1_epi16( coeff[0] ), _mm_set1_epi16( coeff[1] ) );
+ vcoeff[1] = _mm_set1_epi32(coeff[2]);
+
+ for( int col = 0; col < width; col += 4 )
+ {
+ vsrc[0] = _mm_loadl_epi64( ( __m128i const * )&src[col + 0 * srcStride] );
+ vsrc[1] = _mm_loadl_epi64( ( __m128i const * )&src[col + 1 * srcStride] );
+ vsrc[2] = _mm_loadl_epi64( ( __m128i const * )&src[col + 2 * srcStride] );
+ vsrc[3] = _mm_loadl_epi64( ( __m128i const * )&src[col + 3 * srcStride] );
+ vsrc[4] = _mm_loadl_epi64( ( __m128i const * )&src[col + 4 * srcStride] );
+ for( int row = 0; row < height; row++ )
+ {
+#if MCIF_SIMD_NEW
+ _mm_prefetch((const char *)&src[col + (N + 0) * srcStride], _MM_HINT_T0);
+ _mm_prefetch((const char *)&src[col + (N + 1) * srcStride], _MM_HINT_T0);
+#endif
+ vsrc[5] = _mm_loadl_epi64( ( __m128i const * )&src[col + ( N - 1 ) * srcStride] );
+ vsum = vzero;
+ __m128i vsrc0 = _mm_unpacklo_epi16( vsrc[0], vsrc[1] );
+ __m128i vsrc1 = _mm_unpacklo_epi16( vsrc[5], vsrc[4] );
+ vsrc0 = _mm_add_epi16(vsrc0 , vsrc1);
+ vsum = _mm_add_epi32( vsum, _mm_madd_epi16( vsrc0, vcoeff[0] ) );
+ __m128i vsrc2 = _mm_add_epi16( vsrc[2], vsrc[3] );
+ __m128i vsrc3 = _mm_cvtepi16_epi32(vsrc2);
+ vsum = _mm_add_epi32( vsum, _mm_mullo_epi32(vsrc3, vcoeff[1]));
+ vsrc[0] = vsrc[1];
+ vsrc[1] = vsrc[2];
+ vsrc[2] = vsrc[3];
+ vsrc[3] = vsrc[4];
+ vsum = _mm_add_epi32( vsum, voffset );
+ vsum = _mm_srai_epi32( vsum, shift );
+ vsum = _mm_packs_epi32( vsum, vzero );
+ if( shiftBack )
+ {
+ vsum = _mm_min_epi16( vibdimax, _mm_max_epi16( vibdimin, vsum ) );
+ }
+ _mm_storel_epi64( ( __m128i* )&dst[col], vsum );
+ src += srcStride;
+ dst += dstStride;
+ }
+ src = srcOrig;
+ dst = dstOrig;
+ }
+}
+#endif
template<X86_VEXT vext, int N, bool shiftBack>
static void simdInterpolateVerM4( const int16_t *src, int srcStride, int16_t *dst, int dstStride, int width, int height, int shift, int offset, const ClpRng& clpRng, int16_t const *coeff )
{
@@ -1649,6 +2194,90 @@ static void simdInterpolateVerM8( const int16_t *src, int srcStride, int16_t *ds
}
}
+#if JVET_Z0117_CHROMA_IF
+template<X86_VEXT vext, int N, bool shiftBack>
+static void simdInterpolateVerM8_sym_AVX2( const int16_t *src, int srcStride, int16_t *dst, int dstStride, int width, int height, int shift, int offset, const ClpRng& clpRng, int16_t const *coeff )
+{
+#ifdef USE_AVX2
+#if MCIF_SIMD_NEW
+ _mm_prefetch((const char *)&src[0 * srcStride], _MM_HINT_T0);
+ _mm_prefetch((const char *)&src[1 * srcStride], _MM_HINT_T0);
+ if (N >= 2)
+ {
+ _mm_prefetch((const char *)&src[2 * srcStride], _MM_HINT_T0);
+ _mm_prefetch((const char *)&src[3 * srcStride], _MM_HINT_T0);
+ }
+ if (N >= 6)
+ {
+ _mm_prefetch((const char *)&src[4 * srcStride], _MM_HINT_T0);
+ _mm_prefetch((const char *)&src[5 * srcStride], _MM_HINT_T0);
+ }
+ if (N >= 8)
+ {
+ _mm_prefetch((const char *)&src[6 * srcStride], _MM_HINT_T0);
+ _mm_prefetch((const char *)&src[7 * srcStride], _MM_HINT_T0);
+ }
+#endif
+ __m256i voffset = _mm256_set1_epi32( offset );
+ __m128i vibdimin = _mm_set1_epi16( clpRng.min );
+ __m128i vibdimax = _mm_set1_epi16( clpRng.max );
+ __m256i vsum;
+ __m256i vsrc[6];
+ __m256i vcoeff[2];
+ vcoeff[0] = _mm256_unpacklo_epi16( _mm256_set1_epi16( coeff[0] ), _mm256_set1_epi16( coeff[1] ) );
+ vcoeff[1] = _mm256_set1_epi32(coeff[2]);
+ const short *srcOrig = src;
+ int16_t *dstOrig = dst;
+ for( int col = 0; col < width; col+=8 )
+ {
+ for( int i=0; i<N-1; i++ )
+ {
+ vsrc[i]= _mm256_castsi128_si256( _mm_loadu_si128( ( const __m128i * )&src[col + i * srcStride] ) );
+ vsrc[i]= _mm256_permute4x64_epi64( vsrc[i], 0x50 );
+ }
+ for( int row = 0; row < height; row++ )
+ {
+#if MCIF_SIMD_NEW
+ _mm_prefetch((const char *)&src[col + (N + 0) * srcStride], _MM_HINT_T0);
+ _mm_prefetch((const char *)&src[col + (N + 1) * srcStride], _MM_HINT_T0);
+#endif
+ vsrc[5]= _mm256_castsi128_si256( _mm_loadu_si128( ( const __m128i * )&src[col + ( N-1 ) * srcStride] ) );
+ vsrc[5]= _mm256_permute4x64_epi64( vsrc[5], 0x50 );
+ vsum = _mm256_setzero_si256();
+ __m256i vsrc0 = _mm256_unpacklo_epi16( vsrc[0], vsrc[1] );
+ __m256i vsrc1 = _mm256_unpacklo_epi16( vsrc[5], vsrc[4] );
+ vsrc0 = _mm256_add_epi16(vsrc0, vsrc1);
+ vsum = _mm256_add_epi32( vsum, _mm256_madd_epi16( vsrc0, vcoeff[0] ) );
+ __m256i tmpa = _mm256_permute4x64_epi64(vsrc[2], 0x08);
+ __m256i tmpb = _mm256_permute4x64_epi64(vsrc[3], 0x08);
+ __m128i tmpc = _mm256_extracti128_si256(tmpa, 0);
+ __m128i tmpd = _mm256_extracti128_si256(tmpb, 0);
+ tmpc = _mm_add_epi16(tmpc, tmpd);
+ tmpa = _mm256_cvtepi16_epi32(tmpc);
+ tmpb = _mm256_mullo_epi32(tmpa, vcoeff[1]);
+ vsum = _mm256_add_epi32(vsum, tmpb);
+ vsrc[0] = vsrc[1];
+ vsrc[1] = vsrc[2];
+ vsrc[2] = vsrc[3];
+ vsrc[3] = vsrc[4];
+ vsrc[4] = vsrc[5];
+ vsum = _mm256_add_epi32( vsum, voffset );
+ vsum = _mm256_srai_epi32( vsum, shift );
+ __m128i vsump = _mm256_castsi256_si128( _mm256_permute4x64_epi64( _mm256_packs_epi32( vsum, vsum ), 0x88 ) );
+ if( shiftBack )
+ { //clip
+ vsump = _mm_min_epi16( vibdimax, _mm_max_epi16( vibdimin, vsump ) );
+ }
+ _mm_storeu_si128( ( __m128i * )&dst[col], vsump );
+ src += srcStride;
+ dst += dstStride;
+ }
+ src= srcOrig;
+ dst= dstOrig;
+ }
+#endif
+}
+#endif
// template<typename Tdst, int N, bool shiftBack, bool biPred, bool bWeight, bool bChromaIntl>
// static void qpelV16AVX2M8( const short* src, int srcStride, Tdst *dst, int dstStride, int width, int height, int shift, int bitdepth, short const *coeff, wpPredParam *pwp1=NULL, wpPredParam *pwp2=NULL )
template<X86_VEXT vext, int N, bool shiftBack>
@@ -1807,6 +2436,93 @@ static void simdInterpolateVerM16_AVX2( const int16_t *src, int srcStride, int16
#endif
#if MCIF_SIMD_NEW
+#if JVET_Z0117_CHROMA_IF
+template<X86_VEXT vext, int N, bool shiftBack>
+static void simdInterpolateVerM16_sym_AVX2(const int16_t *src, int srcStride, int16_t *dst, int dstStride, int width, int height, int shift, int offset, const ClpRng& clpRng, int16_t const *coeff)
+{
+#ifdef USE_AVX2
+ _mm_prefetch((const char *)&src[0 * srcStride], _MM_HINT_T0);
+ _mm_prefetch((const char *)&src[1 * srcStride], _MM_HINT_T0);
+ if (N >= 2)
+ {
+ _mm_prefetch((const char *)&src[2 * srcStride], _MM_HINT_T0);
+ _mm_prefetch((const char *)&src[3 * srcStride], _MM_HINT_T0);
+ }
+ if (N >= 6)
+ {
+ _mm_prefetch((const char *)&src[4 * srcStride], _MM_HINT_T0);
+ _mm_prefetch((const char *)&src[5 * srcStride], _MM_HINT_T0);
+ }
+ if (N >= 8)
+ {
+ _mm_prefetch((const char *)&src[6 * srcStride], _MM_HINT_T0);
+ _mm_prefetch((const char *)&src[7 * srcStride], _MM_HINT_T0);
+ }
+
+ __m256i voffset = _mm256_set1_epi32(offset);
+ __m256i vibdimin = _mm256_set1_epi16(clpRng.min);
+ __m256i vibdimax = _mm256_set1_epi16(clpRng.max);
+ __m256i vzero = _mm256_setzero_si256();
+ __m256i vsum, vsuma, vsumb;
+ __m256i vsrc[6];
+ __m256i vcoeff[2];
+ vcoeff[0] = _mm256_unpacklo_epi16(_mm256_set1_epi16(coeff[0]), _mm256_set1_epi16(coeff[1]));
+ vcoeff[1] = _mm256_set1_epi32(coeff[2]);
+ const short *srcOrig = src;
+ int16_t *dstOrig = dst;
+
+ for (int col = 0; col < width; col += 16)
+ {
+ vsrc[0] = _mm256_loadu_si256((const __m256i *)&src[col + 0 * srcStride]);
+ vsrc[1] = _mm256_loadu_si256((const __m256i *)&src[col + 1 * srcStride]);
+ vsrc[2] = _mm256_loadu_si256((const __m256i *)&src[col + 2 * srcStride]);
+ vsrc[3] = _mm256_loadu_si256((const __m256i *)&src[col + 3 * srcStride]);
+ vsrc[4] = _mm256_loadu_si256((const __m256i *)&src[col + 4 * srcStride]);
+ for (int row = 0; row < height; row++)
+ {
+ _mm_prefetch((const char *)&src[col + (N + 0) * srcStride], _MM_HINT_T0);
+ _mm_prefetch((const char *)&src[col + (N + 1) * srcStride], _MM_HINT_T0);
+ vsrc[5] = _mm256_loadu_si256((const __m256i *)&src[col + (N - 1) * srcStride]);
+ vsuma = vsumb = vzero;
+ __m256i tmp05 = _mm256_add_epi16(vsrc[0], vsrc[5]);
+ __m256i tmp14 = _mm256_add_epi16(vsrc[1], vsrc[4]);
+ __m256i tmp23 = _mm256_add_epi16(vsrc[2], vsrc[3]);
+ __m256i vsrca = _mm256_unpacklo_epi16(tmp05, tmp14);
+ __m256i vsrcb = _mm256_unpackhi_epi16(tmp05, tmp14);
+ vsuma = _mm256_add_epi32(vsuma, _mm256_madd_epi16(vsrca, vcoeff[0]));
+ vsumb = _mm256_add_epi32(vsumb, _mm256_madd_epi16(vsrcb, vcoeff[0]));
+ tmp23 = _mm256_permute4x64_epi64(tmp23, 0xd8);
+ vsrca = _mm256_cvtepi16_epi32(_mm256_extracti128_si256(tmp23, 0));
+ vsrcb = _mm256_cvtepi16_epi32(_mm256_extracti128_si256(tmp23, 1));
+ vsuma = _mm256_add_epi32(vsuma, _mm256_mullo_epi32(vsrca, vcoeff[1]));
+ vsumb = _mm256_add_epi32(vsumb, _mm256_mullo_epi32(vsrcb, vcoeff[1]));
+ vsrc[0] = vsrc[1];
+ vsrc[1] = vsrc[2];
+ vsrc[2] = vsrc[3];
+ vsrc[3] = vsrc[4];
+ vsrc[4] = vsrc[5];
+ vsuma = _mm256_add_epi32(vsuma, voffset);
+ vsumb = _mm256_add_epi32(vsumb, voffset);
+ vsuma = _mm256_srai_epi32(vsuma, shift);
+ vsumb = _mm256_srai_epi32(vsumb, shift);
+ vsum = _mm256_packs_epi32(vsuma, vsumb);
+ if (shiftBack)
+ { //clip
+ vsum = _mm256_min_epi16(vibdimax, _mm256_max_epi16(vibdimin, vsum));
+ }
+ _mm256_storeu_si256((__m256i *)&dst[col], vsum);
+ src += srcStride;
+ dst += dstStride;
+ }
+ src = srcOrig;
+ dst = dstOrig;
+ }
+#endif
+#if USE_AVX2
+ _mm256_zeroupper();
+#endif
+}
+#endif
template<X86_VEXT vext, int N, bool shiftBack>
static void simdInterpolateVerM16_AVX2(const int16_t *src, int srcStride, int16_t *dst, int dstStride, int width, int height, int shift, int offset, const ClpRng& clpRng, int16_t const *coeff)
{
@@ -2226,6 +2942,9 @@ template<X86_VEXT vext, int N, bool isVertical, bool isFirst, bool isLast>
static void simdFilter( const ClpRng& clpRng, Pel const *src, int srcStride, Pel *dst, int dstStride, int width, int height, TFilterCoeff const *coeff, bool biMCForDMVR)
{
int row, col;
+#if JVET_Z0117_CHROMA_IF
+ bool issym = false;
+#endif
#if IF_12TAP_SIMD
Pel c[16] = { 0, };
#else
@@ -2336,7 +3055,11 @@ static void simdFilter( const ClpRng& clpRng, Pel const *src, int srcStride, Pel
return;
}
#endif
+#if JVET_Z0117_CHROMA_IF
+ else if ((N == 8 || (N == 6)) && !(width & 0x07))
+#else
else if (N == 8 && !(width & 0x07))
+#endif
#else
if (N == 8 && !(width & 0x07))
#endif
@@ -2346,77 +3069,273 @@ static void simdFilter( const ClpRng& clpRng, Pel const *src, int srcStride, Pel
if( vext>= AVX2 )
#if MCIF_SIMD_NEW
if (!(width & 15))
+#if JVET_Z0117_CHROMA_IF
+ {
+ if(issym)
+ {
+ simdInterpolateHorM16_sym_AVX2<vext, 6, isLast>(src, srcStride, dst, dstStride, width, height, shift, offset, clpRng, c);
+ }
+ else if(N == 6)
+ {
+ simdInterpolateHorM16_AVX2<vext, 6, isLast>(src, srcStride, dst, dstStride, width, height, shift, offset, clpRng, c);
+ }
+ else
+ {
+ simdInterpolateHorM16_AVX2<vext, 8, isLast>(src, srcStride, dst, dstStride, width, height, shift, offset, clpRng, c);
+ }
+ }
+#else
simdInterpolateHorM16_AVX2<vext, 8, isLast>(src, srcStride, dst, dstStride, width, height, shift, offset, clpRng, c);
+#endif
else
#endif
+#if JVET_Z0117_CHROMA_IF
+ {
+ if(issym)
+ {
+ simdInterpolateHorM8_sym_AVX2<vext, 6, isLast>( src, srcStride, dst, dstStride, width, height, shift, offset, clpRng, c );
+ }
+ else if(N == 6)
+ {
+ simdInterpolateHorM8_AVX2<vext, 6, isLast>( src, srcStride, dst, dstStride, width, height, shift, offset, clpRng, c );
+ }
+ else
+ {
+ simdInterpolateHorM8_AVX2<vext, 8, isLast>( src, srcStride, dst, dstStride, width, height, shift, offset, clpRng, c );
+ }
+ }
+#else
simdInterpolateHorM8_AVX2<vext, 8, isLast>( src, srcStride, dst, dstStride, width, height, shift, offset, clpRng, c );
+#endif
else
+#if JVET_Z0117_CHROMA_IF
+ if(N == 6)
+ {
+ simdInterpolateHorM8<vext, 6, isLast>( src, srcStride, dst, dstStride, width, height, shift, offset, clpRng, c );
+ }
+ else
+ {
+ simdInterpolateHorM8<vext, 8, isLast>( src, srcStride, dst, dstStride, width, height, shift, offset, clpRng, c );
+ }
+#else
simdInterpolateHorM8<vext, 8, isLast>( src, srcStride, dst, dstStride, width, height, shift, offset, clpRng, c );
+#endif
}
else
{
if( vext>= AVX2 )
#if MCIF_SIMD_NEW
if (!(width & 15))
+#if JVET_Z0117_CHROMA_IF
+ {
+ if(issym)
+ {
+ simdInterpolateVerM16_AVX2<vext, 6, isLast>(src, srcStride, dst, dstStride, width, height, shift, offset, clpRng, c);
+ }
+ else if(N == 6)
+ {
+ simdInterpolateVerM16_AVX2<vext, 6, isLast>(src, srcStride, dst, dstStride, width, height, shift, offset, clpRng, c);
+ }
+ else
+ {
+ simdInterpolateVerM16_AVX2<vext, 8, isLast>(src, srcStride, dst, dstStride, width, height, shift, offset, clpRng, c);
+ }
+ }
+#else
simdInterpolateVerM16_AVX2<vext, 8, isLast>(src, srcStride, dst, dstStride, width, height, shift, offset, clpRng, c);
+#endif
else
#endif
+#if JVET_Z0117_CHROMA_IF
+ {
+ if(issym)
+ {
+ simdInterpolateVerM8_AVX2<vext, 6, isLast>( src, srcStride, dst, dstStride, width, height, shift, offset, clpRng, c );
+ }
+ else if(N == 6)
+ {
+ simdInterpolateVerM8_AVX2<vext, 6, isLast>( src, srcStride, dst, dstStride, width, height, shift, offset, clpRng, c );
+ }
+ else
+ {
+ simdInterpolateVerM8_AVX2<vext, 8, isLast>( src, srcStride, dst, dstStride, width, height, shift, offset, clpRng, c );
+ }
+ }
+#else
simdInterpolateVerM8_AVX2<vext, 8, isLast>( src, srcStride, dst, dstStride, width, height, shift, offset, clpRng, c );
+#endif
else
+#if JVET_Z0117_CHROMA_IF
+ {
+ if(N == 6)
+ {
+ simdInterpolateVerM8<vext, 6, isLast>( src, srcStride, dst, dstStride, width, height, shift, offset, clpRng, c );
+ }
+ else
+ {
+ simdInterpolateVerM8<vext, 8, isLast>( src, srcStride, dst, dstStride, width, height, shift, offset, clpRng, c );
+ }
+ }
+#else
simdInterpolateVerM8<vext, 8, isLast>( src, srcStride, dst, dstStride, width, height, shift, offset, clpRng, c );
+#endif
}
return;
}
+#if JVET_Z0117_CHROMA_IF
+ else if( (N == 8 || N == 6) && !( width & 0x03 ) )
+#else
else if( N == 8 && !( width & 0x03 ) )
+#endif
{
if( !isVertical )
{
+#if JVET_Z0117_CHROMA_IF
+ {
+ if(issym)
+ {
+ simdInterpolateHorM4_sym<vext, 6, isLast>( src, srcStride, dst, dstStride, width, height, shift, offset, clpRng, c );
+ }
+ else if(N == 6)
+ {
+ simdInterpolateHorM4<vext, 6, isLast>( src, srcStride, dst, dstStride, width, height, shift, offset, clpRng, c );
+ }
+ else
+ {
+ simdInterpolateHorM4<vext, 8, isLast>( src, srcStride, dst, dstStride, width, height, shift, offset, clpRng, c );
+ }
+ }
+#else
simdInterpolateHorM4<vext, 8, isLast>( src, srcStride, dst, dstStride, width, height, shift, offset, clpRng, c );
+#endif
}
else
+#if JVET_Z0117_CHROMA_IF
+ {
+ if(issym)
+ {
+ simdInterpolateVerM4<vext, 6, isLast>( src, srcStride, dst, dstStride, width, height, shift, offset, clpRng, c );
+ }
+ else if(N == 6)
+ {
+ simdInterpolateVerM4<vext, 6, isLast>( src, srcStride, dst, dstStride, width, height, shift, offset, clpRng, c );
+ }
+ else
+ {
+ simdInterpolateVerM4<vext, 8, isLast>( src, srcStride, dst, dstStride, width, height, shift, offset, clpRng, c );
+ }
+ }
+#else
simdInterpolateVerM4<vext, 8, isLast>( src, srcStride, dst, dstStride, width, height, shift, offset, clpRng, c );
+#endif
return;
}
#if IF_FILTER_WIDTH_NON_M4
+#if JVET_Z0117_CHROMA_IF
+ else if ((N == 8 || (N == 6)) && width > 16)
+#else
else if (N == 8 && width > 16)
+#endif
{
const int16_t* initialSrc = src;
int16_t* initialDst = dst;
if (!isVertical)
{
+#if JVET_Z0117_CHROMA_IF
+ if(issym)
+ {
+ simdInterpolateHorM16_sym_AVX2<vext, 6, isLast>(src, srcStride, dst, dstStride, width - 1, height, shift, offset, clpRng, c);
+ }
+ else if(N == 6)
+ {
+ simdInterpolateHorM16_AVX2<vext, 6, isLast>(src, srcStride, dst, dstStride, width - 1, height, shift, offset, clpRng, c);
+ }
+ else
+ {
+ simdInterpolateHorM16_AVX2<vext, 8, isLast>(src, srcStride, dst, dstStride, width - 1, height, shift, offset, clpRng, c);
+ }
+#else
simdInterpolateHorM16_AVX2<vext, 8, isLast>(src, srcStride, dst, dstStride, width - 1, height, shift, offset, clpRng, c);
+#endif
}
else
{
+#if JVET_Z0117_CHROMA_IF
+ if(issym)
+ {
+ simdInterpolateVerM16_AVX2<vext, 6, isLast>(src, srcStride, dst, dstStride, width - 1, height, shift, offset, clpRng, c);
+ }
+ else if(N == 6)
+ {
+ simdInterpolateVerM16_AVX2<vext, 6, isLast>(src, srcStride, dst, dstStride, width - 1, height, shift, offset, clpRng, c);
+ }
+ else
+ {
+ simdInterpolateVerM16_AVX2<vext, 8, isLast>(src, srcStride, dst, dstStride, width - 1, height, shift, offset, clpRng, c);
+ }
+#else
simdInterpolateVerM16_AVX2<vext, 8, isLast>(src, srcStride, dst, dstStride, width - 1, height, shift, offset, clpRng, c);
+#endif
}
int sum = 0;
int stride = isVertical ? srcStride : 1;
- for (int row = 0; row < height; row++)
+#if JVET_Z0117_CHROMA_IF
+ if(N >= 8)
{
- for (int col = width - 1; col < width; col++)
+#endif
+ for (int row = 0; row < height; row++)
{
-
- sum = initialSrc[col] * coeff[0];
- sum += initialSrc[col + 1 * stride] * coeff[1];
- sum += initialSrc[col + 2 * stride] * coeff[2];
- sum += initialSrc[col + 3 * stride] * coeff[3];
- sum += initialSrc[col + 4 * stride] * coeff[4];
- sum += initialSrc[col + 5 * stride] * coeff[5];
- sum += initialSrc[col + 6 * stride] * coeff[6];
- sum += initialSrc[col + 7 * stride] * coeff[7];
-
- Pel val = (sum + offset) >> shift;
- if (isLast)
+ for (int col = width - 1; col < width; col++)
{
- val = ClipPel(val, clpRng);
+
+ sum = initialSrc[col] * coeff[0];
+ sum += initialSrc[col + 1 * stride] * coeff[1];
+ sum += initialSrc[col + 2 * stride] * coeff[2];
+ sum += initialSrc[col + 3 * stride] * coeff[3];
+ sum += initialSrc[col + 4 * stride] * coeff[4];
+ sum += initialSrc[col + 5 * stride] * coeff[5];
+ sum += initialSrc[col + 6 * stride] * coeff[6];
+ sum += initialSrc[col + 7 * stride] * coeff[7];
+
+ Pel val = (sum + offset) >> shift;
+ if (isLast)
+ {
+ val = ClipPel(val, clpRng);
+ }
+ initialDst[col] = val;
}
- initialDst[col] = val;
+
+ initialSrc += srcStride;
+ initialDst += dstStride;
}
+#if JVET_Z0117_CHROMA_IF
+ }
+ else if(N == 6){
+ col = width - 1
+ for (int row = 0; row < height; row++)
+ {
+ {
+ sum = initialSrc[col] * coeff[0];
+ sum += initialSrc[col + 1 * stride] * coeff[1];
+ sum += initialSrc[col + 2 * stride] * coeff[2];
+ sum += initialSrc[col + 3 * stride] * coeff[3];
+ sum += initialSrc[col + 4 * stride] * coeff[4];
+ sum += initialSrc[col + 5 * stride] * coeff[5];
+
+ Pel val = (sum + offset) >> shift;
+ if (isLast)
+ {
+ val = ClipPel(val, clpRng);
+ }
+ initialDst[col] = val;
+ }
- initialSrc += srcStride;
- initialDst += dstStride;
+ initialSrc += srcStride;
+ initialDst += dstStride;
+ }
+ }
}
+#endif
}
#endif
else if( N == 4 && !( width & 0x03 ) )
@@ -3117,7 +4036,22 @@ void InterpolationFilter::_initInterpolationFilterX86()
m_filterHor[1][0][1] = simdFilter<vext, 8, false, false, true>;
m_filterHor[1][1][0] = simdFilter<vext, 8, false, true, false>;
m_filterHor[1][1][1] = simdFilter<vext, 8, false, true, true>;
-
+#if JVET_Z0117_CHROMA_IF
+ m_filterHor[2][0][0] = simdFilter<vext, 6, false, false, false>;
+ m_filterHor[2][0][1] = simdFilter<vext, 6, false, false, true>;
+ m_filterHor[2][1][0] = simdFilter<vext, 6, false, true, false>;
+ m_filterHor[2][1][1] = simdFilter<vext, 6, false, true, true>;
+
+ m_filterHor[3][0][0] = simdFilter<vext, 4, false, false, false>;
+ m_filterHor[3][0][1] = simdFilter<vext, 4, false, false, true>;
+ m_filterHor[3][1][0] = simdFilter<vext, 4, false, true, false>;
+ m_filterHor[3][1][1] = simdFilter<vext, 4, false, true, true>;
+
+ m_filterHor[4][0][0] = simdFilter<vext, 2, false, false, false>;
+ m_filterHor[4][0][1] = simdFilter<vext, 2, false, false, true>;
+ m_filterHor[4][1][0] = simdFilter<vext, 2, false, true, false>;
+ m_filterHor[4][1][1] = simdFilter<vext, 2, false, true, true>;
+#else
m_filterHor[2][0][0] = simdFilter<vext, 4, false, false, false>;
m_filterHor[2][0][1] = simdFilter<vext, 4, false, false, true>;
m_filterHor[2][1][0] = simdFilter<vext, 4, false, true, false>;
@@ -3127,7 +4061,7 @@ void InterpolationFilter::_initInterpolationFilterX86()
m_filterHor[3][0][1] = simdFilter<vext, 2, false, false, true>;
m_filterHor[3][1][0] = simdFilter<vext, 2, false, true, false>;
m_filterHor[3][1][1] = simdFilter<vext, 2, false, true, true>;
-
+#endif
m_filterVer[0][0][0] = simdFilter<vext, 12, true, false, false>;
m_filterVer[0][0][1] = simdFilter<vext, 12, true, false, true>;
m_filterVer[0][1][0] = simdFilter<vext, 12, true, true, false>;
@@ -3137,7 +4071,22 @@ void InterpolationFilter::_initInterpolationFilterX86()
m_filterVer[1][0][1] = simdFilter<vext, 8, true, false, true>;
m_filterVer[1][1][0] = simdFilter<vext, 8, true, true, false>;
m_filterVer[1][1][1] = simdFilter<vext, 8, true, true, true>;
-
+#if JVET_Z0117_CHROMA_IF
+ m_filterVer[2][0][0] = simdFilter<vext, 6, true, false, false>;
+ m_filterVer[2][0][1] = simdFilter<vext, 6, true, false, true>;
+ m_filterVer[2][1][0] = simdFilter<vext, 6, true, true, false>;
+ m_filterVer[2][1][1] = simdFilter<vext, 6, true, true, true>;
+
+ m_filterVer[3][0][0] = simdFilter<vext, 4, true, false, false>;
+ m_filterVer[3][0][1] = simdFilter<vext, 4, true, false, true>;
+ m_filterVer[3][1][0] = simdFilter<vext, 4, true, true, false>;
+ m_filterVer[3][1][1] = simdFilter<vext, 4, true, true, true>;
+
+ m_filterVer[4][0][0] = simdFilter<vext, 2, true, false, false>;
+ m_filterVer[4][0][1] = simdFilter<vext, 2, true, false, true>;
+ m_filterVer[4][1][0] = simdFilter<vext, 2, true, true, false>;
+ m_filterVer[4][1][1] = simdFilter<vext, 2, true, true, true>;
+#else
m_filterVer[2][0][0] = simdFilter<vext, 4, true, false, false>;
m_filterVer[2][0][1] = simdFilter<vext, 4, true, false, true>;
m_filterVer[2][1][0] = simdFilter<vext, 4, true, true, false>;
@@ -3147,7 +4096,7 @@ void InterpolationFilter::_initInterpolationFilterX86()
m_filterVer[3][0][1] = simdFilter<vext, 2, true, false, true>;
m_filterVer[3][1][0] = simdFilter<vext, 2, true, true, false>;
m_filterVer[3][1][1] = simdFilter<vext, 2, true, true, true>;
-
+#endif
m_filterCopy[0][0] = simdFilterCopy<vext, false, false>;
m_filterCopy[0][1] = simdFilterCopy<vext, false, true>;
m_filterCopy[1][0] = simdFilterCopy<vext, true, false>;
@@ -3158,7 +4107,22 @@ void InterpolationFilter::_initInterpolationFilterX86()
m_filterHor[0][0][1] = simdFilter<vext, 8, false, false, true>;
m_filterHor[0][1][0] = simdFilter<vext, 8, false, true, false>;
m_filterHor[0][1][1] = simdFilter<vext, 8, false, true, true>;
+#if JVET_Z0117_CHROMA_IF
+ m_filterHor[1][0][0] = simdFilter<vext, 6, false, false, false>;
+ m_filterHor[1][0][1] = simdFilter<vext, 6, false, false, true>;
+ m_filterHor[1][1][0] = simdFilter<vext, 6, false, true, false>;
+ m_filterHor[1][1][1] = simdFilter<vext, 6, false, true, true>;
+
+ m_filterHor[2][0][0] = simdFilter<vext, 4, false, false, false>;
+ m_filterHor[2][0][1] = simdFilter<vext, 4, false, false, true>;
+ m_filterHor[2][1][0] = simdFilter<vext, 4, false, true, false>;
+ m_filterHor[2][1][1] = simdFilter<vext, 4, false, true, true>;
+ m_filterHor[3][0][0] = simdFilter<vext, 2, false, false, false>;
+ m_filterHor[3][0][1] = simdFilter<vext, 2, false, false, true>;
+ m_filterHor[3][1][0] = simdFilter<vext, 2, false, true, false>;
+ m_filterHor[3][1][1] = simdFilter<vext, 2, false, true, true>;
+#else
m_filterHor[1][0][0] = simdFilter<vext, 4, false, false, false>;
m_filterHor[1][0][1] = simdFilter<vext, 4, false, false, true>;
m_filterHor[1][1][0] = simdFilter<vext, 4, false, true, false>;
@@ -3168,12 +4132,28 @@ void InterpolationFilter::_initInterpolationFilterX86()
m_filterHor[2][0][1] = simdFilter<vext, 2, false, false, true>;
m_filterHor[2][1][0] = simdFilter<vext, 2, false, true, false>;
m_filterHor[2][1][1] = simdFilter<vext, 2, false, true, true>;
-
+#endif
m_filterVer[0][0][0] = simdFilter<vext, 8, true, false, false>;
m_filterVer[0][0][1] = simdFilter<vext, 8, true, false, true>;
m_filterVer[0][1][0] = simdFilter<vext, 8, true, true, false>;
m_filterVer[0][1][1] = simdFilter<vext, 8, true, true, true>;
+ m_filterVer[0][1][1] = simdFilter<vext, 8, true, true, true>;
+#if JVET_Z0117_CHROMA_IF
+ m_filterVer[1][0][0] = simdFilter<vext, 6, true, false, false>;
+ m_filterVer[1][0][1] = simdFilter<vext, 6, true, false, true>;
+ m_filterVer[1][1][0] = simdFilter<vext, 6, true, true, false>;
+ m_filterVer[1][1][1] = simdFilter<vext, 6, true, true, true>;
+
+ m_filterVer[2][0][0] = simdFilter<vext, 4, true, false, false>;
+ m_filterVer[2][0][1] = simdFilter<vext, 4, true, false, true>;
+ m_filterVer[2][1][0] = simdFilter<vext, 4, true, true, false>;
+ m_filterVer[2][1][1] = simdFilter<vext, 4, true, true, true>;
+ m_filterVer[3][0][0] = simdFilter<vext, 2, true, false, false>;
+ m_filterVer[3][0][1] = simdFilter<vext, 2, true, false, true>;
+ m_filterVer[3][1][0] = simdFilter<vext, 2, true, true, false>;
+ m_filterVer[3][1][1] = simdFilter<vext, 2, true, true, true>;
+#else
m_filterVer[1][0][0] = simdFilter<vext, 4, true, false, false>;
m_filterVer[1][0][1] = simdFilter<vext, 4, true, false, true>;
m_filterVer[1][1][0] = simdFilter<vext, 4, true, true, false>;
@@ -3183,7 +4163,7 @@ void InterpolationFilter::_initInterpolationFilterX86()
m_filterVer[2][0][1] = simdFilter<vext, 2, true, false, true>;
m_filterVer[2][1][0] = simdFilter<vext, 2, true, true, false>;
m_filterVer[2][1][1] = simdFilter<vext, 2, true, true, true>;
-
+#endif
m_filterCopy[0][0] = simdFilterCopy<vext, false, false>;
m_filterCopy[0][1] = simdFilterCopy<vext, false, true>;
m_filterCopy[1][0] = simdFilterCopy<vext, true, false>;