diff --git a/source/Lib/CommonLib/Contexts.cpp b/source/Lib/CommonLib/Contexts.cpp
index 1bd41905edc89da556daaa387d9d01f3b84b1857..edf20f570ce22ede7d9d7afe8e6678a3f709711e 100644
--- a/source/Lib/CommonLib/Contexts.cpp
+++ b/source/Lib/CommonLib/Contexts.cpp
@@ -1571,7 +1571,17 @@ const CtxSet ContextSetCfg::TransformSkipFlag = ContextSetCfg::addCtxSet
{ 1, 5 },
{ 2, 5 }
});
-
+#if JVET_W0103_INTRA_MTS
+const CtxSet ContextSetCfg::MTSIdx = ContextSetCfg::addCtxSet
+({
+ { 45, 35, 20, 45, },
+ { 38, 35, 35, 38, },
+ { 37, 28, 28, 37, },
+ { 8, 10, 10, 8, },
+ { 8, 10, 10, 8, },
+ { 9, 10, 10, 8, }
+ });
+#else
const CtxSet ContextSetCfg::MTSIdx = ContextSetCfg::addCtxSet
({
{ 37, 25, 34, 40 },
@@ -1581,7 +1591,7 @@ const CtxSet ContextSetCfg::MTSIdx = ContextSetCfg::addCtxSet
{ 9, 0, 10, 8 },
{ 9, 1, 9, 0 }
});
-
+#endif
const CtxSet ContextSetCfg::ISPMode = ContextSetCfg::addCtxSet
({
#if JVET_W0123_TIMD_FUSION
@@ -2579,7 +2589,15 @@ const CtxSet ContextSetCfg::TransformSkipFlag = ContextSetCfg::addCtxSet
{ 25, 9, },
{ 1, 1, },
});
-
+#if JVET_W0103_INTRA_MTS
+const CtxSet ContextSetCfg::MTSIdx = ContextSetCfg::addCtxSet
+({
+ { 45, 35, 20, 45, },
+ { 38, 35, 35, 38, },
+ { 37, 28, 28, 37, },
+ { 8, 9, 9, 8, },
+ });
+#else
const CtxSet ContextSetCfg::MTSIdx = ContextSetCfg::addCtxSet
({
{ 45, 25, 27, 0, },
@@ -2587,7 +2605,7 @@ const CtxSet ContextSetCfg::MTSIdx = ContextSetCfg::addCtxSet
{ 29, 0, 28, 0, },
{ 8, 0, 9, 0, },
});
-
+#endif
const CtxSet ContextSetCfg::ISPMode = ContextSetCfg::addCtxSet
({
#if JVET_W0123_TIMD_FUSION
diff --git a/source/Lib/CommonLib/Rom.cpp b/source/Lib/CommonLib/Rom.cpp
index fb4317e06b126a8385b25016ce3e79ed56104f23..1bf0b8ef990831d35f8f10115c7b540ef8313c31 100644
--- a/source/Lib/CommonLib/Rom.cpp
+++ b/source/Lib/CommonLib/Rom.cpp
@@ -3843,7 +3843,213 @@ TMatrixCoeff g_trCoreDCT2P256[256][256];
TMatrixCoeff g_trCoreDCT8P256[256][256];
TMatrixCoeff g_trCoreDST7P256[256][256];
#endif
+#if JVET_W0103_INTRA_MTS
+TMatrixCoeff g_aiTr2[NUM_TRANS_TYPE][2][2];
+TMatrixCoeff g_aiTr4[NUM_TRANS_TYPE][4][4];
+TMatrixCoeff g_aiTr8[NUM_TRANS_TYPE][8][8];
+TMatrixCoeff g_aiTr16[NUM_TRANS_TYPE][16][16];
+TMatrixCoeff g_aiTr32[NUM_TRANS_TYPE][32][32];
+TMatrixCoeff g_aiTr64[NUM_TRANS_TYPE][64][64];
+TMatrixCoeff g_aiTr128[NUM_TRANS_TYPE][128][128];
+TMatrixCoeff g_aiTr256[NUM_TRANS_TYPE][256][256];
+const uint8_t g_aucIpmToTrSet[16][36] =
+{
+ //0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 MIP
+ { 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 4 }, //4x4
+ { 5, 5, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 9 }, //4x8
+ {10,10,11,11,11,11,11,11,11,11,11,11,11,12,12,12,12,12,12,12,12,12,12,12,13,13,13,13,13,13,13,13,13,13,13,14 }, //4x16
+ {15,15,16,16,16,16,16,16,16,16,16,16,16,17,17,17,17,17,17,17,17,17,17,17,18,18,18,18,18,18,18,18,18,18,18,19 }, //4x32
+ {20,20,21,21,21,21,21,21,21,21,21,21,21,22,22,22,22,22,22,22,22,22,22,22,23,23,23,23,23,23,23,23,23,23,23,24 }, //8x4
+ {25,25,26,26,26,26,26,26,26,26,26,26,26,27,27,27,27,27,27,27,27,27,27,27,28,28,28,28,28,28,28,28,28,28,28,29 }, //8x8
+ {30,30,31,31,31,31,31,31,31,31,31,31,31,32,32,32,32,32,32,32,32,32,32,32,33,33,33,33,33,33,33,33,33,33,33,34 }, //8x16
+ {35,35,36,36,36,36,36,36,36,36,36,36,36,37,37,37,37,37,37,37,37,37,37,37,38,38,38,38,38,38,38,38,38,38,38,39 }, //8x32
+ {40,40,41,41,41,41,41,41,41,41,41,41,41,42,42,42,42,42,42,42,42,42,42,42,43,43,43,43,43,43,43,43,43,43,43,44 }, //16x4
+ {45,45,46,46,46,46,46,46,46,46,46,46,46,47,47,47,47,47,47,47,47,47,47,47,48,48,48,48,48,48,48,48,48,48,48,49 }, //16x8
+ {50,50,51,51,51,51,51,51,51,51,51,51,51,52,52,52,52,52,52,52,52,52,52,52,53,53,53,53,53,53,53,53,53,53,53,54 }, //16x16
+ {55,55,56,56,56,56,56,56,56,56,56,56,56,57,57,57,57,57,57,57,57,57,57,57,58,58,58,58,58,58,58,58,58,58,58,59 }, //16x32
+ {60,60,61,61,61,61,61,61,61,61,61,61,61,62,62,62,62,62,62,62,62,62,62,62,63,63,63,63,63,63,63,63,63,63,63,64 }, //32x4
+ {65,65,66,66,66,66,66,66,66,66,66,66,66,67,67,67,67,67,67,67,67,67,67,67,68,68,68,68,68,68,68,68,68,68,68,69 }, //32x8
+ {70,70,71,71,71,71,71,71,71,71,71,71,71,72,72,72,72,72,72,72,72,72,72,72,73,73,73,73,73,73,73,73,73,73,73,74 }, //32x16
+ {75,75,76,76,76,76,76,76,76,76,76,76,76,77,77,77,77,77,77,77,77,77,77,77,78,78,78,78,78,78,78,78,78,78,78,79 }, //32x32
+};
+const int8_t g_aiIdLut[3][3] =
+{
+ { 8, 6, 4 },{ 8, 8, 6 },{ 4, 2, -1 }
+};
+const uint8_t g_aucTrIdxToTr[25][2] =
+{
+ { DCT8, DCT8 },{ DCT8, DST7 },{ DCT8, DCT5 },{ DCT8, DST4 }, {DCT8, DST1},
+ { DST7, DCT8 },{ DST7, DST7 },{ DST7, DCT5 },{ DST7, DST4 }, {DST7, DST1},
+ { DCT5, DCT8 },{ DCT5, DST7 },{ DCT5, DCT5 },{ DCT5, DST4 }, {DCT5, DST1},
+ { DST4, DCT8 },{ DST4, DST7 },{ DST4, DCT5 },{ DST4, DST4 }, {DST4, DST1},
+ { DST1, DCT8 },{ DST1, DST7 },{ DST1, DCT5 },{ DST1, DST4 }, {DST1, DST1},
+};
+
+const uint8_t g_aucTrSet[80][4] =
+{
+//T0:0, 1, 2, 3,
+{ 17, 18, 23, 24},
+//T1:0, 1, 2, 3,
+{ 3, 7, 18, 22},
+//T2:0, 1, 2, 3,
+{ 2, 17, 18, 22},
+//T3:0, 1, 2, 3,
+{ 3, 15, 17, 18},
+//T4:0, 1, 2, 3,
+{ 3, 12, 18, 19},
+//T5:0, 1, 2, 3,
+{ 12, 18, 19, 23},
+//T6:0, 1, 2, 3,
+{ 2, 12, 17, 18},
+//T7:0, 1, 2, 3,
+{ 2, 17, 18, 22},
+//T8:0, 1, 2, 3,
+{ 2, 11, 17, 18},
+//T9:0, 1, 2, 3,
+{ 12, 18, 19, 23},
+//T10:0, 1, 2, 3,
+{ 12, 13, 16, 24},
+//T11:0, 1, 2, 3,
+{ 2, 11, 16, 23},
+//T12:0, 1, 2, 3,
+{ 2, 13, 17, 22},
+//T13:0, 1, 2, 3,
+{ 2, 11, 17, 21},
+//T14:0, 1, 2, 3,
+{ 13, 16, 19, 22},
+//T15:0, 1, 2, 3,
+{ 7, 12, 13, 18},
+//T16:0, 1, 2, 3,
+{ 1, 11, 12, 16},
+//T17:0, 1, 2, 3,
+{ 3, 13, 17, 22},
+//T18:0, 1, 2, 3,
+{ 1, 6, 12, 22},
+//T19:0, 1, 2, 3,
+{ 12, 13, 15, 16},
+//T20:0, 1, 2, 3,
+{ 18, 19, 23, 24},
+//T21:0, 1, 2, 3,
+{ 2, 17, 18, 24},
+//T22:0, 1, 2, 3,
+{ 3, 4, 17, 22},
+//T23:0, 1, 2, 3,
+{ 12, 18, 19, 23},
+//T24:0, 1, 2, 3,
+{ 12, 18, 19, 23},
+//T25:0, 1, 2, 3,
+{ 6, 12, 18, 24},
+//T26:0, 1, 2, 3,
+{ 2, 6, 12, 21},
+//T27:0, 1, 2, 3,
+{ 1, 11, 17, 22},
+//T28:0, 1, 2, 3,
+{ 3, 11, 16, 17},
+//T29:0, 1, 2, 3,
+{ 8, 12, 19, 23},
+//T30:0, 1, 2, 3,
+{ 7, 13, 16, 23},
+//T31:0, 1, 2, 3,
+{ 1, 6, 11, 12},
+//T32:0, 1, 2, 3,
+{ 1, 11, 17, 21},
+//T33:0, 1, 2, 3,
+{ 6, 11, 17, 21},
+//T34:0, 1, 2, 3,
+{ 8, 11, 14, 17},
+//T35:0, 1, 2, 3,
+{ 6, 11, 12, 21},
+//T36:0, 1, 2, 3,
+{ 1, 6, 11, 12},
+//T37:0, 1, 2, 3,
+{ 2, 6, 11, 12},
+//T38:0, 1, 2, 3,
+{ 1, 6, 11, 21},
+//T39:0, 1, 2, 3,
+{ 7, 11, 12, 16},
+//T40:0, 1, 2, 3,
+{ 8, 12, 19, 24},
+//T41:0, 1, 2, 3,
+{ 1, 13, 18, 22},
+//T42:0, 1, 2, 3,
+{ 2, 6, 17, 21},
+//T43:0, 1, 2, 3,
+{ 11, 12, 16, 19},
+//T44:0, 1, 2, 3,
+{ 8, 12, 17, 24},
+//T45:0, 1, 2, 3,
+{ 6, 12, 19, 21},
+//T46:0, 1, 2, 3,
+{ 6, 12, 13, 21},
+//T47:0, 1, 2, 3,
+{ 2, 16, 17, 21},
+//T48:0, 1, 2, 3,
+{ 6, 17, 19, 23},
+//T49:0, 1, 2, 3,
+{ 6, 12, 14, 17},
+//T50:0, 1, 2, 3,
+{ 6, 7, 11, 21},
+//T51:0, 1, 2, 3,
+{ 1, 11, 12, 16},
+//T52:0, 1, 2, 3,
+{ 1, 6, 11, 12},
+//T53:0, 1, 2, 3,
+{ 6, 11, 12, 21},
+//T54:0, 1, 2, 3,
+{ 7, 8, 9, 11},
+//T55:0, 1, 2, 3,
+{ 6, 7, 11, 12},
+//T56:0, 1, 2, 3,
+{ 6, 7, 11, 12},
+//T57:0, 1, 2, 3,
+{ 1, 11, 12, 16},
+//T58:0, 1, 2, 3,
+{ 6, 11, 17, 21},
+//T59:0, 1, 2, 3,
+{ 6, 7, 11, 12},
+//T60:0, 1, 2, 3,
+{ 12, 14, 18, 21},
+//T61:0, 1, 2, 3,
+{ 1, 11, 16, 22},
+//T62:0, 1, 2, 3,
+{ 1, 11, 16, 22},
+//T63:0, 1, 2, 3,
+{ 7, 13, 15, 16},
+//T64:0, 1, 2, 3,
+{ 1, 8, 12, 19},
+//T65:0, 1, 2, 3,
+{ 6, 7, 9, 12},
+//T66:0, 1, 2, 3,
+{ 2, 6, 12, 13},
+//T67:0, 1, 2, 3,
+{ 1, 12, 16, 21},
+//T68:0, 1, 2, 3,
+{ 7, 11, 16, 19},
+//T69:0, 1, 2, 3,
+{ 7, 8, 11, 12},
+//T70:0, 1, 2, 3,
+{ 6, 7, 11, 12},
+//T71:0, 1, 2, 3,
+{ 6, 7, 11, 12},
+//T72:0, 1, 2, 3,
+{ 1, 6, 11, 12},
+//T73:0, 1, 2, 3,
+{ 6, 7, 11, 16},
+//T74:0, 1, 2, 3,
+{ 6, 7, 11, 12},
+//T75:0, 1, 2, 3,
+{ 6, 7, 11, 12},
+//T76:0, 1, 2, 3,
+{ 6, 11, 12, 21},
+//T77:0, 1, 2, 3,
+{ 1, 6, 11, 12},
+//T78:0, 1, 2, 3,
+{ 6, 7, 11, 12},
+//T79:0, 1, 2, 3,
+{ 6, 7, 11, 12},
+};
+#endif
// initialize ROM variables
void initROM()
{
@@ -4041,7 +4247,7 @@ void initROM()
g_paletteQuant[qp] = (int)(threshQP*0.16 + 0.5);
}
-#if SIGN_PREDICTION
+#if SIGN_PREDICTION && !JVET_W0103_INTRA_MTS
memset(&g_resiBorderTemplate[0][0][0], 0, sizeof(g_resiBorderTemplate));
const int8_t *p_data = g_initRomSignPred;
for( int log2Width = 0; log2Width < 6; ++log2Width)
@@ -4094,6 +4300,72 @@ void initROM()
}
}
#endif
+#if JVET_W0103_INTRA_MTS
+#if LMS_LINEAR_MODEL || TRANSFORM_SIMD_OPT || TU_256
+ c = 2;
+#else
+ int c = 2;
+ const double PI = 3.14159265358979323846;
+#endif
+
+ for (int i = 0; i < 8; i++)
+ {
+ const double s = sqrt((double)c) * (64 << COM16_C806_TRANS_PREC);
+ TMatrixCoeff *iT = NULL;
+
+ switch (i)
+ {
+ case 0: iT = g_aiTr2[0][0]; break;
+ case 1: iT = g_aiTr4[0][0]; break;
+ case 2: iT = g_aiTr8[0][0]; break;
+ case 3: iT = g_aiTr16[0][0]; break;
+ case 4: iT = g_aiTr32[0][0]; break;
+ case 5: iT = g_aiTr64[0][0]; break;
+ case 6: iT = g_aiTr128[0][0]; break;
+ case 7: iT = g_aiTr256[0][0]; break;
+ case 8: exit(0); break;
+ }
+
+ for (int k = 0; k < c; k++)
+ {
+ for (int n = 0; n < c; n++)
+ {
+ double w0, w1, v;
+
+ // DCT-II
+ w0 = k == 0 ? sqrt(0.5) : 1;
+ v = cos(PI*(n + 0.5)*k / c) * w0 * sqrt(2.0 / c);
+ iT[DCT2*c*c + k * c + n] = (short)(s * v + (v > 0 ? 0.5 : -0.5));
+
+ // DCT-V
+ w0 = (k == 0) ? sqrt(0.5) : 1.0;
+ w1 = (n == 0) ? sqrt(0.5) : 1.0;
+ v = cos(PI*n*k / (c - 0.5)) * w0 * w1 * sqrt(2.0 / (c - 0.5));
+ iT[DCT5*c*c + k * c + n] = (short)(s * v + (v > 0 ? 0.5 : -0.5));
+
+ // DCT-VIII
+ v = cos(PI*(k + 0.5)*(n + 0.5) / (c + 0.5)) * sqrt(2.0 / (c + 0.5));
+ iT[DCT8*c*c + k * c + n] = (short)(s * v + (v > 0 ? 0.5 : -0.5));
+
+ // DST-I
+ v = sin(PI*(n + 1)*(k + 1) / (c + 1)) * sqrt(2.0 / (c + 1));
+ iT[DST1*c*c + k * c + n] = (short)(s * v + (v > 0 ? 0.5 : -0.5));
+
+ // DST-VII
+ v = sin(PI*(k + 0.5)*(n + 1) / (c + 0.5)) * sqrt(2.0 / (c + 0.5));
+ iT[DST7*c*c + k * c + n] = (short)(s * v + (v > 0 ? 0.5 : -0.5));
+
+ // DST4
+ v = sin(PI * (k + 0.5) * (n + 0.5) / c) * sqrt(2.0 / c);
+ iT[DST4 * c * c + k * c + n] = (short)(s * v + (v > 0 ? 0.5 : -0.5));
+
+ // ID
+ iT[IDTR * c * c + k * c + n] = (k == n) ? (short)(s + (s > 0 ? 0.5 : -0.5)) : 0;
+ }
+ }
+ c <<= 1;
+ }
+#endif
}
void destroyROM()
diff --git a/source/Lib/CommonLib/Rom.h b/source/Lib/CommonLib/Rom.h
index b41f205ba94e38bffa8a28c550716edd2b20a4dd..b6f4d1faa7693a4397cb75fe0aa7948941d45e33 100644
--- a/source/Lib/CommonLib/Rom.h
+++ b/source/Lib/CommonLib/Rom.h
@@ -166,7 +166,21 @@ extern TMatrixCoeff g_trCoreDCT2P256[256][256];
extern TMatrixCoeff g_trCoreDCT8P256[256][256];
extern TMatrixCoeff g_trCoreDST7P256[256][256];
#endif
-
+#if JVET_W0103_INTRA_MTS
+extern TMatrixCoeff g_aiTr2[NUM_TRANS_TYPE][2][2];
+extern TMatrixCoeff g_aiTr4[NUM_TRANS_TYPE][4][4];
+extern TMatrixCoeff g_aiTr8[NUM_TRANS_TYPE][8][8];
+extern TMatrixCoeff g_aiTr16[NUM_TRANS_TYPE][16][16];
+extern TMatrixCoeff g_aiTr32[NUM_TRANS_TYPE][32][32];
+extern TMatrixCoeff g_aiTr64[NUM_TRANS_TYPE][64][64];
+extern TMatrixCoeff g_aiTr128[NUM_TRANS_TYPE][128][128];
+extern TMatrixCoeff g_aiTr256[NUM_TRANS_TYPE][256][256];
+
+extern const uint8_t g_aucIpmToTrSet[16][36];
+extern const uint8_t g_aucTrSet[80][4];
+extern const int8_t g_aiIdLut[3][3];
+extern const uint8_t g_aucTrIdxToTr[25][2];
+#endif
#if EXTENDED_LFNST
extern const int8_t g_lfnst8x8[ 35 ][ 3 ][ 64 ][ 64 ];
extern const int8_t g_lfnst4x4[ 35 ][ 3 ][ 16 ][ 16 ];
diff --git a/source/Lib/CommonLib/TrQuant.cpp b/source/Lib/CommonLib/TrQuant.cpp
index 26fbda3252bb36edb6394600b284812be9f215d9..4b7586705aa37c25ec7c5ae35142ce09271e4140 100644
--- a/source/Lib/CommonLib/TrQuant.cpp
+++ b/source/Lib/CommonLib/TrQuant.cpp
@@ -297,6 +297,12 @@ void TrQuant::init( const Quant* otherQuant,
{ fastForwardDCT2_B2, fastForwardDCT2_B4, fastForwardDCT2_B8, fastForwardDCT2_B16, fastForwardDCT2_B32, fastForwardDCT2_B64, fastForwardDCT2_B128, fastForwardDCT2_B256 },
{ nullptr, fastForwardDCT8_B4, fastForwardDCT8_B8, fastForwardDCT8_B16, fastForwardDCT8_B32, fastForwardDCT8_B64, fastForwardDCT8_B128, fastForwardDCT8_B256 },
{ nullptr, fastForwardDST7_B4, fastForwardDST7_B8, fastForwardDST7_B16, fastForwardDST7_B32, fastForwardDST7_B64, fastForwardDST7_B128, fastForwardDST7_B256 },
+#if JVET_W0103_INTRA_MTS
+ { nullptr, fastForwardDCT5_B4, fastForwardDCT5_B8, fastForwardDCT5_B16, fastForwardDCT5_B32, fastForwardDCT5_B64, fastForwardDCT5_B128, fastForwardDCT5_B256 },
+ { nullptr, fastForwardDST4_B4, fastForwardDST4_B8, fastForwardDST4_B16, fastForwardDST4_B32, fastForwardDST4_B64, fastForwardDST4_B128, fastForwardDST4_B256 },
+ { nullptr, fastForwardDST1_B4, fastForwardDST1_B8, fastForwardDST1_B16, fastForwardDST1_B32, fastForwardDST1_B64, fastForwardDST1_B128, fastForwardDST1_B256 },
+ { nullptr, fastForwardIDTR_B4, fastForwardIDTR_B8, fastForwardIDTR_B16, fastForwardIDTR_B32, fastForwardIDTR_B64, fastForwardIDTR_B128, fastForwardIDTR_B256 },
+#endif
} };
fastInvTrans =
@@ -304,6 +310,12 @@ void TrQuant::init( const Quant* otherQuant,
{ fastInverseDCT2_B2, fastInverseDCT2_B4, fastInverseDCT2_B8, fastInverseDCT2_B16, fastInverseDCT2_B32, fastInverseDCT2_B64, fastInverseDCT2_B128, fastInverseDCT2_B256 },
{ nullptr, fastInverseDCT8_B4, fastInverseDCT8_B8, fastInverseDCT8_B16, fastInverseDCT8_B32, fastInverseDCT8_B64, fastInverseDCT8_B128, fastInverseDCT8_B256 },
{ nullptr, fastInverseDST7_B4, fastInverseDST7_B8, fastInverseDST7_B16, fastInverseDST7_B32, fastInverseDST7_B64, fastInverseDST7_B128, fastInverseDST7_B256 },
+#if JVET_W0103_INTRA_MTS
+ { nullptr, fastInverseDCT5_B4, fastInverseDCT5_B8, fastInverseDCT5_B16, fastInverseDCT5_B32, fastInverseDCT5_B64, fastInverseDCT5_B128, fastInverseDCT5_B256 },
+ { nullptr, fastInverseDST4_B4, fastInverseDST4_B8, fastInverseDST4_B16, fastInverseDST4_B32, fastInverseDST4_B64, fastInverseDST4_B128, fastInverseDST4_B256 },
+ { nullptr, fastInverseDST1_B4, fastInverseDST1_B8, fastInverseDST1_B16, fastInverseDST1_B32, fastInverseDST1_B64, fastInverseDST1_B128, fastInverseDST1_B256 },
+ { nullptr, fastInverseIDTR_B4, fastInverseIDTR_B8, fastInverseIDTR_B16, fastInverseIDTR_B32, fastInverseIDTR_B64, fastInverseIDTR_B128, fastInverseIDTR_B256 },
+#endif
} };
#else
fastFwdTrans =
@@ -311,6 +323,12 @@ void TrQuant::init( const Quant* otherQuant,
{ fastForwardDCT2_B2, fastForwardDCT2_B4, fastForwardDCT2_B8, fastForwardDCT2_B16, fastForwardDCT2_B32, fastForwardDCT2_B64 },
{ nullptr, fastForwardDCT8_B4, fastForwardDCT8_B8, fastForwardDCT8_B16, fastForwardDCT8_B32, nullptr },
{ nullptr, fastForwardDST7_B4, fastForwardDST7_B8, fastForwardDST7_B16, fastForwardDST7_B32, nullptr },
+#if JVET_W0103_INTRA_MTS
+ { nullptr, fastForwardDCT5_B4, fastForwardDCT5_B8, fastForwardDCT5_B16, fastForwardDCT5_B32, nullptr },
+ { nullptr, fastForwardDST4_B4, fastForwardDST4_B8, fastForwardDST4_B16, fastForwardDST4_B32, nullptr },
+ { nullptr, fastForwardDST1_B4, fastForwardDST1_B8, fastForwardDST1_B16, fastForwardDST1_B32, nullptr },
+ { nullptr, fastForwardIDTR_B4, fastForwardIDTR_B8, fastForwardIDTR_B16, fastForwardIDTR_B32, nullptr },
+#endif
} };
fastInvTrans =
@@ -318,6 +336,12 @@ void TrQuant::init( const Quant* otherQuant,
{ fastInverseDCT2_B2, fastInverseDCT2_B4, fastInverseDCT2_B8, fastInverseDCT2_B16, fastInverseDCT2_B32, fastInverseDCT2_B64 },
{ nullptr, fastInverseDCT8_B4, fastInverseDCT8_B8, fastInverseDCT8_B16, fastInverseDCT8_B32, nullptr },
{ nullptr, fastInverseDST7_B4, fastInverseDST7_B8, fastInverseDST7_B16, fastInverseDST7_B32, nullptr },
+#if JVET_W0103_INTRA_MTS
+ { nullptr, fastInverseDCT5_B4, fastInverseDCT5_B8, fastInverseDCT5_B16, fastInverseDCT5_B32, nullptr },
+ { nullptr, fastInverseDST4_B4, fastInverseDST4_B8, fastInverseDST4_B16, fastInverseDST4_B32, nullptr },
+ { nullptr, fastInverseDST1_B4, fastInverseDST1_B8, fastInverseDST1_B16, fastInverseDST1_B32, nullptr },
+ { nullptr, fastInverseIDTR_B4, fastInverseIDTR_B8, fastInverseIDTR_B16, fastInverseIDTR_B32, nullptr },
+#endif
} };
#endif
@@ -1323,6 +1347,64 @@ void TrQuant::getTrTypes(const TransformUnit tu, const ComponentID compID, int &
if (isExplicitMTS)
{
+#if JVET_W0103_INTRA_MTS
+ if (tu.mtsIdx[compID] > MTS_SKIP && CU::isIntra(*tu.cu))
+ {
+ CHECK(compID != COMPONENT_Y, " MTS activated for chroma");
+ uint32_t width = tu.blocks[compID].width;
+ uint32_t height = tu.blocks[compID].height;
+ int TrIdx = (tu.mtsIdx[compID] - MTS_DST7_DST7);
+ CHECK(width < 4 || height < 4, "width < 4 || height < 4 for MTS");
+ uint8_t nSzIdxW = std::min(3, (floorLog2(width) - 2));
+ uint8_t nSzIdxH = std::min(3, (floorLog2(height) - 2));
+ const CompArea& area = tu.blocks[compID];
+ int predMode = PU::getFinalIntraMode(*tu.cs->getPU(area.pos(), toChannelType(compID)), toChannelType(compID));
+#if JVET_W0123_TIMD_FUSION
+ if (tu.cu->timd && compID == COMPONENT_Y)
+ {
+ predMode = MAP131TO67(predMode);
+ }
+#endif
+ int ucMode;
+ int nMdIdx;
+ bool isTrTransposed = false;
+ if (tu.cu->mipFlag) //MIP is treated as planar.
+ {
+ ucMode = 0;
+ nMdIdx = 35;
+ isTrTransposed = (tu.cs->getPU(area.pos(), toChannelType(compID)))->mipTransposedFlag;
+ }
+ else
+ {
+ ucMode = predMode; //"ucMode" is the signaled Mode.
+ predMode = PU::getWideAngle(tu, (uint32_t)predMode, compID);
+ CHECK(predMode < -(NUM_EXT_LUMA_MODE >> 1) && predMode >= NUM_LUMA_MODE + (NUM_EXT_LUMA_MODE >> 1), "luma mode out of range");
+ predMode = (predMode < 0) ? 2 : (predMode >= NUM_LUMA_MODE) ? 66 : predMode;
+ nMdIdx = predMode > DIA_IDX ? (NUM_LUMA_MODE + 1 - predMode) : predMode;
+ isTrTransposed = (predMode > DIA_IDX) ? true : false;
+ }
+ uint8_t nSzIdx = isTrTransposed ? (nSzIdxH * 4 + nSzIdxW) : (nSzIdxW * 4 + nSzIdxH);
+ CHECK(nSzIdx >= 16, "nSzIdx >= 16");
+ CHECK(nMdIdx >= 36, "nMdIdx >= 36");
+ uint8_t nTrSet = g_aucIpmToTrSet[nSzIdx][nMdIdx];
+ CHECK(nTrSet >= 80, "nTrSet >= 80");
+ trTypeVer = g_aucTrIdxToTr[g_aucTrSet[nTrSet][TrIdx]][predMode > DIA_IDX ? 1 : 0];
+ trTypeHor = g_aucTrIdxToTr[g_aucTrSet[nTrSet][TrIdx]][predMode > DIA_IDX ? 0 : 1];
+ predMode = ucMode; //to Check IDTR criteria, signaled mode should be used to check the difference
+ if (TrIdx == 3 && width <= 16 && height <= 16)
+ {
+ if (abs(predMode - HOR_IDX) <= g_aiIdLut[floorLog2(width) - 2][floorLog2(height) - 2])
+ {
+ trTypeVer = IDTR;
+ }
+ if (abs(predMode - VER_IDX) <= g_aiIdLut[floorLog2(width) - 2][floorLog2(height) - 2])
+ {
+ trTypeHor = IDTR;
+ }
+ }
+ }
+ else
+#endif
if (tu.mtsIdx[compID] > MTS_SKIP)
{
int indHor = (tu.mtsIdx[compID] - MTS_DST7_DST7) & 1;
@@ -1621,7 +1703,35 @@ void TrQuant::transformNxN( TransformUnit& tu, const ComponentID& compID, const
itC++;
}
}
+#if JVET_W0103_INTRA_MTS
+// does transform for MTS candidates and return absSum of unquant Coeffs.
+uint64_t TrQuant::transformNxN(TransformUnit& tu)
+{
+ CHECK(!tu.cu->mtsFlag, "mtsFlag should be on for selection");
+ CodingStructure &cs = *tu.cs;
+ const CompArea &rect = tu.blocks[COMPONENT_Y];
+ const uint32_t uiWidth = rect.width;
+ const uint32_t uiHeight = rect.height;
+
+ const CPelBuf resiBuf = cs.getResiBuf(rect);
+ CoeffBuf tempCoeff(m_mtsCoeffs[tu.mtsIdx[0]], rect);
+ xT(tu, COMPONENT_Y, resiBuf, tempCoeff, uiWidth, uiHeight);
+
+ const TCoeff *dstCoeffBuf = tempCoeff.buf;
+ const int dstCoeffStride = tempCoeff.stride;
+ uint64_t coeffAbsSum = 0;
+
+ for (int y = 0; y < uiHeight; y++)
+ {
+ for (int x = 0; x < uiWidth; x++)
+ {
+ coeffAbsSum += abs(dstCoeffBuf[(y * dstCoeffStride) + x]);
+ }
+ }
+ return coeffAbsSum;
+}
+#endif
void TrQuant::transformNxN( TransformUnit& tu, const ComponentID& compID, const QpParam& cQP, TCoeff& uiAbsSum, const Ctx& ctx, const bool loadTr )
{
CodingStructure &cs = *tu.cs;
@@ -1803,7 +1913,11 @@ void TrQuant::predCoeffSigns(TransformUnit &tu, const ComponentID compID, const
PelBuf bufResiTemplateReshape(predResiTemplateReshape, uiWidth + uiHeight - 1, 1);
int trHor, trVer;
getTrTypes(tu, residCompID, trHor, trVer);
+#if JVET_W0103_INTRA_MTS
+ int actualTrIdx = trHor * NUM_TRANS_TYPE + trVer;
+#else
int actualTrIdx = trHor * 3 + trVer;
+#endif
int log2Width = floorLog2(uiWidth);
int log2Height = floorLog2(uiHeight);
if(!g_resiBorderTemplate[log2Width-2][log2Height-2][actualTrIdx])
diff --git a/source/Lib/CommonLib/TrQuant.h b/source/Lib/CommonLib/TrQuant.h
index 8556825e4e00241ab087347b8a5609876dfea999..77d99074a3565ce3cce5c314fb2b589fe8dd38fd 100644
--- a/source/Lib/CommonLib/TrQuant.h
+++ b/source/Lib/CommonLib/TrQuant.h
@@ -157,6 +157,9 @@ public:
void invTransformNxN (TransformUnit &tu, const ComponentID &compID, PelBuf &pResi, const QpParam &cQPs);
void transformNxN ( TransformUnit& tu, const ComponentID& compID, const QpParam& cQP, std::vector<TrMode>* trModes, const int maxCand );
void transformNxN ( TransformUnit& tu, const ComponentID& compID, const QpParam& cQP, TCoeff& uiAbsSum, const Ctx& ctx, const bool loadTr = false );
+#if JVET_W0103_INTRA_MTS
+ uint64_t transformNxN(TransformUnit& tu);
+#endif
void transformSkipQuantOneSample(TransformUnit &tu, const ComponentID &compID, const TCoeff &resiDiff, TCoeff &coeff, const uint32_t &uiPos, const QpParam &cQP, const bool bUseHalfRoundingPoint);
void invTrSkipDeQuantOneSample (TransformUnit &tu, const ComponentID &compID, const TCoeff &pcCoeff, Pel &reconSample, const uint32_t &uiPos, const QpParam &cQP);
diff --git a/source/Lib/CommonLib/TrQuant_EMT.cpp b/source/Lib/CommonLib/TrQuant_EMT.cpp
index e0e8ef8ccdb917ca379a3dc5a217694a0c2cbd6a..6912cb56b000dc05c8137b546b633b1fe45cea87 100644
--- a/source/Lib/CommonLib/TrQuant_EMT.cpp
+++ b/source/Lib/CommonLib/TrQuant_EMT.cpp
@@ -1973,4 +1973,289 @@ void fastInverseDCT8_B256( const TCoeff *src, TCoeff *dst, int shift, int line,
{
_fastInverseMM< 256 >( src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_trCoreDCT8P256[0] );
}
+#endif
+#if JVET_W0103_INTRA_MTS
+//DCT5
+void fastForwardDCT5_B4(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
+{
+ _fastForwardMM< 4 >(src, dst, shift, line, iSkipLine, iSkipLine2, g_aiTr4[DCT5][0]);
+}
+
+void fastInverseDCT5_B4(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
+{
+ _fastInverseMM< 4 >(src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_aiTr4[DCT5][0]);
+}
+
+void fastForwardDCT5_B8(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
+{
+ _fastForwardMM< 8 >(src, dst, shift, line, iSkipLine, iSkipLine2, g_aiTr8[DCT5][0]);
+}
+
+void fastInverseDCT5_B8(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
+{
+ _fastInverseMM< 8 >(src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_aiTr8[DCT5][0]);
+}
+
+void fastForwardDCT5_B16(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
+{
+ _fastForwardMM< 16 >(src, dst, shift, line, iSkipLine, iSkipLine2, g_aiTr16[DCT5][0]);
+}
+
+void fastInverseDCT5_B16(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
+{
+ _fastInverseMM< 16 >(src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_aiTr16[DCT5][0]);
+}
+
+void fastForwardDCT5_B32(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
+{
+ _fastForwardMM< 32 >(src, dst, shift, line, iSkipLine, iSkipLine2, g_aiTr32[DCT5][0]);
+}
+
+void fastInverseDCT5_B32(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
+{
+ _fastInverseMM< 32 >(src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_aiTr32[DCT5][0]);
+}
+
+void fastForwardDCT5_B64(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
+{
+ _fastForwardMM< 64 >(src, dst, shift, line, iSkipLine, iSkipLine2, g_aiTr64[DCT5][0]);
+}
+
+void fastInverseDCT5_B64(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
+{
+ _fastInverseMM< 64 >(src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_aiTr64[DCT5][0]);
+}
+
+void fastForwardDCT5_B128(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
+{
+ _fastForwardMM< 128 >(src, dst, shift, line, iSkipLine, iSkipLine2, g_aiTr128[DCT5][0]);
+}
+
+void fastInverseDCT5_B128(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
+{
+ _fastInverseMM< 128 >(src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_aiTr128[DCT5][0]);
+}
+
+void fastForwardDCT5_B256(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
+{
+ _fastForwardMM< 256 >(src, dst, shift, line, iSkipLine, iSkipLine2, g_aiTr256[DCT5][0]);
+}
+
+void fastInverseDCT5_B256(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
+{
+ _fastInverseMM< 256 >(src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_aiTr256[DCT5][0]);
+}
+
+//DST4
+void fastForwardDST4_B4(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
+{
+ _fastForwardMM< 4 >(src, dst, shift, line, iSkipLine, iSkipLine2, g_aiTr4[DST4][0]);
+}
+
+void fastInverseDST4_B4(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
+{
+ _fastInverseMM< 4 >(src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_aiTr4[DST4][0]);
+}
+
+void fastForwardDST4_B8(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
+{
+ _fastForwardMM< 8 >(src, dst, shift, line, iSkipLine, iSkipLine2, g_aiTr8[DST4][0]);
+}
+
+void fastInverseDST4_B8(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
+{
+ _fastInverseMM< 8 >(src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_aiTr8[DST4][0]);
+}
+
+void fastForwardDST4_B16(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
+{
+ _fastForwardMM< 16 >(src, dst, shift, line, iSkipLine, iSkipLine2, g_aiTr16[DST4][0]);
+}
+
+void fastInverseDST4_B16(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
+{
+ _fastInverseMM< 16 >(src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_aiTr16[DST4][0]);
+}
+
+void fastForwardDST4_B32(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
+{
+ _fastForwardMM< 32 >(src, dst, shift, line, iSkipLine, iSkipLine2, g_aiTr32[DST4][0]);
+}
+
+void fastInverseDST4_B32(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
+{
+ _fastInverseMM< 32 >(src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_aiTr32[DST4][0]);
+}
+
+void fastForwardDST4_B64(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
+{
+ _fastForwardMM< 64 >(src, dst, shift, line, iSkipLine, iSkipLine2, g_aiTr64[DST4][0]);
+}
+
+void fastInverseDST4_B64(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
+{
+ _fastInverseMM< 64 >(src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_aiTr64[DST4][0]);
+}
+
+void fastForwardDST4_B128(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
+{
+ _fastForwardMM< 128 >(src, dst, shift, line, iSkipLine, iSkipLine2, g_aiTr128[DST4][0]);
+}
+
+void fastInverseDST4_B128(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
+{
+ _fastInverseMM< 128 >(src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_aiTr128[DST4][0]);
+}
+
+void fastForwardDST4_B256(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
+{
+ _fastForwardMM< 256 >(src, dst, shift, line, iSkipLine, iSkipLine2, g_aiTr256[DST4][0]);
+}
+
+void fastInverseDST4_B256(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
+{
+ _fastInverseMM< 256 >(src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_aiTr256[DST4][0]);
+}
+
+//DST1
+void fastForwardDST1_B4(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
+{
+ _fastForwardMM< 4 >(src, dst, shift, line, iSkipLine, iSkipLine2, g_aiTr4[DST1][0]);
+}
+
+void fastInverseDST1_B4(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
+{
+ _fastInverseMM< 4 >(src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_aiTr4[DST1][0]);
+}
+
+void fastForwardDST1_B8(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
+{
+ _fastForwardMM< 8 >(src, dst, shift, line, iSkipLine, iSkipLine2, g_aiTr8[DST1][0]);
+}
+
+void fastInverseDST1_B8(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
+{
+ _fastInverseMM< 8 >(src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_aiTr8[DST1][0]);
+}
+
+void fastForwardDST1_B16(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
+{
+ _fastForwardMM< 16 >(src, dst, shift, line, iSkipLine, iSkipLine2, g_aiTr16[DST1][0]);
+}
+
+void fastInverseDST1_B16(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
+{
+ _fastInverseMM< 16 >(src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_aiTr16[DST1][0]);
+}
+
+void fastForwardDST1_B32(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
+{
+ _fastForwardMM< 32 >(src, dst, shift, line, iSkipLine, iSkipLine2, g_aiTr32[DST1][0]);
+}
+
+void fastInverseDST1_B32(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
+{
+ _fastInverseMM< 32 >(src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_aiTr32[DST1][0]);
+}
+
+void fastForwardDST1_B64(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
+{
+ _fastForwardMM< 64 >(src, dst, shift, line, iSkipLine, iSkipLine2, g_aiTr64[DST1][0]);
+}
+
+void fastInverseDST1_B64(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
+{
+ _fastInverseMM< 64 >(src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_aiTr64[DST1][0]);
+}
+
+void fastForwardDST1_B128(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
+{
+ _fastForwardMM< 128 >(src, dst, shift, line, iSkipLine, iSkipLine2, g_aiTr128[DST1][0]);
+}
+
+void fastInverseDST1_B128(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
+{
+ _fastInverseMM< 128 >(src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_aiTr128[DST1][0]);
+}
+
+void fastForwardDST1_B256(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
+{
+ _fastForwardMM< 256 >(src, dst, shift, line, iSkipLine, iSkipLine2, g_aiTr256[DST1][0]);
+}
+
+void fastInverseDST1_B256(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
+{
+ _fastInverseMM< 256 >(src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_aiTr256[DST1][0]);
+}
+
+//IDTR
+void fastForwardIDTR_B4(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
+{
+ _fastForwardMM< 4 >(src, dst, shift, line, iSkipLine, iSkipLine2, g_aiTr4[IDTR][0]);
+}
+
+void fastInverseIDTR_B4(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
+{
+ _fastInverseMM< 4 >(src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_aiTr4[IDTR][0]);
+}
+
+void fastForwardIDTR_B8(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
+{
+ _fastForwardMM< 8 >(src, dst, shift, line, iSkipLine, iSkipLine2, g_aiTr8[IDTR][0]);
+}
+
+void fastInverseIDTR_B8(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
+{
+ _fastInverseMM< 8 >(src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_aiTr8[IDTR][0]);
+}
+
+void fastForwardIDTR_B16(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
+{
+ _fastForwardMM< 16 >(src, dst, shift, line, iSkipLine, iSkipLine2, g_aiTr16[IDTR][0]);
+}
+
+void fastInverseIDTR_B16(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
+{
+ _fastInverseMM< 16 >(src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_aiTr16[IDTR][0]);
+}
+
+void fastForwardIDTR_B32(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
+{
+ _fastForwardMM< 32 >(src, dst, shift, line, iSkipLine, iSkipLine2, g_aiTr32[IDTR][0]);
+}
+
+void fastInverseIDTR_B32(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
+{
+ _fastInverseMM< 32 >(src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_aiTr32[IDTR][0]);
+}
+
+void fastForwardIDTR_B64(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
+{
+ _fastForwardMM< 64 >(src, dst, shift, line, iSkipLine, iSkipLine2, g_aiTr64[IDTR][0]);
+}
+
+void fastInverseIDTR_B64(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
+{
+ _fastInverseMM< 64 >(src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_aiTr64[IDTR][0]);
+}
+
+void fastForwardIDTR_B128(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
+{
+ _fastForwardMM< 128 >(src, dst, shift, line, iSkipLine, iSkipLine2, g_aiTr128[IDTR][0]);
+}
+
+void fastInverseIDTR_B128(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
+{
+ _fastInverseMM< 128 >(src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_aiTr128[IDTR][0]);
+}
+
+void fastForwardIDTR_B256(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
+{
+ _fastForwardMM< 256 >(src, dst, shift, line, iSkipLine, iSkipLine2, g_aiTr256[IDTR][0]);
+}
+
+void fastInverseIDTR_B256(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
+{
+ _fastInverseMM< 256 >(src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_aiTr256[IDTR][0]);
+}
#endif
\ No newline at end of file
diff --git a/source/Lib/CommonLib/TrQuant_EMT.h b/source/Lib/CommonLib/TrQuant_EMT.h
index 03d7491dce4ee3196c0cf5e17e339eb5cb1e6af1..2379795dacd6027157d2c3f085a199854d67b670 100644
--- a/source/Lib/CommonLib/TrQuant_EMT.h
+++ b/source/Lib/CommonLib/TrQuant_EMT.h
@@ -97,5 +97,78 @@ void fastForwardDCT8_B256( const TCoeff *src, TCoeff *dst, int shift, int line,
void fastInverseDCT8_B256( const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum );
#endif
+#if JVET_W0103_INTRA_MTS
+//DST-IV transforms
+void fastForwardDST4_B4(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2);
+void fastInverseDST4_B4(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum);
+void fastForwardDST4_B8(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2);
+void fastInverseDST4_B8(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum);
+void fastForwardDST4_B16(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2);
+void fastInverseDST4_B16(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum);
+void fastForwardDST4_B32(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2);
+void fastInverseDST4_B32(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum);
+void fastForwardDST4_B64(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2);
+void fastInverseDST4_B64(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum);
+#if TU_256
+void fastForwardDST4_B128(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2);
+void fastInverseDST4_B128(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum);
+void fastForwardDST4_B256(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2);
+void fastInverseDST4_B256(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum);
+#endif
+
+//DST-I transforms
+void fastForwardDST1_B4(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2);
+void fastInverseDST1_B4(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum);
+void fastForwardDST1_B8(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2);
+void fastInverseDST1_B8(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum);
+void fastForwardDST1_B16(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2);
+void fastInverseDST1_B16(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum);
+void fastForwardDST1_B32(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2);
+void fastInverseDST1_B32(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum);
+void fastForwardDST1_B64(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2);
+void fastInverseDST1_B64(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum);
+#if TU_256
+void fastForwardDST1_B128(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2);
+void fastInverseDST1_B128(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum);
+void fastForwardDST1_B256(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2);
+void fastInverseDST1_B256(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum);
+#endif
+
+//DCT-V transforms
+void fastForwardDCT5_B4(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2);
+void fastInverseDCT5_B4(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum);
+void fastForwardDCT5_B8(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2);
+void fastInverseDCT5_B8(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum);
+void fastForwardDCT5_B16(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2);
+void fastInverseDCT5_B16(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum);
+void fastForwardDCT5_B32(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2);
+void fastInverseDCT5_B32(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum);
+void fastForwardDCT5_B64(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2);
+void fastInverseDCT5_B64(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum);
+#if TU_256
+void fastForwardDCT5_B128(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2);
+void fastInverseDCT5_B128(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum);
+void fastForwardDCT5_B256(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2);
+void fastInverseDCT5_B256(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum);
+#endif
+
+//IDTR transforms
+void fastForwardIDTR_B4(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2);
+void fastInverseIDTR_B4(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum);
+void fastForwardIDTR_B8(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2);
+void fastInverseIDTR_B8(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum);
+void fastForwardIDTR_B16(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2);
+void fastInverseIDTR_B16(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum);
+void fastForwardIDTR_B32(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2);
+void fastInverseIDTR_B32(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum);
+void fastForwardIDTR_B64(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2);
+void fastInverseIDTR_B64(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum);
+#if TU_256
+void fastForwardIDTR_B128(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2);
+void fastInverseIDTR_B128(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum);
+void fastForwardIDTR_B256(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2);
+void fastInverseIDTR_B256(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum);
+#endif
+#endif
#endif // __TRQUANT__
diff --git a/source/Lib/CommonLib/TypeDef.h b/source/Lib/CommonLib/TypeDef.h
index 63d251c2662434aba131f547bb649856a53767d5..83d7c38c4c00b07468f3815a5f00b0e51a6bad89 100644
--- a/source/Lib/CommonLib/TypeDef.h
+++ b/source/Lib/CommonLib/TypeDef.h
@@ -52,10 +52,13 @@
+
+
#define BASE_ENCODER 1
#define BASE_NORMATIVE 1
#define TOOLS 1
+
#if BASE_ENCODER
// Lossy encoder speedups
#define AFFINE_ENC_OPT 1 // Affine encoder optimization
@@ -132,6 +135,8 @@
#define EXTENDED_LFNST 1 // Extended LFNST
#define SIGN_PREDICTION 1 // transform coefficients sign prediction
+#define JVET_W0103_INTRA_MTS 1 // JVET W0103: Extended Intra MTS
+
// Entropy Coding
#define EC_HIGH_PRECISION 1 // CABAC high precision
#define SLICE_TYPE_WIN_SIZE 1 // Context window initialization based on slice type
@@ -157,7 +162,6 @@
#if JVET_V0094_BILATERAL_FILTER
#define ENABLE_SIMD_BILATERAL_FILTER 1
#endif
-
#endif // tools
@@ -541,11 +545,23 @@ enum QuantFlags
//EMT transform tags
enum TransType
{
+#if JVET_W0103_INTRA_MTS
+ DCT2 = 0,
+ DCT8 = 1,
+ DST7 = 2,
+ DCT5 = 3,
+ DST4 = 4,
+ DST1 = 5,
+ IDTR = 6,
+ NUM_TRANS_TYPE = 7,
+ DCT2_EMT = 8
+#else
DCT2 = 0,
DCT8 = 1,
DST7 = 2,
NUM_TRANS_TYPE = 3,
DCT2_EMT = 4
+#endif
};
enum MTSIdx
diff --git a/source/Lib/CommonLib/x86/TrQuantX86.h b/source/Lib/CommonLib/x86/TrQuantX86.h
index 5f7238fc60ea6aa9385a797399af4a7ccb3b524f..97f5215e77fc940b6dee55727666c623d283dc88 100644
--- a/source/Lib/CommonLib/x86/TrQuantX86.h
+++ b/source/Lib/CommonLib/x86/TrQuantX86.h
@@ -547,6 +547,12 @@ void TrQuant::_initTrQuantX86()
{ g_trCoreDCT2P2[TRANSFORM_FORWARD][0], g_trCoreDCT2P4[TRANSFORM_FORWARD][0], g_trCoreDCT2P8[TRANSFORM_FORWARD][0], g_trCoreDCT2P16[TRANSFORM_FORWARD][0], g_trCoreDCT2P32[TRANSFORM_FORWARD][0], g_trCoreDCT2P64[TRANSFORM_FORWARD][0], g_trCoreDCT2P128[TRANSFORM_FORWARD][0], g_trCoreDCT2P256[0] },
{ nullptr, g_trCoreDCT8P4[TRANSFORM_FORWARD][0], g_trCoreDCT8P8[TRANSFORM_FORWARD][0], g_trCoreDCT8P16[TRANSFORM_FORWARD][0], g_trCoreDCT8P32[TRANSFORM_FORWARD][0], g_trCoreDCT8P64[TRANSFORM_FORWARD][0], g_trCoreDCT8P128[TRANSFORM_FORWARD][0], g_trCoreDCT8P256[0] },
{ nullptr, g_trCoreDST7P4[TRANSFORM_FORWARD][0], g_trCoreDST7P8[TRANSFORM_FORWARD][0], g_trCoreDST7P16[TRANSFORM_FORWARD][0], g_trCoreDST7P32[TRANSFORM_FORWARD][0], g_trCoreDST7P64[TRANSFORM_FORWARD][0], g_trCoreDST7P128[TRANSFORM_FORWARD][0], g_trCoreDST7P256[0] },
+#if JVET_W0103_INTRA_MTS
+ { nullptr, g_aiTr4[DCT5][0], g_aiTr8[DCT5][0], g_aiTr16[DCT5][0], g_aiTr32[DCT5][0], g_aiTr64[DCT5][0], g_aiTr128[DCT5][0], g_aiTr256[DCT5][0] },
+ { nullptr, g_aiTr4[DST4][0], g_aiTr8[DST4][0], g_aiTr16[DST4][0], g_aiTr32[DST4][0], g_aiTr64[DST4][0], g_aiTr128[DST4][0], g_aiTr256[DST4][0] },
+ { nullptr, g_aiTr4[DST1][0], g_aiTr8[DST1][0], g_aiTr16[DST1][0], g_aiTr32[DST1][0], g_aiTr64[DST1][0], g_aiTr128[DST1][0], g_aiTr256[DST1][0] },
+ { nullptr, g_aiTr4[IDTR][0], g_aiTr8[IDTR][0], g_aiTr16[IDTR][0], g_aiTr32[IDTR][0], g_aiTr64[IDTR][0], g_aiTr128[IDTR][0], g_aiTr256[IDTR][0] },
+#endif
} };
m_inverseTransformKernels =
@@ -554,6 +560,12 @@ void TrQuant::_initTrQuantX86()
{ g_trCoreDCT2P2[TRANSFORM_INVERSE][0], g_trCoreDCT2P4[TRANSFORM_INVERSE][0], g_trCoreDCT2P8[TRANSFORM_INVERSE][0], g_trCoreDCT2P16[TRANSFORM_INVERSE][0], g_trCoreDCT2P32[TRANSFORM_INVERSE][0], g_trCoreDCT2P64[TRANSFORM_INVERSE][0], g_trCoreDCT2P128[TRANSFORM_INVERSE][0], g_trCoreDCT2P256[0] },
{ nullptr, g_trCoreDCT8P4[TRANSFORM_INVERSE][0], g_trCoreDCT8P8[TRANSFORM_INVERSE][0], g_trCoreDCT8P16[TRANSFORM_INVERSE][0], g_trCoreDCT8P32[TRANSFORM_INVERSE][0], g_trCoreDCT8P64[TRANSFORM_INVERSE][0], g_trCoreDCT8P128[TRANSFORM_INVERSE][0], g_trCoreDCT8P256[0] },
{ nullptr, g_trCoreDST7P4[TRANSFORM_INVERSE][0], g_trCoreDST7P8[TRANSFORM_INVERSE][0], g_trCoreDST7P16[TRANSFORM_INVERSE][0], g_trCoreDST7P32[TRANSFORM_INVERSE][0], g_trCoreDST7P64[TRANSFORM_INVERSE][0], g_trCoreDST7P128[TRANSFORM_INVERSE][0], g_trCoreDST7P256[0] },
+#if JVET_W0103_INTRA_MTS
+ { nullptr, g_aiTr4[DCT5][0], g_aiTr8[DCT5][0], g_aiTr16[DCT5][0], g_aiTr32[DCT5][0], g_aiTr64[DCT5][0], g_aiTr128[DCT5][0], g_aiTr256[DCT5][0] },
+ { nullptr, g_aiTr4[DST4][0], g_aiTr8[DST4][0], g_aiTr16[DST4][0], g_aiTr32[DST4][0], g_aiTr64[DST4][0], g_aiTr128[DST4][0], g_aiTr256[DST4][0] },
+ { nullptr, g_aiTr4[DST1][0], g_aiTr8[DST1][0], g_aiTr16[DST1][0], g_aiTr32[DST1][0], g_aiTr64[DST1][0], g_aiTr128[DST1][0], g_aiTr256[DST1][0] },
+ { nullptr, g_aiTr4[IDTR][0], g_aiTr8[IDTR][0], g_aiTr16[IDTR][0], g_aiTr32[IDTR][0], g_aiTr64[IDTR][0], g_aiTr128[IDTR][0], g_aiTr256[IDTR][0] },
+#endif
} };
fastFwdTrans[0][0] = fastForwardTransform_SIMD<DCT2, 2>;
@@ -583,6 +595,44 @@ void TrQuant::_initTrQuantX86()
fastFwdTrans[2][6] = fastForwardTransform_SIMD<DST7, 128>;
fastFwdTrans[2][7] = fastForwardTransform_SIMD<DST7, 256>;
+#if JVET_W0103_INTRA_MTS
+ fastFwdTrans[3][0] = nullptr;
+ fastFwdTrans[3][1] = fastForwardTransform_SIMD<DCT5, 4>;
+ fastFwdTrans[3][2] = fastForwardTransform_SIMD<DCT5, 8>;
+ fastFwdTrans[3][3] = fastForwardTransform_SIMD<DCT5, 16>;
+ fastFwdTrans[3][4] = fastForwardTransform_SIMD<DCT5, 32>;
+ fastFwdTrans[3][5] = fastForwardTransform_SIMD<DCT5, 64>;
+ fastFwdTrans[3][6] = fastForwardTransform_SIMD<DCT5, 128>;
+ fastFwdTrans[3][7] = fastForwardTransform_SIMD<DCT5, 256>;
+
+ fastFwdTrans[4][0] = nullptr;
+ fastFwdTrans[4][1] = fastForwardTransform_SIMD<DST4, 4>;
+ fastFwdTrans[4][2] = fastForwardTransform_SIMD<DST4, 8>;
+ fastFwdTrans[4][3] = fastForwardTransform_SIMD<DST4, 16>;
+ fastFwdTrans[4][4] = fastForwardTransform_SIMD<DST4, 32>;
+ fastFwdTrans[4][5] = fastForwardTransform_SIMD<DST4, 64>;
+ fastFwdTrans[4][6] = fastForwardTransform_SIMD<DST4, 128>;
+ fastFwdTrans[4][7] = fastForwardTransform_SIMD<DST4, 256>;
+
+ fastFwdTrans[5][0] = nullptr;
+ fastFwdTrans[5][1] = fastForwardTransform_SIMD<DST1, 4>;
+ fastFwdTrans[5][2] = fastForwardTransform_SIMD<DST1, 8>;
+ fastFwdTrans[5][3] = fastForwardTransform_SIMD<DST1, 16>;
+ fastFwdTrans[5][4] = fastForwardTransform_SIMD<DST1, 32>;
+ fastFwdTrans[5][5] = fastForwardTransform_SIMD<DST1, 64>;
+ fastFwdTrans[5][6] = fastForwardTransform_SIMD<DST1, 128>;
+ fastFwdTrans[5][7] = fastForwardTransform_SIMD<DST1, 256>;
+
+ fastFwdTrans[6][0] = nullptr;
+ fastFwdTrans[6][1] = fastForwardTransform_SIMD<IDTR, 4>;
+ fastFwdTrans[6][2] = fastForwardTransform_SIMD<IDTR, 8>;
+ fastFwdTrans[6][3] = fastForwardTransform_SIMD<IDTR, 16>;
+ fastFwdTrans[6][4] = fastForwardTransform_SIMD<IDTR, 32>;
+ fastFwdTrans[6][5] = fastForwardTransform_SIMD<IDTR, 64>;
+ fastFwdTrans[6][6] = fastForwardTransform_SIMD<IDTR, 128>;
+ fastFwdTrans[6][7] = fastForwardTransform_SIMD<IDTR, 256>;
+#endif
+
fastInvTrans[0][0] = fastInverseTransform_SIMD<DCT2, 2>;
fastInvTrans[0][1] = fastInverseTransform_SIMD<DCT2, 4>;
fastInvTrans[0][2] = fastInverseTransform_SIMD<DCT2, 8>;
@@ -609,6 +659,44 @@ void TrQuant::_initTrQuantX86()
fastInvTrans[2][5] = fastInverseTransform_SIMD<DST7, 64>;
fastInvTrans[2][6] = fastInverseTransform_SIMD<DST7, 128>;
fastInvTrans[2][7] = fastInverseTransform_SIMD<DST7, 256>;
+
+#if JVET_W0103_INTRA_MTS
+ fastInvTrans[3][0] = nullptr;
+ fastInvTrans[3][1] = fastInverseTransform_SIMD<DCT5, 4>;
+ fastInvTrans[3][2] = fastInverseTransform_SIMD<DCT5, 8>;
+ fastInvTrans[3][3] = fastInverseTransform_SIMD<DCT5, 16>;
+ fastInvTrans[3][4] = fastInverseTransform_SIMD<DCT5, 32>;
+ fastInvTrans[3][5] = fastInverseTransform_SIMD<DCT5, 64>;
+ fastInvTrans[3][6] = fastInverseTransform_SIMD<DCT5, 128>;
+ fastInvTrans[3][7] = fastInverseTransform_SIMD<DCT5, 256>;
+
+ fastInvTrans[4][0] = nullptr;
+ fastInvTrans[4][1] = fastInverseTransform_SIMD<DST4, 4>;
+ fastInvTrans[4][2] = fastInverseTransform_SIMD<DST4, 8>;
+ fastInvTrans[4][3] = fastInverseTransform_SIMD<DST4, 16>;
+ fastInvTrans[4][4] = fastInverseTransform_SIMD<DST4, 32>;
+ fastInvTrans[4][5] = fastInverseTransform_SIMD<DST4, 64>;
+ fastInvTrans[4][6] = fastInverseTransform_SIMD<DST4, 128>;
+ fastInvTrans[4][7] = fastInverseTransform_SIMD<DST4, 256>;
+
+ fastInvTrans[5][0] = nullptr;
+ fastInvTrans[5][1] = fastInverseTransform_SIMD<DST1, 4>;
+ fastInvTrans[5][2] = fastInverseTransform_SIMD<DST1, 8>;
+ fastInvTrans[5][3] = fastInverseTransform_SIMD<DST1, 16>;
+ fastInvTrans[5][4] = fastInverseTransform_SIMD<DST1, 32>;
+ fastInvTrans[5][5] = fastInverseTransform_SIMD<DST1, 64>;
+ fastInvTrans[5][6] = fastInverseTransform_SIMD<DST1, 128>;
+ fastInvTrans[5][7] = fastInverseTransform_SIMD<DST1, 256>;
+
+ fastInvTrans[6][0] = nullptr;
+ fastInvTrans[6][1] = fastInverseTransform_SIMD<IDTR, 4>;
+ fastInvTrans[6][2] = fastInverseTransform_SIMD<IDTR, 8>;
+ fastInvTrans[6][3] = fastInverseTransform_SIMD<IDTR, 16>;
+ fastInvTrans[6][4] = fastInverseTransform_SIMD<IDTR, 32>;
+ fastInvTrans[6][5] = fastInverseTransform_SIMD<IDTR, 64>;
+ fastInvTrans[6][6] = fastInverseTransform_SIMD<IDTR, 128>;
+ fastInvTrans[6][7] = fastInverseTransform_SIMD<IDTR, 256>;
+#endif
#else
m_forwardTransformKernels =
{ {
diff --git a/source/Lib/DecoderLib/CABACReader.cpp b/source/Lib/DecoderLib/CABACReader.cpp
index a0ced2ca8a07d59063fdaf837ccf7d5bb7354ec7..2a7113eee1f8149d4147ed596a9b19ad3845ba2d 100644
--- a/source/Lib/DecoderLib/CABACReader.cpp
+++ b/source/Lib/DecoderLib/CABACReader.cpp
@@ -3858,11 +3858,21 @@ void CABACReader::mts_idx( CodingUnit& cu, CUCtx& cuCtx )
cuCtx.mtsLastScanPos && cu.lfnstIdx == 0 && mtsIdx != MTS_SKIP)
{
RExt__DECODER_DEBUG_BIT_STATISTICS_CREATE_SET_SIZE2( STATS__CABAC_BITS__MTS_FLAGS, tu.blocks[COMPONENT_Y], COMPONENT_Y );
+#if JVET_W0103_INTRA_MTS
+ int ctxIdx = (cu.mipFlag) ? 3 : 0;
+#else
int ctxIdx = 0;
+#endif
int symbol = m_BinDecoder.decodeBin( Ctx::MTSIdx(ctxIdx));
if( symbol )
{
+#if JVET_W0103_INTRA_MTS
+ int bins[2];
+ bins[0] = m_BinDecoder.decodeBin(Ctx::MTSIdx(1));
+ bins[1] = m_BinDecoder.decodeBin(Ctx::MTSIdx(2));
+ mtsIdx = MTS_DST7_DST7 + (bins[0] << 1) + bins[1];
+#else
ctxIdx = 1;
mtsIdx = MTS_DST7_DST7; // mtsIdx = 2 -- 4
for( int i = 0; i < 3; i++, ctxIdx++ )
@@ -3875,6 +3885,7 @@ void CABACReader::mts_idx( CodingUnit& cu, CUCtx& cuCtx )
break;
}
}
+#endif
}
}
diff --git a/source/Lib/EncoderLib/CABACWriter.cpp b/source/Lib/EncoderLib/CABACWriter.cpp
index 4904308cf850d1f40e3d256220d054e1791b8dc5..01b001838402ac20b70d35d1f1d8fcb418cfdc66 100644
--- a/source/Lib/EncoderLib/CABACWriter.cpp
+++ b/source/Lib/EncoderLib/CABACWriter.cpp
@@ -3604,12 +3604,22 @@ void CABACWriter::mts_idx( const CodingUnit& cu, CUCtx* cuCtx )
cuCtx->mtsLastScanPos && cu.lfnstIdx == 0 && mtsIdx != MTS_SKIP)
{
int symbol = mtsIdx != MTS_DCT2_DCT2 ? 1 : 0;
+#if JVET_W0103_INTRA_MTS
+ int ctxIdx = (cu.mipFlag) ? 3 : 0;
+#else
int ctxIdx = 0;
+#endif
m_BinEncoder.encodeBin( symbol, Ctx::MTSIdx(ctxIdx));
if( symbol )
{
+#if JVET_W0103_INTRA_MTS
+ int TrIdx = (tu.mtsIdx[COMPONENT_Y] - MTS_DST7_DST7);
+ CHECK(TrIdx < 0 || TrIdx >= 4, "TrIdx outside range");
+ m_BinEncoder.encodeBin(TrIdx >> 1, Ctx::MTSIdx(1));
+ m_BinEncoder.encodeBin(TrIdx & 1, Ctx::MTSIdx(2));
+#else
ctxIdx = 1;
for( int i = 0; i < 3; i++, ctxIdx++ )
{
@@ -3621,6 +3631,7 @@ void CABACWriter::mts_idx( const CodingUnit& cu, CUCtx* cuCtx )
break;
}
}
+#endif
}
}
DTRACE( g_trace_ctx, D_SYNTAX, "mts_idx() etype=%d pos=(%d,%d) mtsIdx=%d\n", COMPONENT_Y, tu.cu->lx(), tu.cu->ly(), mtsIdx);
diff --git a/source/Lib/EncoderLib/EncCu.cpp b/source/Lib/EncoderLib/EncCu.cpp
index 2142260474e0634d82ada0799cc17cc8bad0f82a..d6dc2b317dd81897ab461b923c9eca00e19e1048 100644
--- a/source/Lib/EncoderLib/EncCu.cpp
+++ b/source/Lib/EncoderLib/EncCu.cpp
@@ -1832,6 +1832,9 @@ bool EncCu::xCheckRDCostIntra(CodingStructure *&tempCS, CodingStructure *&bestCS
bool trGrpCheck [ 4 ] = { true, true, true, true };
int startMTSIdx [ 4 ] = { 0, 1, 2, 3 };
int endMTSIdx [ 4 ] = { 0, 1, 2, 3 };
+#if JVET_W0103_INTRA_MTS
+ endMTSIdx[0] = 3; //put all MTS candidates in "Grp 0"
+#endif
double trGrpStopThreshold[ 3 ] = { 1.001, 1.001, 1.001 };
int bestMtsFlag = 0;
int bestLfnstIdx = 0;
@@ -1857,8 +1860,11 @@ bool EncCu::xCheckRDCostIntra(CodingStructure *&tempCS, CodingStructure *&bestCS
bool skipOtherLfnst = false;
int startLfnstIdx = 0;
int endLfnstIdx = sps.getUseLFNST() ? maxLfnstIdx : 0;
-
+#if JVET_W0103_INTRA_MTS
+ int grpNumMax = 1;
+#else
int grpNumMax = sps.getUseLFNST() ? m_pcEncCfg->getMTSIntraMaxCand() : 1;
+#endif
m_modeCtrl->setISPWasTested(false);
m_pcIntraSearch->invalidateBestModeCost();
if (sps.getUseColorTrans() && !CS::isDualITree(*tempCS))
@@ -2222,7 +2228,6 @@ bool EncCu::xCheckRDCostIntra(CodingStructure *&tempCS, CodingStructure *&bestCS
xCalDebCost( *tempCS, partitioner );
tempCS->useDbCost = m_pcEncCfg->getUseEncDbOpt();
-
#if WCG_EXT
DTRACE_MODE_COST( *tempCS, m_pcRdCost->getLambda( true ) );
#else
@@ -2237,6 +2242,7 @@ bool EncCu::xCheckRDCostIntra(CodingStructure *&tempCS, CodingStructure *&bestCS
bestCS->tmpColorSpaceIntraCost[colorSpaceIdx] = tempCS->cost;
}
}
+
if( !sps.getUseLFNST() )
{
xCheckBestMode( tempCS, bestCS, partitioner, encTestMode );
@@ -2249,6 +2255,7 @@ bool EncCu::xCheckRDCostIntra(CodingStructure *&tempCS, CodingStructure *&bestCS
bestSelFlag [ trGrpIdx ] = true;
bestMtsFlag = mtsFlag;
bestLfnstIdx = lfnstIdx;
+
if( bestCS->cus.size() == 1 )
{
CodingUnit &cu = *bestCS->cus.front();
@@ -2261,7 +2268,7 @@ bool EncCu::xCheckRDCostIntra(CodingStructure *&tempCS, CodingStructure *&bestCS
}
}
}
-
+
//we decide to skip the non-DCT-II transforms and LFNST according to the ISP results
if ((endMtsFlag > 0 || endLfnstIdx > 0) && (cu.ispMode || (bestCS && bestCS->cus[0]->ispMode)) && tempCS->slice->isIntra() && m_pcEncCfg->getUseFastISP())
{
@@ -2269,7 +2276,7 @@ bool EncCu::xCheckRDCostIntra(CodingStructure *&tempCS, CodingStructure *&bestCS
double bestIspCost = m_modeCtrl->getIspCost();
CHECKD( bestCostDct2NoIsp <= bestIspCost, "wrong cost!" );
double threshold = 1.4;
-
+
double lfnstThreshold = 1.01 * threshold;
if( m_modeCtrl->getStopNonDCT2Transforms() || bestCostDct2NoIsp > bestIspCost*lfnstThreshold )
{
@@ -2294,6 +2301,16 @@ bool EncCu::xCheckRDCostIntra(CodingStructure *&tempCS, CodingStructure *&bestCS
break;
}
}
+#if JVET_W0103_INTRA_MTS
+ if (lfnstIdx && m_modeCtrl->getMtsFirstPassNoIspCost() != MAX_DOUBLE && isLuma(partitioner.chType))
+ {
+ double threshold = 1.5;
+ if (m_modeCtrl->getMtsFirstPassNoIspCost() > threshold * bestCS->cost)
+ {
+ endLfnstIdx = lfnstIdx;
+ }
+ }
+#endif
}
} //for emtCuFlag
diff --git a/source/Lib/EncoderLib/IntraSearch.cpp b/source/Lib/EncoderLib/IntraSearch.cpp
index aa634aa9dcd750952c764c140dd7b54ea5af8cd0..23c47b245bcc8b23b11a4dfb0283591275026a20 100644
--- a/source/Lib/EncoderLib/IntraSearch.cpp
+++ b/source/Lib/EncoderLib/IntraSearch.cpp
@@ -380,6 +380,45 @@ double IntraSearch::findInterCUCost( CodingUnit &cu )
return COST_UNKNOWN;
}
#endif
+#if JVET_W0103_INTRA_MTS
+bool IntraSearch::testISPforCurrCU(const CodingUnit &cu)
+{
+ CodingStructure &cs = *cu.cs;
+ auto &pu = *cu.firstPU;
+ const CompArea &area = pu.Y();
+ PelBuf piOrg = cs.getOrgBuf(area);
+
+ Pel* pOrg = piOrg.buf;
+ int uiWidth = area.width;
+ int uiHeight = area.height;
+ int iStride = piOrg.stride;
+ int Gsum = 0;
+ int nPix = (uiWidth - 2) * (uiHeight - 2);
+ for (int y = 1; y < (uiHeight - 1); y++)
+ {
+ for (int x = 1; x < (uiWidth - 1); x++)
+ {
+ const Pel *p = pOrg + y * iStride + x;
+
+ int iDy = p[-iStride - 1] + 2 * p[-1] + p[iStride - 1] - p[-iStride + 1] - 2 * p[+1] - p[iStride + 1];
+ int iDx = p[iStride - 1] + 2 * p[iStride] + p[iStride + 1] - p[-iStride - 1] - 2 * p[-iStride] - p[-iStride + 1];
+
+ if (iDy == 0 && iDx == 0)
+ continue;
+
+ int iAmp = (int)(abs(iDx) + abs(iDy));
+ Gsum += iAmp;
+ }
+ }
+ Gsum = (Gsum + (nPix >> 1)) / nPix;
+
+ bool testISP = true;
+ CHECK(m_numModesISPRDO != -1, "m_numModesISPRDO!=-1");
+
+ m_numModesISPRDO = (Gsum < 50 && uiWidth >= 16 && uiHeight >= 16) ? 1 : 2;
+ return testISP;
+}
+#endif
bool IntraSearch::estIntraPredLumaQT(CodingUnit &cu, Partitioner &partitioner, const double bestCostSoFar, bool mtsCheckRangeFlag, int mtsFirstCheckId, int mtsLastCheckId, bool moreProbMTSIdxFirst, CodingStructure* bestCS)
{
CodingStructure &cs = *cu.cs;
@@ -441,6 +480,18 @@ bool IntraSearch::estIntraPredLumaQT(CodingUnit &cu, Partitioner &partitioner, c
mtsUsageFlag = 0;
}
+#if JVET_W0103_INTRA_MTS
+ if (!cu.mtsFlag && !cu.lfnstIdx)
+ {
+ m_globalBestCostStore = MAX_DOUBLE;
+ m_globalBestCostValid = false;
+ if (bestCS->getCU(partitioner.chType) != NULL && bestCS->getCU(partitioner.chType)->predMode != MODE_INTRA && bestCostSoFar != MAX_DOUBLE)
+ {
+ m_globalBestCostStore = bestCostSoFar;
+ m_globalBestCostValid = true;
+ }
+ }
+#endif
const bool colorTransformIsEnabled = sps.getUseColorTrans() && !CS::isDualITree(cs);
const bool isFirstColorSpace = colorTransformIsEnabled && ((m_pcEncCfg->getRGBFormatFlag() && cu.colorTransform) || (!m_pcEncCfg->getRGBFormatFlag() && !cu.colorTransform));
const bool isSecondColorSpace = colorTransformIsEnabled && ((m_pcEncCfg->getRGBFormatFlag() && !cu.colorTransform) || (!m_pcEncCfg->getRGBFormatFlag() && cu.colorTransform));
@@ -449,7 +500,13 @@ bool IntraSearch::estIntraPredLumaQT(CodingUnit &cu, Partitioner &partitioner, c
bool ispCanBeUsed = sps.getUseISP() && cu.mtsFlag == 0 && cu.lfnstIdx == 0 && CU::canUseISP(width, height, cu.cs->sps->getMaxTbSize());
bool saveDataForISP = ispCanBeUsed && (!colorTransformIsEnabled || isFirstColorSpace);
bool testISP = ispCanBeUsed && (!colorTransformIsEnabled || !cu.colorTransform);
-
+#if JVET_W0103_INTRA_MTS
+ if (testISP && m_pcEncCfg->getUseFastISP())
+ {
+ m_numModesISPRDO = -1;
+ testISP &= testISPforCurrCU(cu);
+ }
+#endif
if ( saveDataForISP )
{
//reset the intra modes lists variables
@@ -1016,7 +1073,25 @@ bool IntraSearch::estIntraPredLumaQT(CodingUnit &cu, Partitioner &partitioner, c
if ((m_pcEncCfg->getUseFastLFNST() || !cu.slice->isIntra()) && m_bestModeCostValid[lfnstIdx])
{
numModesForFullRD = 0;
+#if JVET_W0103_INTRA_MTS
+ double thresholdSkipMode = 1.0 + ((cu.lfnstIdx > 0) ? 0.1 : 0.8) * (1.4 / sqrt((double)(width * height)));
+ std::vector<std::pair<ModeInfo, double>> ModeInfoWithDCT2Cost(m_savedNumRdModes[0]);
+ for (int i = 0; i < m_savedNumRdModes[0]; i++)
+ {
+ ModeInfoWithDCT2Cost[i] = { m_savedRdModeList[0][i], m_modeCostStore[0][i] };
+ }
+ std::stable_sort(ModeInfoWithDCT2Cost.begin(), ModeInfoWithDCT2Cost.end(), [](const std::pair<ModeInfo, double> & l, const std::pair<ModeInfo, double> & r) {return l.second < r.second; });
+ // **Reorder the modes** and Skip checking the modes with much larger R-D cost than the best mode
+ for (int i = 0; i < m_savedNumRdModes[0]; i++)
+ {
+ if (ModeInfoWithDCT2Cost[i].second <= thresholdSkipMode * ModeInfoWithDCT2Cost[0].second)
+ {
+ uiRdModeList.push_back(ModeInfoWithDCT2Cost[i].first);
+ numModesForFullRD++;
+ }
+ }
+#else
double thresholdSkipMode = 1.0 + ((cu.lfnstIdx > 0) ? 0.1 : 1.0) * (1.4 / sqrt((double) (width * height)));
// Skip checking the modes with much larger R-D cost than the best mode
@@ -1028,6 +1103,7 @@ bool IntraSearch::estIntraPredLumaQT(CodingUnit &cu, Partitioner &partitioner, c
numModesForFullRD++;
}
}
+#endif
}
else // this is necessary because we skip the candidates list calculation, since it was already obtained for
// the DCT-II. Now we load it
@@ -1060,7 +1136,6 @@ bool IntraSearch::estIntraPredLumaQT(CodingUnit &cu, Partitioner &partitioner, c
CHECK(numModesForFullRD != uiRdModeList.size(), "Inconsistent state!");
#endif
// after this point, don't use numModesForFullRD
-
// PBINTRA fast
if (m_pcEncCfg->getUsePbIntraFast() && !cs.slice->isIntra() && uiRdModeList.size() < numModesAvailable
&& !cs.slice->getDisableSATDForRD() && (mtsUsageFlag != 2 || lfnstIdx > 0))
@@ -1270,6 +1345,9 @@ bool IntraSearch::estIntraPredLumaQT(CodingUnit &cu, Partitioner &partitioner, c
#endif
#if JVET_V0130_INTRA_TMP
cu.tmpFlag = uiOrgMode.tmpFlag;
+#if JVET_W0103_INTRA_MTS
+ if (cu.tmpFlag && cu.mtsFlag) continue;
+#endif
#endif
cu.mipFlag = uiOrgMode.mipFlg;
pu.mipTransposedFlag = uiOrgMode.mipTrFlg;
@@ -1426,6 +1504,7 @@ bool IntraSearch::estIntraPredLumaQT(CodingUnit &cu, Partitioner &partitioner, c
sortRdModeListFirstColorSpace(uiOrgMode, csTemp->cost, cu.bdpcmMode, m_savedRdModeFirstColorSpace[m_savedRdModeIdx], m_savedRdCostFirstColorSpace[m_savedRdModeIdx], m_savedBDPCMModeFirstColorSpace[m_savedRdModeIdx], m_numSavedRdModeFirstColorSpace[m_savedRdModeIdx]);
}
}
+
// check r-d cost
if( csTemp->cost < csBest->cost )
{
@@ -1444,6 +1523,13 @@ bool IntraSearch::estIntraPredLumaQT(CodingUnit &cu, Partitioner &partitioner, c
m_bestModeCostStore[ lfnstIdx ] = csBest->cost; //cs.cost;
m_bestModeCostValid[ lfnstIdx ] = true;
}
+#if JVET_W0103_INTRA_MTS
+ if (sps.getUseLFNST() && m_globalBestCostStore > csBest->cost)
+ {
+ m_globalBestCostStore = csBest->cost;
+ m_globalBestCostValid = true;
+ }
+#endif
if( csBest->cost < bestCurrentCost )
{
bestCurrentCost = csBest->cost;
@@ -3411,7 +3497,98 @@ uint64_t IntraSearch::xGetIntraFracBitsQTChroma(TransformUnit& currTU, const Com
uint64_t fracBits = m_CABACEstimator->getEstFracBits();
return fracBits;
}
+#if JVET_W0103_INTRA_MTS
+void IntraSearch::xSelectAMTForFullRD(TransformUnit &tu)
+{
+ if (!tu.blocks[COMPONENT_Y].valid())
+ {
+ return;
+ }
+
+ if (!tu.cu->mtsFlag)
+ {
+ return;
+ }
+
+ CodingStructure &cs = *tu.cs;
+ m_pcRdCost->setChromaFormat(cs.sps->getChromaFormatIdc());
+
+ const CompArea &area = tu.blocks[COMPONENT_Y];
+
+ const ChannelType chType = toChannelType(COMPONENT_Y);
+
+
+ PelBuf piOrg = cs.getOrgBuf(area);
+ PelBuf piPred = cs.getPredBuf(area);
+ PelBuf piResi = cs.getResiBuf(area);
+
+ const PredictionUnit &pu = *cs.getPU(area.pos(), chType);
+
+ //===== init availability pattern =====
+
+ PelBuf sharedPredTS(m_pSharedPredTransformSkip[COMPONENT_Y], area);
+ initIntraPatternChType(*tu.cu, area);
+
+ //===== get prediction signal =====
+ if (PU::isMIP(pu, chType))
+ {
+ initIntraMip(pu, area);
+ predIntraMip(COMPONENT_Y, piPred, pu);
+ }
+ else
+ {
+ predIntraAng(COMPONENT_Y, piPred, pu);
+ }
+
+
+ // save prediction
+ sharedPredTS.copyFrom(piPred);
+
+ const Slice &slice = *cs.slice;
+ //===== get residual signal =====
+ piResi.copyFrom(piOrg);
+ if (slice.getLmcsEnabledFlag() && m_pcReshape->getCTUFlag())
+ {
+ piResi.rspSignal(m_pcReshape->getFwdLUT());
+ piResi.subtract(piPred);
+ }
+ else
+ {
+ piResi.subtract(piPred);
+ }
+ // do transform and calculate Coeff AbsSum for all MTS candidates
+ std::vector<std::pair<int, uint64_t>> CoeffAbsSum(4);
+
+ for (int i = 0; i < 4; i++)
+ {
+ tu.mtsIdx[0] = i + MTS_DST7_DST7;
+ uint64_t AbsSum = m_pcTrQuant->transformNxN(tu);
+ CoeffAbsSum[i] = { i, AbsSum };
+ }
+ std::stable_sort(CoeffAbsSum.begin(), CoeffAbsSum.end(), [](const std::pair<int, uint64_t> & l, const std::pair<int, uint64_t> & r) {return l.second < r.second; });
+
+ for (int i = 0; i < 4; i++)
+ {
+ m_TestAMTForFullRD[i] = CoeffAbsSum[i].first;
+ }
+ m_numCandAMTForFullRD = 4;
+
+ if (m_pcEncCfg->getUseFastLFNST())
+ {
+ double skipThreshold = 1.0 + 1.0 / sqrt((double)(area.width*area.height));
+ skipThreshold = std::max(skipThreshold, 1.03);
+ for (int i = 1; i < m_numCandAMTForFullRD; i++)
+ {
+ if (CoeffAbsSum[i].second > skipThreshold * CoeffAbsSum[0].second)
+ {
+ m_numCandAMTForFullRD = i;
+ break;
+ }
+ }
+ }
+}
+#endif
void IntraSearch::xIntraCodingTUBlock(TransformUnit &tu, const ComponentID &compID, Distortion& ruiDist, const int &default0Save1Load2, uint32_t* numSig, std::vector<TrMode>* trModes, const bool loadTr)
{
if (!tu.blocks[compID].valid())
@@ -3532,7 +3709,11 @@ void IntraSearch::xIntraCodingTUBlock(TransformUnit &tu, const ComponentID &comp
const Slice &slice = *cs.slice;
bool flag = slice.getLmcsEnabledFlag() && (slice.isIntra() || (!slice.isIntra() && m_pcReshape->getCTUFlag()));
+#if JVET_W0103_INTRA_MTS
+ if (!tu.cu->mtsFlag && isLuma(compID))
+#else
if (isLuma(compID))
+#endif
{
//===== get residual signal =====
piResi.copyFrom( piOrg );
@@ -4254,13 +4435,21 @@ bool IntraSearch::xRecurIntraCodingLumaQT( CodingStructure &cs, Partitioner &par
bool cbfBestMode = false;
bool cbfBestModeValid = false;
bool cbfDCT2 = true;
-
+#if JVET_W0103_INTRA_MTS
+ if (sps.getUseLFNST() && cu.mtsFlag) xSelectAMTForFullRD(tu);
+#endif
double bestDCT2cost = MAX_DOUBLE;
double threshold = m_pcEncCfg->getUseFastISP() && !cu.ispMode && ispIsCurrentWinner && nNumTransformCands > 1 ? 1 + 1.4 / sqrt( cu.lwidth() * cu.lheight() ) : 1;
for( int modeId = firstCheckId; modeId <= ( sps.getUseLFNST() ? lastCheckId : ( nNumTransformCands - 1 ) ); modeId++ )
{
uint8_t transformIndex = modeId;
-
+#if JVET_W0103_INTRA_MTS
+ if (sps.getUseLFNST() && cu.mtsFlag)
+ {
+ if (modeId >= m_numCandAMTForFullRD) continue;
+ transformIndex = m_TestAMTForFullRD[modeId];
+ }
+#endif
if( sps.getUseLFNST() )
{
if( ( transformIndex < lastCheckId ) || ( ( transformIndex == lastCheckId ) && !checkTransformSkip ) ) //we avoid this if the mode is transformSkip
@@ -4370,6 +4559,12 @@ bool IntraSearch::xRecurIntraCodingLumaQT( CodingStructure &cs, Partitioner &par
}
}
}
+#if JVET_W0103_INTRA_MTS
+ else if (cu.mtsFlag)
+ {
+ xIntraCodingTUBlock(tu, COMPONENT_Y, singleDistTmpLuma, 2, &numSig, nullptr, true);
+ }
+#endif
else
{
xIntraCodingTUBlock( tu, COMPONENT_Y, singleDistTmpLuma, default0Save1Load2, &numSig );
@@ -4443,7 +4638,21 @@ bool IntraSearch::xRecurIntraCodingLumaQT( CodingStructure &cs, Partitioner &par
{
bestDCT2cost = singleCostTmp;
}
-
+#if JVET_W0103_INTRA_MTS
+ if (sps.getUseLFNST() && cu.mtsFlag)
+ {
+ if (singleCostTmp != MAX_DOUBLE)
+ {
+ const CompArea& area = tu.blocks[COMPONENT_Y];
+ double skipThreshold = 1.0 + 1.0 / sqrt((double)(area.width*area.height));
+ skipThreshold = std::max(skipThreshold, !m_pcEncCfg->getUseFastLFNST()? 1.06: 1.03);
+ if (singleCostTmp > skipThreshold * m_globalBestCostStore)
+ {
+ m_numCandAMTForFullRD = modeId + 1;
+ }
+ }
+ }
+#endif
if (singleCostTmp < dSingleCost)
{
dSingleCost = singleCostTmp;
@@ -6512,13 +6721,18 @@ bool IntraSearch::xSortISPCandList(double bestCostSoFar, double bestNonISPCost,
ModeInfo(refMode.mipFlg, refMode.mipTrFlg, refMode.mRefId, refMode.ispMod, bestISPModeInRelCU));
modeIsInList[bestISPModeInRelCU] = true;
}
-
// Planar
+#if JVET_W0103_INTRA_MTS
+ // push planar later when FastISP is on.
+ if (!m_pcEncCfg->getUseFastISP() && !modeIsInList[mode1])
+#else
if (!modeIsInList[mode1])
+#endif
{
destListPtr->push_back(ModeInfo(refMode.mipFlg, refMode.mipTrFlg, refMode.mRefId, refMode.ispMod, mode1));
modeIsInList[mode1] = true;
}
+
// Best angle in regular intra
if (mode2 != -1 && !modeIsInList[mode2])
{
@@ -6543,7 +6757,14 @@ bool IntraSearch::xSortISPCandList(double bestCostSoFar, double bestNonISPCost,
}
}
}
-
+#if JVET_W0103_INTRA_MTS
+ // Planar (after angular modes when FastISP is on)
+ if (!modeIsInList[mode1])
+ {
+ destListPtr->push_back(ModeInfo(refMode.mipFlg, refMode.mipTrFlg, refMode.mRefId, refMode.ispMod, mode1));
+ modeIsInList[mode1] = true;
+ }
+#endif
// DC is added after the angles from regular intra
if (dcModeIndex != -1 && !modeIsInList[DC_IDX])
{
@@ -6555,6 +6776,12 @@ bool IntraSearch::xSortISPCandList(double bestCostSoFar, double bestNonISPCost,
for (int j = 0; j < NUM_LFNST_NUM_PER_SET; j++)
{
m_ispTestedModes[j].numOrigModesToTest = (int)destListPtr->size();
+#if JVET_W0103_INTRA_MTS
+ if (m_pcEncCfg->getUseFastISP() && m_numModesISPRDO != -1 && destListPtr->size() > m_numModesISPRDO)
+ {
+ m_ispTestedModes[j].numOrigModesToTest = m_numModesISPRDO;
+ }
+#endif
}
const int addedModesFromHadList = 3;
int newModesAdded = 0;
diff --git a/source/Lib/EncoderLib/IntraSearch.h b/source/Lib/EncoderLib/IntraSearch.h
index 31b813dabccf8007b977a5b4d48c8ed6f2e78a89..6afe99d63fa7dc88ec2be3722b77c29fa091e9ca 100644
--- a/source/Lib/EncoderLib/IntraSearch.h
+++ b/source/Lib/EncoderLib/IntraSearch.h
@@ -198,7 +198,10 @@ class IntraSearch : public IntraPrediction
private:
EncModeCtrl *m_modeCtrl;
Pel* m_pSharedPredTransformSkip[MAX_NUM_TBLOCKS];
-
+#if JVET_W0103_INTRA_MTS
+ int m_TestAMTForFullRD[4];
+ int m_numCandAMTForFullRD;
+#endif
XUCache m_unitCache;
CodingStructure ****m_pSplitCS;
@@ -371,7 +374,11 @@ private:
double m_modeCostStore[ NUM_LFNST_NUM_PER_SET ][ NUM_LUMA_MODE ]; // RD cost of each mode for each PU using DCT2
ModeInfo m_savedRdModeList[ NUM_LFNST_NUM_PER_SET ][ NUM_LUMA_MODE ];
int32_t m_savedNumRdModes[ NUM_LFNST_NUM_PER_SET ];
-
+#if JVET_W0103_INTRA_MTS
+ double m_globalBestCostStore;
+ bool m_globalBestCostValid;
+ int m_numModesISPRDO; //full modes for ISP testing.
+#endif
ModeInfo m_savedRdModeFirstColorSpace[4 * NUM_LFNST_NUM_PER_SET * 2][FAST_UDI_MAX_RDMODE_NUM];
char m_savedBDPCMModeFirstColorSpace[4 * NUM_LFNST_NUM_PER_SET * 2][FAST_UDI_MAX_RDMODE_NUM];
double m_savedRdCostFirstColorSpace[4 * NUM_LFNST_NUM_PER_SET * 2][FAST_UDI_MAX_RDMODE_NUM];
@@ -494,7 +501,10 @@ protected:
void xIntraCodingTUBlock (TransformUnit &tu, const ComponentID &compID, Distortion& ruiDist, const int &default0Save1Load2 = 0, uint32_t* numSig = nullptr, std::vector<TrMode>* trModes=nullptr, const bool loadTr=false );
void xIntraCodingACTTUBlock(TransformUnit &tu, const ComponentID &compID, Distortion& ruiDist, std::vector<TrMode>* trModes = nullptr, const bool loadTr = false);
-
+#if JVET_W0103_INTRA_MTS
+ void xSelectAMTForFullRD(TransformUnit &tu);
+ bool testISPforCurrCU(const CodingUnit &cu);
+#endif
ChromaCbfs xRecurIntraChromaCodingQT( CodingStructure &cs, Partitioner& pm, const double bestCostSoFar = MAX_DOUBLE, const PartSplit ispType = TU_NO_ISP );
bool xRecurIntraCodingLumaQT ( CodingStructure &cs, Partitioner& pm, const double bestCostSoFar = MAX_DOUBLE, const int subTuIdx = -1, const PartSplit ispType = TU_NO_ISP, const bool ispIsCurrentWinner = false, bool mtsCheckRangeFlag = false, int mtsFirstCheckId = 0, int mtsLastCheckId = 0, bool moreProbMTSIdxFirst = false );
bool xRecurIntraCodingACTQT(CodingStructure &cs, Partitioner& pm, bool mtsCheckRangeFlag = false, int mtsFirstCheckId = 0, int mtsLastCheckId = 0, bool moreProbMTSIdxFirst = false);