diff --git a/source/App/DecoderApp/decmain.cpp b/source/App/DecoderApp/decmain.cpp
index c8a6e3bd7070cdcbd867a95d3f39c27cca6739e1..0a664aa461019162caa0c5c7d1d09092bce5eb47 100644
--- a/source/App/DecoderApp/decmain.cpp
+++ b/source/App/DecoderApp/decmain.cpp
@@ -54,7 +54,7 @@ int main(int argc, char* argv[])
// print information
fprintf( stdout, "\n" );
- fprintf( stdout, "VVCSoftware: VTM Decoder Version %s ", VTM_VERSION );
+ fprintf( stdout, "VVCSoftware: ECM Decoder Version %s (VTM-%s) ", ECM_VERSION, VTM_VERSION );
fprintf( stdout, NVM_ONOS );
fprintf( stdout, NVM_COMPILEDBY );
fprintf( stdout, NVM_BITS );
diff --git a/source/App/EncoderApp/EncApp.cpp b/source/App/EncoderApp/EncApp.cpp
index 02bedf8fa1bb0f68541d7a1267d166976744de50..378561a7a65918bc4935c41dab9ab3711c3d642f 100644
--- a/source/App/EncoderApp/EncApp.cpp
+++ b/source/App/EncoderApp/EncApp.cpp
@@ -788,9 +788,9 @@ void EncApp::xInitLibCfg()
m_cEncLib.setUseWrapAround ( m_wrapAround );
m_cEncLib.setWrapAroundOffset ( m_wrapAroundOffset );
-#if IDCC_TPM_JEM
- m_cEncLib.setUseIntraTMP(m_IntraTMP);
- m_cEncLib.setIntraTMPMaxSize(m_IntraTMP_MaxSize);
+#if JVET_V0130_INTRA_TMP
+ m_cEncLib.setUseIntraTMP ( m_intraTMP );
+ m_cEncLib.setIntraTMPMaxSize ( m_intraTmpMaxSize );
#endif
#if JVET_V0094_BILATERAL_FILTER
m_cEncLib.setUseBIF ( m_BIF );
diff --git a/source/App/EncoderApp/EncAppCfg.cpp b/source/App/EncoderApp/EncAppCfg.cpp
index 21890e532d76457d352cbe6d6015d70421bf7758..33be266cc7f211f1f44df7dd5c67d63811482e8e 100644
--- a/source/App/EncoderApp/EncAppCfg.cpp
+++ b/source/App/EncoderApp/EncAppCfg.cpp
@@ -1044,9 +1044,9 @@ bool EncAppCfg::parseCfg( int argc, char* argv[] )
("AdditionalInterHypRefFrames", m_maxNumAddHypRefFrames, 4, "max. number of ref frames for additional inter hypotheseis")
("AdditionalInterHypTries", m_addHypTries, 1, "number of tries for additional inter prediction hypotheseis")
#endif
-#if IDCC_TPM_JEM
- ("IntraTMP", m_IntraTMP, false, "intra Template Matching (0: off, 1:on) [default: on]")
- ("IntraTMPMaxSize", m_IntraTMP_MaxSize, 64u, "intra Template Matching max CU size [default: 64]")
+#if JVET_V0130_INTRA_TMP
+ ("IntraTMP", m_intraTMP, false, "intra Template Matching (0: off, 1:on) [default: on]")
+ ("IntraTMPMaxSize", m_intraTmpMaxSize, 64u, "intra Template Matching max CU size [default: 64]")
#endif
#if JVET_V0094_BILATERAL_FILTER
("BIF", m_BIF, true, "bilateral filter (0: off, 1:on) [default: on]")
@@ -4176,12 +4176,12 @@ void EncAppCfg::xPrintParameter()
}
#endif
}
-#if IDCC_TPM_JEM
- msg(DETAILS, "Intra TMP: %d\n", m_IntraTMP);
- msg(DETAILS, "Max CU size of TMP: %d\n", m_IntraTMP_MaxSize);
+#if INTRA_TEMPLATE_MATCHING
+ msg(DETAILS, "Intra TMP: %d\n", m_intraTMP);
+ msg(DETAILS, "Max CU size of TMP: %d\n", m_intraTmpMaxSize);
msg(DETAILS, "dynamic search range with fixed comparison per pixel: \n");
- msg(DETAILS, " searchRangeWidth = %d*Width \n", IDCC_SearchRangeMultFactor);
- msg(DETAILS, " searchRangeHeight = %d*Heigh \n", IDCC_SearchRangeMultFactor);
+ msg(DETAILS, " searchRangeWidth = %d*Width \n", TMP_SEARCH_RANGE_MULT_FACTOR );
+ msg(DETAILS, " searchRangeHeight = %d*Heigh \n", TMP_SEARCH_RANGE_MULT_FACTOR );
#endif
msg( DETAILS, "Max Num Merge Candidates : %d\n", m_maxNumMergeCand );
@@ -4316,9 +4316,9 @@ void EncAppCfg::xPrintParameter()
{
msg( VERBOSE, "WrapAroundOffset:%d ", m_wrapAroundOffset );
}
-#if IDCC_TPM_JEM
- msg( VERBOSE, "IntraTMP:%d ", m_IntraTMP);
- msg( VERBOSE, "IntraTMP_MaxSize:%d ", m_IntraTMP_MaxSize);
+#if INTRA_TEMPLATE_MATCHING
+ msg( VERBOSE, "IntraTMP:%d ", m_intraTMP);
+ msg( VERBOSE, "IntraTmpMaxSize:%d ", m_intraTmpMaxSize);
#endif
#if JVET_V0094_BILATERAL_FILTER
msg( VERBOSE, "BIF:%d ", m_BIF);
diff --git a/source/App/EncoderApp/EncAppCfg.h b/source/App/EncoderApp/EncAppCfg.h
index a035b4d4c277f54a8360ef341a66000d81015815..83b5967d6cfb4d41bb9ed786457ce56db66ab722 100644
--- a/source/App/EncoderApp/EncAppCfg.h
+++ b/source/App/EncoderApp/EncAppCfg.h
@@ -411,9 +411,9 @@ protected:
int m_maxNumAddHypRefFrames; ///< max. number of ref frames for additional inter hypotheseis
int m_addHypTries; ///< max. number of tries for additional inter hypotheseis
#endif
-#if IDCC_TPM_JEM
- bool m_IntraTMP; ///< intra Template Matching
- unsigned m_IntraTMP_MaxSize; ///< max CU size for which intra TMP is allowed
+#if JVET_V0130_INTRA_TMP
+ bool m_intraTMP; ///< intra Template Matching
+ unsigned m_intraTmpMaxSize; ///< max CU size for which intra TMP is allowed
#endif
#if JVET_V0094_BILATERAL_FILTER
bool m_BIF; ///< bilateral filter
diff --git a/source/App/EncoderApp/encmain.cpp b/source/App/EncoderApp/encmain.cpp
index 0dfefebecaeb29ce3ebcea6a9e55c2bbe2672ab9..44a68dc8f4bbf7432d8622841ed5abd2db83a58b 100644
--- a/source/App/EncoderApp/encmain.cpp
+++ b/source/App/EncoderApp/encmain.cpp
@@ -85,7 +85,7 @@ int main(int argc, char* argv[])
{
// print information
fprintf( stdout, "\n" );
- fprintf( stdout, "VVCSoftware: VTM Encoder Version %s ", VTM_VERSION );
+ fprintf( stdout, "VVCSoftware: ECM Encoder Version %s (VTM-%s) ", ECM_VERSION, VTM_VERSION );
fprintf( stdout, NVM_ONOS );
fprintf( stdout, NVM_COMPILEDBY );
fprintf( stdout, NVM_BITS );
diff --git a/source/Lib/CommonLib/BilateralFilter.cpp b/source/Lib/CommonLib/BilateralFilter.cpp
index f5b7f0f3151c1666053530dbdf09ab11980649b2..38058a4da2d79cee6a2c2678f08320ee32b93a45 100755
--- a/source/Lib/CommonLib/BilateralFilter.cpp
+++ b/source/Lib/CommonLib/BilateralFilter.cpp
@@ -50,6 +50,13 @@
BilateralFilter::BilateralFilter()
{
+ m_bilateralFilterDiamond5x5 = blockBilateralFilterDiamond5x5;
+
+#if ENABLE_SIMD_BILATERAL_FILTER
+#ifdef TARGET_SIMD_X86
+ initBilateralFilterX86();
+#endif
+#endif
}
BilateralFilter::~BilateralFilter()
@@ -64,292 +71,193 @@ void BilateralFilter::destroy()
{
}
-void BilateralFilter::simdFilterDiamond5x5(uint32_t uiWidth, uint32_t uiHeight, int16_t block[], int16_t blkFilt[], const ClpRng& clpRng, Pel* recPtr, int recStride, int iWidthExtSIMD, int bfac, int bif_round_add, int bif_round_shift, bool isRDO)
+const char* BilateralFilter::getFilterLutParameters( const int size, const PredMode predMode, const int32_t qp, int& bfac )
{
- int pad = 2;
- int padwidth = iWidthExtSIMD;
-
- __m128i center, left, right, up, down, lu, ld, ru, rd, diffabs, four, fifteen, lut, acc, temp, round_add, clipmin, clipmax, inputVals;
- __m128i ll, rr, uu, dd;
-
- four = _mm_set1_epi16(4);
- fifteen = _mm_set1_epi16(15);
- round_add = _mm_set1_epi16(bif_round_add);
- clipmin = _mm_set1_epi16(clpRng.min);
- clipmax = _mm_set1_epi16(clpRng.max);
+ if( size <= 4 )
+ {
+ bfac = 3;
+ }
+ else if( size >= 16 )
+ {
+ bfac = 1;
+ }
+ else
+ {
+ bfac = 2;
+ }
- lut = _mm_loadu_si128((__m128i*)(LUTrowPtr));
- acc = _mm_set1_epi32(0);
-
- // Copy back parameters
- Pel *tempBlockPtr = (short*)blkFilt + (((padwidth+4) << 1) + 2);
- int tempBlockStride = padwidth+4;
-
-
- for (int col = 0; col < uiWidth; col += 8)
+ if( predMode == MODE_INTER )
{
- for (int row = 0; row < uiHeight; row++)
+ if( size <= 4 )
{
- acc = _mm_set1_epi32(0);
- int16_t *point = &block[(row + pad)*padwidth + pad + col];
-
- center = _mm_loadu_si128((__m128i*)(point));
-
- //load neighbours
- left = _mm_loadu_si128((__m128i*)(point - 1));
- right = _mm_loadu_si128((__m128i*)(point + 1));
- up = _mm_loadu_si128((__m128i*)(point - padwidth));
- down = _mm_loadu_si128((__m128i*)(point + padwidth));
-
- lu = _mm_loadu_si128((__m128i*)(point - 1 - padwidth));
- ld = _mm_loadu_si128((__m128i*)(point - 1 + padwidth));
- ru = _mm_loadu_si128((__m128i*)(point + 1 - padwidth));
- rd = _mm_loadu_si128((__m128i*)(point + 1 + padwidth));
-
- ll = _mm_loadu_si128((__m128i*)(point - 2));
- rr = _mm_loadu_si128((__m128i*)(point + 2));
- uu = _mm_loadu_si128((__m128i*)(point - 2*padwidth));
- dd = _mm_loadu_si128((__m128i*)(point + 2*padwidth));
-
- //calculate diffs
- left = _mm_sub_epi16(left, center);
- right = _mm_sub_epi16(right, center);
- up = _mm_sub_epi16(up, center);
- down = _mm_sub_epi16(down, center);
-
- lu = _mm_sub_epi16(lu, center);
- ld = _mm_sub_epi16(ld, center);
- ru = _mm_sub_epi16(ru, center);
- rd = _mm_sub_epi16(rd, center);
-
- ll = _mm_sub_epi16(ll, center);
- rr = _mm_sub_epi16(rr, center);
- uu = _mm_sub_epi16(uu, center);
- dd = _mm_sub_epi16(dd, center);
-
- //LEFT!
- //calculate abs
- diffabs = _mm_abs_epi16(left); //abs
- diffabs = _mm_add_epi16(diffabs, four); //+4
- diffabs = _mm_srai_epi16(diffabs, 3); //>>3
- diffabs = _mm_min_epi16(diffabs, fifteen); //min(x,15)
- diffabs = _mm_packus_epi16(diffabs, diffabs); //convert to 8
- diffabs = _mm_shuffle_epi8(lut, diffabs);//lut
- diffabs = _mm_cvtepi8_epi16(diffabs);//back to 16-bit
- diffabs = _mm_sign_epi16(diffabs, left);//fix sign!
- acc = _mm_add_epi16(diffabs, acc); //add to acc
- //RIGHT!
- //calculate abs
- diffabs = _mm_abs_epi16(right); //abs
- diffabs = _mm_add_epi16(diffabs, four); //+4
- diffabs = _mm_srai_epi16(diffabs, 3); //>>3
- diffabs = _mm_min_epi16(diffabs, fifteen); //min(x,15)
- diffabs = _mm_packus_epi16(diffabs, diffabs); //convert to 8
- diffabs = _mm_shuffle_epi8(lut, diffabs);//lut
- diffabs = _mm_cvtepi8_epi16(diffabs);//back to 16-bit
- diffabs = _mm_sign_epi16(diffabs, right);//fix sign!
- acc = _mm_add_epi16(diffabs, acc); //add to acc
- //UP!
- //calculate abs
- diffabs = _mm_abs_epi16(up); //abs
- diffabs = _mm_add_epi16(diffabs, four); //+4
- diffabs = _mm_srai_epi16(diffabs, 3); //>>3
- diffabs = _mm_min_epi16(diffabs, fifteen); //min(x,15)
- diffabs = _mm_packus_epi16(diffabs, diffabs); //convert to 8
- diffabs = _mm_shuffle_epi8(lut, diffabs);//lut
- diffabs = _mm_cvtepi8_epi16(diffabs);//back to 16-bit
- diffabs = _mm_sign_epi16(diffabs, up);//fix sign!
- acc = _mm_add_epi16(diffabs, acc); //add to acc
-
- //DOWN!
- //calculate abs
- diffabs = _mm_abs_epi16(down); //abs
- diffabs = _mm_add_epi16(diffabs, four); //+4
- diffabs = _mm_srai_epi16(diffabs, 3); //>>3
- diffabs = _mm_min_epi16(diffabs, fifteen); //min(x,15)
- diffabs = _mm_packus_epi16(diffabs, diffabs); //convert to 8
- diffabs = _mm_shuffle_epi8(lut, diffabs);//lut
- diffabs = _mm_cvtepi8_epi16(diffabs);//back to 16-bit
- diffabs = _mm_sign_epi16(diffabs, down);//fix sign!
- acc = _mm_add_epi16(diffabs, acc); //add to acc
-
- //lu!
- //calculate abs
- diffabs = _mm_abs_epi16(lu); //abs
- diffabs = _mm_add_epi16(diffabs, four); //+4
- diffabs = _mm_srai_epi16(diffabs, 3); //>>3
- diffabs = _mm_min_epi16(diffabs, fifteen); //min(x,15)
- diffabs = _mm_packus_epi16(diffabs, diffabs); //convert to 8
- diffabs = _mm_shuffle_epi8(lut, diffabs);//lut
- diffabs = _mm_cvtepi8_epi16(diffabs);//back to 16-bit
- diffabs = _mm_srai_epi16(diffabs, 1);//diagonal shift!
- diffabs = _mm_sign_epi16(diffabs, lu);//fix sign!
- acc = _mm_add_epi16(diffabs, acc); //add to acc
- //ld!
- //calculate abs
- diffabs = _mm_abs_epi16(ld); //abs
- diffabs = _mm_add_epi16(diffabs, four); //+4
- diffabs = _mm_srai_epi16(diffabs, 3); //>>3
- diffabs = _mm_min_epi16(diffabs, fifteen); //min(x,15)
- diffabs = _mm_packus_epi16(diffabs, diffabs); //convert to 8
- diffabs = _mm_shuffle_epi8(lut, diffabs);//lut
- diffabs = _mm_cvtepi8_epi16(diffabs);//back to 16-bit
- diffabs = _mm_srai_epi16(diffabs, 1);//diagonal shift!
- diffabs = _mm_sign_epi16(diffabs, ld);//fix sign!
- acc = _mm_add_epi16(diffabs, acc); //add to acc
- //ru!
- //calculate abs
- diffabs = _mm_abs_epi16(ru); //abs
- diffabs = _mm_add_epi16(diffabs, four); //+4
- diffabs = _mm_srai_epi16(diffabs, 3); //>>3
- diffabs = _mm_min_epi16(diffabs, fifteen); //min(x,15)
- diffabs = _mm_packus_epi16(diffabs, diffabs); //convert to 8
- diffabs = _mm_shuffle_epi8(lut, diffabs);//lut
- diffabs = _mm_cvtepi8_epi16(diffabs);//back to 16-bit
- diffabs = _mm_srai_epi16(diffabs, 1);//diagonal shift!
- diffabs = _mm_sign_epi16(diffabs, ru);//fix sign!
- acc = _mm_add_epi16(diffabs, acc); //add to acc
- //rd!
- //calculate abs
- diffabs = _mm_abs_epi16(rd); //abs
- diffabs = _mm_add_epi16(diffabs, four); //+4
- diffabs = _mm_srai_epi16(diffabs, 3); //>>3
- diffabs = _mm_min_epi16(diffabs, fifteen); //min(x,15)
- diffabs = _mm_packus_epi16(diffabs, diffabs); //convert to 8
- diffabs = _mm_shuffle_epi8(lut, diffabs);//lut
- diffabs = _mm_cvtepi8_epi16(diffabs);//back to 16-bit
- diffabs = _mm_srai_epi16(diffabs, 1);//diagonal shift!
- diffabs = _mm_sign_epi16(diffabs, rd);//fix sign!
- acc = _mm_add_epi16(diffabs, acc); //add to acc
-
- //ll!
- //calculate abs
- diffabs = _mm_abs_epi16(ll); //abs
- diffabs = _mm_add_epi16(diffabs, four); //+4
- diffabs = _mm_srai_epi16(diffabs, 3); //>>3
- diffabs = _mm_min_epi16(diffabs, fifteen); //min(x,15)
- diffabs = _mm_packus_epi16(diffabs, diffabs); //convert to 8
- diffabs = _mm_shuffle_epi8(lut, diffabs);//lut
- diffabs = _mm_cvtepi8_epi16(diffabs);//back to 16-bit
- diffabs = _mm_srai_epi16(diffabs, 1);//diagonal shift!
- diffabs = _mm_sign_epi16(diffabs, ll);//fix sign!
- acc = _mm_add_epi16(diffabs, acc); //add to acc
- //rr!
- //calculate abs
- diffabs = _mm_abs_epi16(rr); //abs
- diffabs = _mm_add_epi16(diffabs, four); //+4
- diffabs = _mm_srai_epi16(diffabs, 3); //>>3
- diffabs = _mm_min_epi16(diffabs, fifteen); //min(x,15)
- diffabs = _mm_packus_epi16(diffabs, diffabs); //convert to 8
- diffabs = _mm_shuffle_epi8(lut, diffabs);//lut
- diffabs = _mm_cvtepi8_epi16(diffabs);//back to 16-bit
- diffabs = _mm_srai_epi16(diffabs, 1);//diagonal shift!
- diffabs = _mm_sign_epi16(diffabs, rr);//fix sign!
- acc = _mm_add_epi16(diffabs, acc); //add to acc
- //uu!
- //calculate abs
- diffabs = _mm_abs_epi16(uu); //abs
- diffabs = _mm_add_epi16(diffabs, four); //+4
- diffabs = _mm_srai_epi16(diffabs, 3); //>>3
- diffabs = _mm_min_epi16(diffabs, fifteen); //min(x,15)
- diffabs = _mm_packus_epi16(diffabs, diffabs); //convert to 8
- diffabs = _mm_shuffle_epi8(lut, diffabs);//lut
- diffabs = _mm_cvtepi8_epi16(diffabs);//back to 16-bit
- diffabs = _mm_srai_epi16(diffabs, 1);//diagonal shift!
- diffabs = _mm_sign_epi16(diffabs, uu);//fix sign!
- acc = _mm_add_epi16(diffabs, acc); //add to acc
- //dd!
- //calculate abs
- diffabs = _mm_abs_epi16(dd); //abs
- diffabs = _mm_add_epi16(diffabs, four); //+4
- diffabs = _mm_srai_epi16(diffabs, 3); //>>3
- diffabs = _mm_min_epi16(diffabs, fifteen); //min(x,15)
- diffabs = _mm_packus_epi16(diffabs, diffabs); //convert to 8
- diffabs = _mm_shuffle_epi8(lut, diffabs);//lut
- diffabs = _mm_cvtepi8_epi16(diffabs);//back to 16-bit
- diffabs = _mm_srai_epi16(diffabs, 1);//diagonal shift!
- diffabs = _mm_sign_epi16(diffabs, dd);//fix sign!
- acc = _mm_add_epi16(diffabs, acc); //add to acc
-
- if (bfac == 2)
- {
- acc = _mm_slli_epi16(acc, 1); // Shift left to get 2*
- }
- else if (bfac == 3)
- {
- temp = _mm_slli_epi16(acc, 1); // Multiply by two by shifting left
- acc = _mm_add_epi16(acc, temp); // Add original value to get 3*
- }
-
- // Add 16 and shift 5
- acc = _mm_add_epi16(acc, round_add);
- acc = _mm_srai_epi16(acc, bif_round_shift);
-
- // Instead we add our input values to the delta
- if(isRDO)
- {
- acc = _mm_add_epi16(acc, center);
- }
- else
- {
- int16_t *recpoint = &recPtr[row * recStride + col];
- inputVals = _mm_loadu_si128((__m128i*)(recpoint));
- acc = _mm_add_epi16(acc, inputVals);
- }
-
- // Clip
- acc = _mm_max_epi16(acc, clipmin);
- acc = _mm_min_epi16(acc, clipmax);
-
- _mm_store_si128((__m128i*)(blkFilt + (row + pad) * (padwidth + 4) + col + pad), acc);
+ bfac = 2;
+ }
+ else if( size >= 16 )
+ {
+ bfac = 1;
+ }
+ else
+ {
+ bfac = 2;
}
}
-
- // Copy back from tempbufFilter to recBuf
- int onerow = uiWidth * sizeof(Pel);
- for(uint32_t yy = 0; yy < uiHeight; yy++)
+
+ int sqp = qp;
+
+ if( sqp < 17 )
+ {
+ sqp = 17;
+ }
+
+ if( sqp > 42 )
{
- std::memcpy(recPtr, tempBlockPtr, onerow);
- recPtr += recStride;
- tempBlockPtr += tempBlockStride;
+ sqp = 42;
}
+
+ return m_wBIF[sqp - 17];
}
-void BilateralFilter::blockBilateralFilterDiamond5x5(uint32_t uiWidth, uint32_t uiHeight, int16_t block[], int16_t blkFilt[], const ClpRng& clpRng, Pel* recPtr, int recStride, int iWidthExtSIMD, int bfac, int bif_round_add, int bif_round_shift, bool isRDO)
+void BilateralFilter::blockBilateralFilterDiamond5x5( uint32_t uiWidth, uint32_t uiHeight, int16_t block[], int16_t blkFilt[], const ClpRng& clpRng, Pel* recPtr, int recStride, int iWidthExtSIMD, int bfac, int bif_round_add, int bif_round_shift, bool isRDO, const char* LUTrowPtr )
{
int pad = 2;
-#ifdef TARGET_SIMD_X86
- if ((uiWidth >= 8) || (!isRDO && (uiWidth >= 4)))
+ int padwidth = iWidthExtSIMD;
+ int downbuffer[64];
+ int downleftbuffer[65];
+ int downrightbuffer[2][65];
+ int Shift, sg0, v0, idx, w0;
+ Shift = sizeof( int ) * 8 - 1;
+ downbuffer[0] = 0;
+
+ for( int x = 0; x < uiWidth; x++ )
{
- simdFilterDiamond5x5(uiWidth, uiHeight, block, blkFilt, clpRng, recPtr, recStride, iWidthExtSIMD, bfac, bif_round_add, bif_round_shift, isRDO);
+ int pixel = block[(-1 + pad)*padwidth + x + pad];
+ int below = block[(-1 + pad + 1)*padwidth + x + pad];
+ int diff = below - pixel;
+ sg0 = diff >> Shift;
+ v0 = (diff + sg0) ^ sg0;
+ v0 = (v0 + 4) >> 3;
+ idx = 15 + ((v0 - 15)&((v0 - 15) >> Shift));
+ w0 = LUTrowPtr[idx];
+ int mod = (w0 + sg0) ^ sg0;
+ downbuffer[x] = mod;
+
+ int belowright = block[(-1 + pad + 1)*padwidth + x + pad + 1];
+ diff = belowright - pixel;
+ sg0 = diff >> Shift;
+ v0 = (diff + sg0) ^ sg0;
+ v0 = (v0 + 4) >> 3;
+ idx = 15 + ((v0 - 15)&((v0 - 15) >> Shift));
+ w0 = LUTrowPtr[idx] >> 1;
+ mod = (w0 + sg0) ^ sg0;
+ downrightbuffer[1][x + 1] = mod;
+
+ int belowleft = block[(-1 + pad + 1)*padwidth + x + pad - 1];
+ diff = belowleft - pixel;
+ sg0 = diff >> Shift;
+ v0 = (diff + sg0) ^ sg0;
+ v0 = (v0 + 4) >> 3;
+ idx = 15 + ((v0 - 15)&((v0 - 15) >> Shift));
+ w0 = LUTrowPtr[idx] >> 1;
+ mod = (w0 + sg0) ^ sg0;
+ downleftbuffer[x] = mod;
}
- else
-#endif
+ int width = uiWidth;
+ for( int y = 0; y < uiHeight; y++ )
{
-
- int padwidth = iWidthExtSIMD;
- int downbuffer[64];
- int downleftbuffer[65];
- int downrightbuffer[2][65];
- int Shift, sg0, v0, idx, w0;
- Shift = sizeof(int) * 8 - 1;
- downbuffer[0] = 0;
-
- for (int x = 0; x < uiWidth; x++)
- {
- int pixel = block[(-1 + pad)*padwidth + x + pad];
- int below = block[(-1 + pad + 1)*padwidth + x + pad];
- int diff = below - pixel;
+ int diff;
+
+ int16_t *rowStart = &block[(y + pad)*padwidth + pad];
+
+ int pixel = rowStart[-1];
+
+ int right = rowStart[0];
+ diff = right - pixel;
+ sg0 = diff >> Shift;
+ v0 = (diff + sg0) ^ sg0;
+ v0 = (v0 + 4) >> 3;
+ idx = 15 + ((v0 - 15)&((v0 - 15) >> Shift));
+ w0 = LUTrowPtr[idx];
+ int mod = (w0 + sg0) ^ sg0;
+ int rightmod = mod;
+
+ pixel = rowStart[-padwidth - 1];
+ int belowright = right;
+ diff = belowright - pixel;
+ sg0 = diff >> Shift;
+ v0 = (diff + sg0) ^ sg0;
+ v0 = (v0 + 4) >> 3;
+ idx = 15 + ((v0 - 15)&((v0 - 15) >> Shift));
+ w0 = LUTrowPtr[idx] >> 1;
+ mod = (w0 + sg0) ^ sg0;
+ downrightbuffer[(y + 1) % 2][0] = mod;
+
+ pixel = rowStart[-padwidth + width];
+ int belowleft = rowStart[width - 1];
+ diff = belowleft - pixel;
+ sg0 = diff >> Shift;
+ v0 = (diff + sg0) ^ sg0;
+ v0 = (v0 + 4) >> 3;
+ idx = 15 + ((v0 - 15)&((v0 - 15) >> Shift));
+ w0 = LUTrowPtr[idx] >> 1;
+ mod = (w0 + sg0) ^ sg0;
+ downleftbuffer[width] = mod;
+
+ for( int x = 0; x < uiWidth; x++ )
+ {
+ pixel = rowStart[x];
+ int modsum = 0;
+
+ int abovemod = -downbuffer[x];
+ modsum += abovemod;
+
+ int leftmod = -rightmod;
+ modsum += leftmod;
+
+ right = rowStart[x + 1];
+ diff = right - pixel;
sg0 = diff >> Shift;
v0 = (diff + sg0) ^ sg0;
v0 = (v0 + 4) >> 3;
idx = 15 + ((v0 - 15)&((v0 - 15) >> Shift));
w0 = LUTrowPtr[idx];
- int mod = (w0 + sg0) ^ sg0;
+ mod = (w0 + sg0) ^ sg0;
+
+ modsum += mod;
+ rightmod = mod;
+
+ int below = rowStart[x + padwidth];
+ diff = below - pixel;
+ sg0 = diff >> Shift;
+ v0 = (diff + sg0) ^ sg0;
+ v0 = (v0 + 4) >> 3;
+ idx = 15 + ((v0 - 15)&((v0 - 15) >> Shift));
+ w0 = LUTrowPtr[idx];
+ mod = (w0 + sg0) ^ sg0;
+ modsum += mod;
downbuffer[x] = mod;
-
- int belowright = block[(-1 + pad + 1)*padwidth + x + pad + 1];
+
+ int aboverightmod = -downleftbuffer[x + 1];
+ // modsum += ((int16_t)((uint16_t)((aboverightmod) >> 1)));
+ modsum += aboverightmod;
+
+ int aboveleftmod = -downrightbuffer[(y + 1) % 2][x];
+ // modsum += ((int16_t)((uint16_t)((aboveleftmod) >> 1)));
+ modsum += aboveleftmod;
+
+ int belowleft = rowStart[x + padwidth - 1];
+ diff = belowleft - pixel;
+ sg0 = diff >> Shift;
+ v0 = (diff + sg0) ^ sg0;
+ v0 = (v0 + 4) >> 3;
+ idx = 15 + ((v0 - 15)&((v0 - 15) >> Shift));
+ w0 = LUTrowPtr[idx] >> 1;
+ mod = (w0 + sg0) ^ sg0;
+ // modsum += ((int16_t)((uint16_t)((mod) >> 1)));
+ modsum += mod;
+ downleftbuffer[x] = mod;
+
+ int belowright = rowStart[x + padwidth + 1];
diff = belowright - pixel;
sg0 = diff >> Shift;
v0 = (diff + sg0) ^ sg0;
@@ -357,207 +265,87 @@ void BilateralFilter::blockBilateralFilterDiamond5x5(uint32_t uiWidth, uint32_t
idx = 15 + ((v0 - 15)&((v0 - 15) >> Shift));
w0 = LUTrowPtr[idx] >> 1;
mod = (w0 + sg0) ^ sg0;
- downrightbuffer[1][x + 1] = mod;
-
- int belowleft = block[(-1 + pad + 1)*padwidth + x + pad - 1];
- diff = belowleft - pixel;
+ //modsum += ((int16_t)((uint16_t)((mod) >> 1)));
+ modsum += mod;
+ downrightbuffer[y % 2][x + 1] = mod;
+
+ // For samples two pixels out, we do not reuse previously calculated
+ // values even though that is possible. Doing so would likely increase
+ // speed when SIMD is turned off.
+
+ int above = rowStart[x - 2 * padwidth];
+ diff = above - pixel;
sg0 = diff >> Shift;
v0 = (diff + sg0) ^ sg0;
v0 = (v0 + 4) >> 3;
idx = 15 + ((v0 - 15)&((v0 - 15) >> Shift));
w0 = LUTrowPtr[idx] >> 1;
mod = (w0 + sg0) ^ sg0;
- downleftbuffer[x] = mod;
- }
- int width = uiWidth;
- for (int y = 0; y < uiHeight; y++)
- {
- int diff;
-
- int16_t *rowStart = &block[(y + pad)*padwidth + pad];
-
- int pixel = rowStart[-1];
-
- int right = rowStart[0];
- diff = right - pixel;
+ modsum += mod;
+
+ below = rowStart[x + 2 * padwidth];
+ diff = below - pixel;
sg0 = diff >> Shift;
v0 = (diff + sg0) ^ sg0;
v0 = (v0 + 4) >> 3;
idx = 15 + ((v0 - 15)&((v0 - 15) >> Shift));
- w0 = LUTrowPtr[idx];
- int mod = (w0 + sg0) ^ sg0;
- int rightmod = mod;
-
- pixel = rowStart[-padwidth - 1];
- int belowright = right;
- diff = belowright - pixel;
+ w0 = LUTrowPtr[idx] >> 1;
+ mod = (w0 + sg0) ^ sg0;
+ modsum += mod;
+
+ int left = rowStart[x - 2];
+ diff = left - pixel;
sg0 = diff >> Shift;
v0 = (diff + sg0) ^ sg0;
v0 = (v0 + 4) >> 3;
idx = 15 + ((v0 - 15)&((v0 - 15) >> Shift));
w0 = LUTrowPtr[idx] >> 1;
mod = (w0 + sg0) ^ sg0;
- downrightbuffer[(y + 1) % 2][0] = mod;
-
- pixel = rowStart[-padwidth + width];
- int belowleft = rowStart[width - 1];
- diff = belowleft - pixel;
+ modsum += mod;
+
+ right = rowStart[x + 2];
+ diff = right - pixel;
sg0 = diff >> Shift;
v0 = (diff + sg0) ^ sg0;
v0 = (v0 + 4) >> 3;
idx = 15 + ((v0 - 15)&((v0 - 15) >> Shift));
w0 = LUTrowPtr[idx] >> 1;
mod = (w0 + sg0) ^ sg0;
- downleftbuffer[width] = mod;
-
-
- for (int x = 0; x < uiWidth; x++)
- {
-
- pixel = rowStart[x];
-
- int modsum = 0;
-
-
- int abovemod = -downbuffer[x];
- modsum += abovemod;
-
- int leftmod = -rightmod;
- modsum += leftmod;
-
- right = rowStart[x + 1];
- diff = right - pixel;
- sg0 = diff >> Shift;
- v0 = (diff + sg0) ^ sg0;
- v0 = (v0 + 4) >> 3;
- idx = 15 + ((v0 - 15)&((v0 - 15) >> Shift));
- w0 = LUTrowPtr[idx];
- mod = (w0 + sg0) ^ sg0;
-
- modsum += mod;
- rightmod = mod;
-
- int below = rowStart[x + padwidth];
- diff = below - pixel;
- sg0 = diff >> Shift;
- v0 = (diff + sg0) ^ sg0;
- v0 = (v0 + 4) >> 3;
- idx = 15 + ((v0 - 15)&((v0 - 15) >> Shift));
- w0 = LUTrowPtr[idx];
- mod = (w0 + sg0) ^ sg0;
- modsum += mod;
- downbuffer[x] = mod;
-
- int aboverightmod = -downleftbuffer[x + 1];
- // modsum += ((int16_t)((uint16_t)((aboverightmod) >> 1)));
- modsum += aboverightmod;
-
- int aboveleftmod = -downrightbuffer[(y + 1) % 2][x];
- // modsum += ((int16_t)((uint16_t)((aboveleftmod) >> 1)));
- modsum += aboveleftmod;
-
- int belowleft = rowStart[x + padwidth - 1];
- diff = belowleft - pixel;
- sg0 = diff >> Shift;
- v0 = (diff + sg0) ^ sg0;
- v0 = (v0 + 4) >> 3;
- idx = 15 + ((v0 - 15)&((v0 - 15) >> Shift));
- w0 = LUTrowPtr[idx] >> 1;
- mod = (w0 + sg0) ^ sg0;
- // modsum += ((int16_t)((uint16_t)((mod) >> 1)));
- modsum += mod;
- downleftbuffer[x] = mod;
-
- int belowright = rowStart[x + padwidth + 1];
- diff = belowright - pixel;
- sg0 = diff >> Shift;
- v0 = (diff + sg0) ^ sg0;
- v0 = (v0 + 4) >> 3;
- idx = 15 + ((v0 - 15)&((v0 - 15) >> Shift));
- w0 = LUTrowPtr[idx] >> 1;
- mod = (w0 + sg0) ^ sg0;
- //modsum += ((int16_t)((uint16_t)((mod) >> 1)));
- modsum += mod;
- downrightbuffer[y % 2][x + 1] = mod;
-
- // For samples two pixels out, we do not reuse previously calculated
- // values even though that is possible. Doing so would likely increase
- // speed when SIMD is turned off.
-
- int above = rowStart[x - 2*padwidth];
- diff = above - pixel;
- sg0 = diff >> Shift;
- v0 = (diff + sg0) ^ sg0;
- v0 = (v0 + 4) >> 3;
- idx = 15 + ((v0 - 15)&((v0 - 15) >> Shift));
- w0 = LUTrowPtr[idx] >> 1;
- mod = (w0 + sg0) ^ sg0;
- modsum += mod;
-
- below = rowStart[x + 2*padwidth];
- diff = below - pixel;
- sg0 = diff >> Shift;
- v0 = (diff + sg0) ^ sg0;
- v0 = (v0 + 4) >> 3;
- idx = 15 + ((v0 - 15)&((v0 - 15) >> Shift));
- w0 = LUTrowPtr[idx] >> 1;
- mod = (w0 + sg0) ^ sg0;
- modsum += mod;
-
- int left = rowStart[x - 2];
- diff = left - pixel;
- sg0 = diff >> Shift;
- v0 = (diff + sg0) ^ sg0;
- v0 = (v0 + 4) >> 3;
- idx = 15 + ((v0 - 15)&((v0 - 15) >> Shift));
- w0 = LUTrowPtr[idx] >> 1;
- mod = (w0 + sg0) ^ sg0;
- modsum += mod;
-
- right = rowStart[x + 2];
- diff = right - pixel;
- sg0 = diff >> Shift;
- v0 = (diff + sg0) ^ sg0;
- v0 = (v0 + 4) >> 3;
- idx = 15 + ((v0 - 15)&((v0 - 15) >> Shift));
- w0 = LUTrowPtr[idx] >> 1;
- mod = (w0 + sg0) ^ sg0;
- modsum += mod;
-
- blkFilt[(y + pad)*(padwidth+4) + x + pad] = ((int16_t)((uint16_t)((modsum*bfac + bif_round_add) >> bif_round_shift)));
- }
+ modsum += mod;
+
+ blkFilt[(y + pad)*(padwidth + 4) + x + pad] = (( int16_t ) (( uint16_t ) ((modsum*bfac + bif_round_add) >> bif_round_shift)));
}
+ }
- // Copy back
- Pel *tempBlockPtr = (short*)blkFilt + (((padwidth+4) << 1) + 2);
- int tempBlockStride = padwidth+4;
- if(isRDO)
+ // Copy back
+ Pel *tempBlockPtr = ( short* ) blkFilt + (((padwidth + 4) << 1) + 2);
+ int tempBlockStride = padwidth + 4;
+ if( isRDO )
+ {
+ Pel *srcBlockPtr = ( short* ) block + (((padwidth) << 1) + 2);
+ int srcBlockStride = padwidth;
+ for( uint32_t yy = 0; yy < uiHeight; yy++ )
{
- Pel *srcBlockPtr = (short*)block + (((padwidth) << 1) + 2);
- int srcBlockStride = padwidth;
- for(uint32_t yy = 0; yy < uiHeight; yy++)
+ for( uint32_t xx = 0; xx < uiWidth; xx++ )
{
- for(uint32_t xx = 0; xx < uiWidth; xx++)
- {
- recPtr[xx] = ClipPel(srcBlockPtr[xx] + tempBlockPtr[xx], clpRng);
- }
- recPtr += recStride;
- tempBlockPtr += tempBlockStride;
- srcBlockPtr += srcBlockStride;
+ recPtr[xx] = ClipPel( srcBlockPtr[xx] + tempBlockPtr[xx], clpRng );
}
+ recPtr += recStride;
+ tempBlockPtr += tempBlockStride;
+ srcBlockPtr += srcBlockStride;
}
- else
+ }
+ else
+ {
+ for( uint32_t yy = 0; yy < uiHeight; yy++ )
{
- for(uint32_t yy = 0; yy < uiHeight; yy++)
+ for( uint32_t xx = 0; xx < uiWidth; xx++ )
{
- for(uint32_t xx = 0; xx < uiWidth; xx++)
- {
- // new result = old result (which is SAO-treated already) + diff due to bilateral filtering
- recPtr[xx] = ClipPel<int>(recPtr[xx] + tempBlockPtr[xx], clpRng);
- }
- recPtr += recStride;
- tempBlockPtr += tempBlockStride;
+ // new result = old result (which is SAO-treated already) + diff due to bilateral filtering
+ recPtr[xx] = ClipPel<int>( recPtr[xx] + tempBlockPtr[xx], clpRng );
}
+ recPtr += recStride;
+ tempBlockPtr += tempBlockStride;
}
}
}
@@ -568,34 +356,10 @@ void BilateralFilter::bilateralFilterRDOdiamond5x5(PelBuf& resiBuf, const CPelBu
const unsigned uiHeight = predBuf.height;
int bfac = 1;
-
- int size = std::min(uiWidth, uiHeight);
- if (size <= 4)
- bfac = 3;
- else if (size >= 16)
- bfac = 1;
- else
- bfac = 2;
- if (currTU.cu->predMode == MODE_INTER)
- {
- if (size <= 4)
- bfac = 2;
- else if (size >= 16)
- bfac = 1;
- else
- bfac = 2;
- }
-
- qp = qp + currTU.cs->pps->getBIFQPOffset();
int bif_round_add = (BIF_ROUND_ADD) >> (currTU.cs->pps->getBIFStrength());
- int bif_round_shift = (BIF_ROUND_SHIFT) - (currTU.cs->pps->getBIFStrength());
-
- int sqp = qp;
- if(sqp<17)
- sqp = 17;
- if(sqp>42)
- sqp = 42;
- LUTrowPtr = wBIF[sqp-17];
+ int bif_round_shift = ( BIF_ROUND_SHIFT ) -(currTU.cs->pps->getBIFStrength());
+
+ const char* LUTrowPtr = getFilterLutParameters( std::min( uiWidth, uiHeight ), currTU.cu->predMode, qp + currTU.cs->pps->getBIFQPOffset(), bfac );
const unsigned uiPredStride = predBuf.stride;
const unsigned uiStrideRes = resiBuf.stride;
@@ -612,7 +376,6 @@ void BilateralFilter::bilateralFilterRDOdiamond5x5(PelBuf& resiBuf, const CPelBu
Pel *piRecoTemp = piReco;
// Reco = Pred + Resi
-
Pel *tempBlockPtr;
uint32_t uiWidthExt = uiWidth + (NUMBER_PADDED_SAMPLES << 1);
@@ -702,9 +465,13 @@ void BilateralFilter::bilateralFilterRDOdiamond5x5(PelBuf& resiBuf, const CPelBu
{
// copy 4 pixels one line above block from block to blockx + 3
std::copy(piRecIPred - (uiRecIPredStride)+blockx, piRecIPred - (uiRecIPredStride)+blockx + 1, tempblock + 2 + uiWidthExt + blockx);
- if(doReshape)
- for(int xx = 0; xx < 1; xx++)
- tempblock[2+uiWidthExt+blockx+xx] = pLUT[tempblock[2+uiWidthExt+blockx+xx]];
+ if( doReshape )
+ {
+ for( int xx = 0; xx < 1; xx++ )
+ {
+ tempblock[2 + uiWidthExt + blockx + xx] = pLUT[tempblock[2 + uiWidthExt + blockx + xx]];
+ }
+ }
}
}
else if (subTuHor)
@@ -716,13 +483,23 @@ void BilateralFilter::bilateralFilterRDOdiamond5x5(PelBuf& resiBuf, const CPelBu
const Pel *earlierPel = earlierHalfBuf.buf + (currTU.prev->lheight() - 1)*earlierStride;
std::copy(earlierPel, earlierPel + area.width, tempblock + 2 + uiWidthExt);
- if(doReshape)
- for(int xx = 0; xx < area.width; xx++)
- tempblock[2+uiWidthExt+xx] = pLUT[tempblock[2+uiWidthExt+xx]];
+ if( doReshape )
+ {
+ for( int xx = 0; xx < area.width; xx++ )
+ {
+ tempblock[2 + uiWidthExt + xx] = pLUT[tempblock[2 + uiWidthExt + xx]];
+ }
+ }
+
std::copy(earlierPel - earlierStride, earlierPel - earlierStride + area.width, tempblock + 2);
- if(doReshape)
- for(int xx = 0; xx < area.width; xx++)
- tempblock[2+xx] = pLUT[tempblock[2+xx]];
+
+ if( doReshape )
+ {
+ for( int xx = 0; xx < area.width; xx++ )
+ {
+ tempblock[2 + xx] = pLUT[tempblock[2 + xx]];
+ }
+ }
}
// left column
if (leftAvailable)
@@ -767,7 +544,7 @@ void BilateralFilter::bilateralFilterRDOdiamond5x5(PelBuf& resiBuf, const CPelBu
std::copy(tempblock + uiWidthExt, tempblock + uiWidthExt + uiWidthExt, tempblock);
std::copy(tempblock + uiWidthExt*(uiHeightExt-2), tempblock + uiWidthExt*(uiHeightExt-2) + uiWidthExt, tempblock + uiWidthExt*(uiHeightExt-1));
- blockBilateralFilterDiamond5x5(uiWidth, uiHeight, tempblock, tempblockFiltered, clpRng, piReco, uiRecStride, uiWidth + 4, bfac, bif_round_add, bif_round_shift, true);
+ m_bilateralFilterDiamond5x5(uiWidth, uiHeight, tempblock, tempblockFiltered, clpRng, piReco, uiRecStride, uiWidth + 4, bfac, bif_round_add, bif_round_shift, true, LUTrowPtr );
if (!useReco)
{
@@ -802,36 +579,9 @@ void BilateralFilter::bilateralFilterDiamond5x5(const CPelUnitBuf& src, PelUnitB
int recStride = rec.get(COMPONENT_Y).stride;
Pel *recPtr = rec.get(COMPONENT_Y).bufAt(compArea);
-
- int size = std::min(uiWidth, uiHeight);
- int bfac = 1;
-
- if (size <= 4)
- bfac = 3;
- else if (size >= 16)
- bfac = 1;
- else
- bfac = 2;
- if (currTU.cu->predMode == MODE_INTER)
- {
- if (size <= 4)
- bfac = 2;
- else if (size >= 16)
- bfac = 1;
- else
- bfac = 2;
- }
-
- // Offset qp before deciding on LUT:
- qp = qp + currTU.cs->pps->getBIFQPOffset();
-
- int sqp = qp;
- if(sqp<17)
- sqp = 17;
- if(sqp>42)
- sqp = 42;
- LUTrowPtr = wBIF[sqp-17];
+ int bfac = 1;
+ const char* LUTrowPtr = getFilterLutParameters( std::min( uiWidth, uiHeight ), currTU.cu->predMode, qp + currTU.cs->pps->getBIFQPOffset(), bfac );
int bif_round_add = (BIF_ROUND_ADD) >> (currTU.cs->pps->getBIFStrength());
int bif_round_shift = (BIF_ROUND_SHIFT) - (currTU.cs->pps->getBIFStrength());
@@ -846,8 +596,10 @@ void BilateralFilter::bilateralFilterDiamond5x5(const CPelUnitBuf& src, PelUnitB
uint32_t uiHeightExt = uiHeight + (NUMBER_PADDED_SAMPLES << 1);
int iWidthExtSIMD = uiWidthExt;
- if(uiWidth < 8)
+ if( uiWidth < 8 )
+ {
iWidthExtSIMD = 8 + (NUMBER_PADDED_SAMPLES << 1);
+ }
Pel *tempBlockPtr;
@@ -902,7 +654,7 @@ void BilateralFilter::bilateralFilterDiamond5x5(const CPelUnitBuf& src, PelUnitB
{
std::memcpy(tempBlockPtr, srcPtr, (uiWidthExt) * sizeof(Pel));
}
- return blockBilateralFilterDiamond5x5(uiWidth, uiHeight, tempblock, tempblockFiltered, clpRng, recPtr, recStride, iWidthExtSIMD, bfac, bif_round_add, bif_round_shift, false);
+ return m_bilateralFilterDiamond5x5(uiWidth, uiHeight, tempblock, tempblockFiltered, clpRng, recPtr, recStride, iWidthExtSIMD, bfac, bif_round_add, bif_round_shift, false, LUTrowPtr );
}
else
{
@@ -1034,7 +786,7 @@ void BilateralFilter::bilateralFilterDiamond5x5(const CPelUnitBuf& src, PelUnitB
}
}
- blockBilateralFilterDiamond5x5(uiWidth, uiHeight, tempblock, tempblockFiltered, clpRng, recPtr, recStride, iWidthExtSIMD, bfac, bif_round_add, bif_round_shift, false);
+ m_bilateralFilterDiamond5x5(uiWidth, uiHeight, tempblock, tempblockFiltered, clpRng, recPtr, recStride, iWidthExtSIMD, bfac, bif_round_add, bif_round_shift, false, LUTrowPtr );
}
void BilateralFilter::clipNotBilaterallyFilteredBlocks(const CPelUnitBuf& src, PelUnitBuf& rec, const ClpRng& clpRng, TransformUnit & currTU)
@@ -1225,10 +977,14 @@ void BilateralFilter::bilateralFilterPicRDOperCTU(CodingStructure& cs, PelUnitBu
rec.copyFrom(src);
}
- if (bifParams.frmOn == 0)
- std::fill(bifParams.ctuOn.begin(), bifParams.ctuOn.end(), 0);
- else if (bifParams.allCtuOn)
- std::fill(bifParams.ctuOn.begin(), bifParams.ctuOn.end(), 1);
+ if( bifParams.frmOn == 0 )
+ {
+ std::fill( bifParams.ctuOn.begin(), bifParams.ctuOn.end(), 0 );
+ }
+ else if( bifParams.allCtuOn )
+ {
+ std::fill( bifParams.ctuOn.begin(), bifParams.ctuOn.end(), 1 );
+ }
}
#endif
diff --git a/source/Lib/CommonLib/BilateralFilter.h b/source/Lib/CommonLib/BilateralFilter.h
index 0504013e76fec51da5db3ac5c9ff019188456647..4cbd41be7cb19150bafba652de4e71334c4ec209 100755
--- a/source/Lib/CommonLib/BilateralFilter.h
+++ b/source/Lib/CommonLib/BilateralFilter.h
@@ -64,12 +64,10 @@ private:
// = 2313 128-bit words which has been rounded up to 2320 above.
short *tempblockFiltered = &tempblockFilteredTemp[-2];
- void blockBilateralFilterDiamond5x5(uint32_t uiWidth, uint32_t uiHeight, int16_t block[], int16_t blkFilt[], const ClpRng& clpRng, Pel* recPtr, int recStride, int iWidthExtSIMD, int bfac, int bif_round_add, int bif_round_shift, bool isRDO);
- void simdFilterDiamond5x5(uint32_t uiWidth, uint32_t uiHeight, int16_t block[], int16_t blkFilt[], const ClpRng& clpRng, Pel* recPtr, int recStride, int iWidthExtSIMD, int bfac, int bif_round_add, int bif_round_shift, bool isRDO);
-
- char *LUTrowPtr;
+ void (*m_bilateralFilterDiamond5x5)( uint32_t uiWidth, uint32_t uiHeight, int16_t block[], int16_t blkFilt[], const ClpRng& clpRng, Pel* recPtr, int recStride, int iWidthExtSIMD, int bfac, int bif_round_add, int bif_round_shift, bool isRDO, const char* LUTrowPtr);
+ static void blockBilateralFilterDiamond5x5(uint32_t uiWidth, uint32_t uiHeight, int16_t block[], int16_t blkFilt[], const ClpRng& clpRng, Pel* recPtr, int recStride, int iWidthExtSIMD, int bfac, int bif_round_add, int bif_round_shift, bool isRDO, const char* LUTrowPtr );
- char wBIF[26][16] = {
+ char m_wBIF[26][16] = {
{ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, },
{ 0, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, },
{ 0, 2, 2, 2, 1, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, },
@@ -112,6 +110,19 @@ public:
void bilateralFilterDiamond5x5(const CPelUnitBuf& src, PelUnitBuf& rec, int32_t qp, const ClpRng& clpRng, TransformUnit & currTU);
void clipNotBilaterallyFilteredBlocks(const CPelUnitBuf& src, PelUnitBuf& rec, const ClpRng& clpRng, TransformUnit & currTU);
+ const char* getFilterLutParameters( const int size, const PredMode predMode, const int qp, int& bfac );
+
+#if ENABLE_SIMD_BILATERAL_FILTER
+#ifdef TARGET_SIMD_X86
+ template<X86_VEXT vext>
+ static void simdFilterDiamond5x5( uint32_t uiWidth, uint32_t uiHeight, int16_t block[], int16_t blkFilt[], const ClpRng& clpRng, Pel* recPtr, int recStride, int iWidthExtSIMD, int bfac, int bif_round_add, int bif_round_shift, bool isRDO, const char* LUTrowPtr );
+
+ void initBilateralFilterX86();
+ template <X86_VEXT vext>
+ void _initBilateralFilterX86();
+#endif
+#endif
+
};
#endif
diff --git a/source/Lib/CommonLib/CommonDef.h b/source/Lib/CommonLib/CommonDef.h
index 7210ce71c7abed003447fe9083e549086a3dfcc7..7a26296aafdd575c98bcc7e5a92fffb67e651a08 100644
--- a/source/Lib/CommonLib/CommonDef.h
+++ b/source/Lib/CommonLib/CommonDef.h
@@ -477,11 +477,6 @@ static const int ALF_VB_POS_ABOVE_CTUROW_CHMA = 2;
static const int MAX_ENCODER_DEBLOCKING_QUALITY_LAYERS = 8 ;
#endif
-#if IDCC_TPM_JEM
-static const int USE_MORE_BLOCKSIZE_DEPTH_MAX = IDCC_TMP_MaxSize_Depth - 1;
-static const int INIT_THRESHOULD_SHIFTBITS = 2; ///< (default 2) Early skip threshold for checking distance.
-#endif
-
#if SHARP_LUMA_DELTA_QP
static const uint32_t LUMA_LEVEL_TO_DQP_LUT_MAXSIZE = 1024; ///< max LUT size for QP offset based on luma
@@ -955,4 +950,12 @@ static const int MAX_FILTER_LENGTH_FIXED = 13;
static const int FIX_FILTER_NUM_COEFF = 42;
#endif
+#if JVET_V0130_INTRA_TMP
+static const int TMP_TEMPLATE_SIZE = 4; // must be multiple of 4 for SIMD
+static const int TMP_MAXSIZE_DEPTH = 6; // should be log2(TMP_TEMPLATE_SIZE): keep as 6 to avoid any error
+static const int USE_MORE_BLOCKSIZE_DEPTH_MAX = TMP_MAXSIZE_DEPTH - 1;
+static const int INIT_THRESHOULD_SHIFTBITS = 2; ///< (default 2) Early skip threshold for checking distance.
+static const int TMP_SEARCH_RANGE_MULT_FACTOR = 5;
+#endif
+
#endif // end of #ifndef __COMMONDEF__
diff --git a/source/Lib/CommonLib/ContextModelling.cpp b/source/Lib/CommonLib/ContextModelling.cpp
index 7855d636f03d74734a51f7f92acb5460e63cabfc..895e739d4da85b93e85c63642f9cebff728cca06 100644
--- a/source/Lib/CommonLib/ContextModelling.cpp
+++ b/source/Lib/CommonLib/ContextModelling.cpp
@@ -778,17 +778,17 @@ void MergeCtx::setMmvdMergeCandiInfo(PredictionUnit& pu, int candIdx)
PU::restrictBiPredMergeCandsOne(pu);
}
-#if IDCC_TPM_JEM
+#if JVET_V0130_INTRA_TMP
unsigned DeriveCtx::CtxTmpFlag(const CodingUnit& cu)
{
const CodingStructure* cs = cu.cs;
unsigned ctxId = 0;
const CodingUnit* cuLeft = cs->getCURestricted(cu.lumaPos().offset(-1, 0), cu, CH_L);
- ctxId = (cuLeft && cuLeft->TmpFlag) ? 1 : 0;
+ ctxId = (cuLeft && cuLeft->tmpFlag) ? 1 : 0;
const CodingUnit* cuAbove = cs->getCURestricted(cu.lumaPos().offset(0, -1), cu, CH_L);
- ctxId += (cuAbove && cuAbove->TmpFlag) ? 1 : 0;
+ ctxId += (cuAbove && cuAbove->tmpFlag) ? 1 : 0;
ctxId = (cu.lwidth() > 2 * cu.lheight() || cu.lheight() > 2 * cu.lwidth()) ? 3 : ctxId;
diff --git a/source/Lib/CommonLib/ContextModelling.h b/source/Lib/CommonLib/ContextModelling.h
index 6e3ad4616ecfce51ec1eacf7418eb2608fce7406..834e03af34a3c377352cdbcc5d6afa0f90870959 100644
--- a/source/Lib/CommonLib/ContextModelling.h
+++ b/source/Lib/CommonLib/ContextModelling.h
@@ -611,7 +611,7 @@ unsigned CtxAffineFlag( const CodingUnit& cu );
unsigned CtxPredModeFlag( const CodingUnit& cu );
unsigned CtxIBCFlag(const CodingUnit& cu);
unsigned CtxMipFlag ( const CodingUnit& cu );
-#if IDCC_TPM_JEM
+#if JVET_V0130_INTRA_TMP
unsigned CtxTmpFlag(const CodingUnit& cu);
#endif
unsigned CtxPltCopyFlag( const unsigned prevRunType, const unsigned dist );
diff --git a/source/Lib/CommonLib/Contexts.cpp b/source/Lib/CommonLib/Contexts.cpp
index 5b61e1b97c98302457599341a7218b96a4771802..dfce0eb5fa9ac8c61ce60203397cadb392f9f0dd 100644
--- a/source/Lib/CommonLib/Contexts.cpp
+++ b/source/Lib/CommonLib/Contexts.cpp
@@ -1000,15 +1000,15 @@ const CtxSet ContextSetCfg::MipFlag = ContextSetCfg::addCtxSet
{ 9, 9, 8, 6 },
{ 10, 10, 9, 6 }
});
-#if IDCC_TPM_JEM
+#if JVET_V0130_INTRA_TMP
const CtxSet ContextSetCfg::TmpFlag = ContextSetCfg::addCtxSet
({
{ CNU, CNU, CNU, CNU, },
{ CNU, CNU, CNU, CNU, },
{ CNU, CNU, CNU, CNU, },
- { DWS, DWS, DWS, DWS, },
- { DWS, DWS, DWS, DWS, },
- { DWS, DWS, DWS, DWS, },
+ { DWS, DWS, DWS, DWS, },
+ { DWS, DWS, DWS, DWS, },
+ { DWS, DWS, DWS, DWS, },
});
#endif
@@ -2089,13 +2089,13 @@ const CtxSet ContextSetCfg::MipFlag = ContextSetCfg::addCtxSet
{ 33, 49, 50, 25, },
{ 9, 10, 9, 6, },
});
-#if IDCC_TPM_JEM
+#if JVET_V0130_INTRA_TMP
const CtxSet ContextSetCfg::TmpFlag = ContextSetCfg::addCtxSet
({
{ CNU, CNU, CNU, CNU, },
{ CNU, CNU, CNU, CNU, },
{ CNU, CNU, CNU, CNU, },
- { DWS, DWS, DWS, DWS, },
+ { DWS, DWS, DWS, DWS, },
});
#endif
diff --git a/source/Lib/CommonLib/Contexts.h b/source/Lib/CommonLib/Contexts.h
index 0506ecf1983adfc686a63adacecea0f47646e06f..8993e11bbff8dccf4930e823bf6bb63ff935030c 100644
--- a/source/Lib/CommonLib/Contexts.h
+++ b/source/Lib/CommonLib/Contexts.h
@@ -245,7 +245,7 @@ public:
static const CtxSet CclmModeIdx;
static const CtxSet IntraChromaPredMode;
static const CtxSet MipFlag;
-#if IDCC_TPM_JEM
+#if JVET_V0130_INTRA_TMP
static const CtxSet TmpFlag;
#endif
#if MMLM
diff --git a/source/Lib/CommonLib/IntraPrediction.cpp b/source/Lib/CommonLib/IntraPrediction.cpp
index 4df14069b127922b9e232a2631d6b7e8356ddf01..bc5c898ed0d09a8e270f0db2f8d5176041b8e4dc 100644
--- a/source/Lib/CommonLib/IntraPrediction.cpp
+++ b/source/Lib/CommonLib/IntraPrediction.cpp
@@ -686,7 +686,7 @@ void IntraPrediction::initPredIntraParams(const PredictionUnit & pu, const CompA
if( sps.getSpsRangeExtension().getIntraSmoothingDisabledFlag()
|| !isLuma( chType )
|| useISP
-#if IDCC_TPM_JEM
+#if JVET_V0130_INTRA_TMP
|| PU::isTmp(pu, chType)
#endif
|| PU::isMIP( pu, chType )
@@ -1394,7 +1394,7 @@ void IntraPrediction::initIntraPatternChTypeISP(const CodingUnit& cu, const Comp
}
}
-#if IDCC_TPM_JEM
+#if JVET_V0130_INTRA_TMP
bool IntraPrediction::isRefTemplateAvailable(CodingUnit& cu, CompArea& area)
{
const ChannelType chType = toChannelType(area.compID);
@@ -1402,14 +1402,12 @@ bool IntraPrediction::isRefTemplateAvailable(CodingUnit& cu, CompArea& area)
const SPS& sps = *cs.sps;
const PreCalcValues& pcv = *cs.pcv;
-
const int tuWidth = area.width;
const int tuHeight = area.height;
const int predSize = m_topRefLength;
const int predHSize = m_leftRefLength;
//const int predStride = predSize;
-
const int unitWidth = pcv.minCUWidth >> getComponentScaleX(area.compID, sps.getChromaFormatIdc());
const int unitHeight = pcv.minCUHeight >> getComponentScaleY(area.compID, sps.getChromaFormatIdc());
@@ -1421,8 +1419,10 @@ bool IntraPrediction::isRefTemplateAvailable(CodingUnit& cu, CompArea& area)
const int numAboveRightUnits = totalAboveUnits - numAboveUnits;
const int numLeftBelowUnits = totalLeftUnits - numLeftUnits;
- if (numAboveUnits <= 0 || numLeftUnits <= 0 || numAboveRightUnits <= 0 || numLeftBelowUnits <= 0)
- return false;
+ if( numAboveUnits <= 0 || numLeftUnits <= 0 || numAboveRightUnits <= 0 || numLeftBelowUnits <= 0 )
+ {
+ return false;
+ }
// ----- Step 1: analyze neighborhood -----
const Position posLT = area;
diff --git a/source/Lib/CommonLib/IntraPrediction.h b/source/Lib/CommonLib/IntraPrediction.h
index 5d7ae3173634dd702b00ccf6248c3021e7e5f51e..95a98729e8b67700dfec4de8d5f5f4ad4614de51 100644
--- a/source/Lib/CommonLib/IntraPrediction.h
+++ b/source/Lib/CommonLib/IntraPrediction.h
@@ -144,7 +144,7 @@ protected:
void xPredIntraBDPCM ( const CPelBuf &pSrc, PelBuf &pDst, const uint32_t dirMode, const ClpRng& clpRng );
Pel xGetPredValDc ( const CPelBuf &pSrc, const Size &dstSize );
-#if IDCC_TPM_JEM
+#if JVET_V0130_INTRA_TMP
bool isRefTemplateAvailable(CodingUnit& cu, CompArea& area);
#endif
diff --git a/source/Lib/CommonLib/Slice.cpp b/source/Lib/CommonLib/Slice.cpp
index 94ef8a9abcb4733eb8b16024af158ac1cfdbf6a5..934d14c39ac50bc84650f081b54be54510cfe0e7 100644
--- a/source/Lib/CommonLib/Slice.cpp
+++ b/source/Lib/CommonLib/Slice.cpp
@@ -3113,9 +3113,9 @@ SPS::SPS()
#if ENABLE_DIMD
, m_dimd ( false )
#endif
-#if IDCC_TPM_JEM
-, m_IntraTMP ( false )
-, m_IntraTMP_MaxSize ( 64 )
+#if JVET_V0130_INTRA_TMP
+, m_intraTMP ( false )
+, m_intraTmpMaxSize ( 64 )
#endif
#if ENABLE_OBMC
, m_OBMC ( false )
diff --git a/source/Lib/CommonLib/Slice.h b/source/Lib/CommonLib/Slice.h
index ec74b504c951a6289b65f4852c511a86dc6f5efe..f53244b1dbbedad876acfc6f5385cf719faa4399 100644
--- a/source/Lib/CommonLib/Slice.h
+++ b/source/Lib/CommonLib/Slice.h
@@ -1651,9 +1651,9 @@ private:
#if ENABLE_DIMD
bool m_dimd;
#endif
-#if IDCC_TPM_JEM
- bool m_IntraTMP; ///< intra Template Matching
- unsigned m_IntraTMP_MaxSize; ///< max CU size for which intra TMP is allowed
+#if JVET_V0130_INTRA_TMP
+ bool m_intraTMP; ///< intra Template Matching
+ unsigned m_intraTmpMaxSize; ///< max CU size for which intra TMP is allowed
#endif
#if ENABLE_OBMC
bool m_OBMC;
@@ -2077,11 +2077,11 @@ void setCCALFEnabledFlag( bool b )
void setUseDimd ( bool b ) { m_dimd = b; }
bool getUseDimd () const { return m_dimd; }
#endif
-#if IDCC_TPM_JEM
- void setUseIntraTMP(bool b) { m_IntraTMP = b; }
- bool getUseIntraTMP() const { return m_IntraTMP; }
- void setIntraTMPMaxSize(unsigned n) { m_IntraTMP_MaxSize = n; }
- unsigned getIntraTMPMaxSize() const { return m_IntraTMP_MaxSize; }
+#if JVET_V0130_INTRA_TMP
+ void setUseIntraTMP (bool b) { m_intraTMP = b; }
+ bool getUseIntraTMP () const { return m_intraTMP; }
+ void setIntraTMPMaxSize (unsigned n) { m_intraTmpMaxSize = n; }
+ unsigned getIntraTMPMaxSize () const { return m_intraTmpMaxSize; }
#endif
#if ENABLE_OBMC
void setUseOBMC ( bool b ) { m_OBMC = b; }
diff --git a/source/Lib/CommonLib/TrQuant.cpp b/source/Lib/CommonLib/TrQuant.cpp
index c74ab4c431e67373d95234b32e93db2c78fd177f..5889375a2e2f21adafaf099111dfe9810d04b936 100644
--- a/source/Lib/CommonLib/TrQuant.cpp
+++ b/source/Lib/CommonLib/TrQuant.cpp
@@ -55,16 +55,14 @@
#include "CommonLib/CodingStatistics.h"
#endif
-#if IDCC_TMP_SIMD
+#if ENABLE_SIMD_TMP
#include "CommonDefX86.h"
#endif
-#if IDCC_TPM_JEM
-
+#if JVET_V0130_INTRA_TMP
unsigned int g_uiDepth2Width[5] = { 4, 8, 16, 32, 64 };
#endif
-
struct coeffGroupRDStats
{
int iNNZbeforePos0;
@@ -197,8 +195,8 @@ TrQuant::TrQuant() : m_quant( nullptr )
m_fwdICT[-2] = fwdTransformCbCr<-2>;
m_fwdICT[ 3] = fwdTransformCbCr< 3>;
m_fwdICT[-3] = fwdTransformCbCr<-3>;
-#if IDCC_TPM_JEM
- m_pppTarPatch = NULL;
+#if JVET_V0130_INTRA_TMP
+ m_pppTarPatch = NULL;
#endif
}
}
@@ -210,17 +208,15 @@ TrQuant::~TrQuant()
delete m_quant;
m_quant = nullptr;
}
-#if IDCC_TPM_JEM
-#endif
-#if IDCC_TPM_JEM
+#if JVET_V0130_INTRA_TMP
if (m_pppTarPatch != NULL)
{
for (unsigned int uiDepth = 0; uiDepth < USE_MORE_BLOCKSIZE_DEPTH_MAX; uiDepth++)
{
unsigned int blkSize = g_uiDepth2Width[uiDepth];
- unsigned int patchSize = blkSize + IDCC_TemplateSize;
+ unsigned int patchSize = blkSize + TMP_TEMPLATE_SIZE;
for (unsigned int uiRow = 0; uiRow < patchSize; uiRow++)
{
if (m_pppTarPatch[uiDepth][uiRow] != NULL)
@@ -275,7 +271,7 @@ void TrQuant::init( const Quant* otherQuant,
}
-#if IDCC_TPM_JEM
+#if JVET_V0130_INTRA_TMP
unsigned int blkSize;
if (m_pppTarPatch == NULL)
@@ -285,7 +281,7 @@ void TrQuant::init( const Quant* otherQuant,
{
blkSize = g_uiDepth2Width[uiDepth];
- unsigned int patchSize = blkSize + IDCC_TemplateSize;
+ unsigned int patchSize = blkSize + TMP_TEMPLATE_SIZE;
m_pppTarPatch[uiDepth] = new Pel * [patchSize];
for (unsigned int uiRow = 0; uiRow < patchSize; uiRow++)
{
@@ -328,6 +324,9 @@ void TrQuant::init( const Quant* otherQuant,
#if ENABLE_SIMD_SIGN_PREDICTION
m_computeSAD = xComputeSAD;
#endif
+#if INTRA_TEMPLATE_MATCHING
+ m_calcTemplateDiff = calcTemplateDiff;
+#endif
#if ENABLE_SIMD_SIGN_PREDICTION || TRANSFORM_SIMD_OPT
#ifdef TARGET_SIMD_X86
@@ -435,32 +434,30 @@ void TrQuant::invLfnstNxN( int* src, int* dst, const uint32_t mode, const uint32
}
}
-#if IDCC_TPM_JEM
-void insertNode(DistType diff, int& iXOffset, int& iYOffset, DistType& pDiff, int& pX, int& pY, short& pId, unsigned int& setId)
+#if JVET_V0130_INTRA_TMP
+void insertNode(int diff, int& iXOffset, int& iYOffset, int& pDiff, int& pX, int& pY, short& pId, unsigned int& setId)
{
pDiff = diff;
pX = iXOffset;
pY = iYOffset;
pId = setId;
}
-#if IDCC_TPM_JEM
+
void clipMvIntraConstraint(CodingUnit* pcCU, int regionId, int& iHorMin, int& iHorMax, int& iVerMin, int& iVerMax, unsigned int uiTemplateSize, unsigned int uiBlkWidth, unsigned int uiBlkHeight, int iCurrY, int iCurrX, int offsetLCUY, int offsetLCUX)
{
- int SearchRange_Height, SearchRange_Width;
-
- SearchRange_Width = IDCC_SearchRangeMultFactor * uiBlkWidth;
- SearchRange_Height = IDCC_SearchRangeMultFactor * uiBlkHeight;
- int iMvShift = 0;
+ int searchRangeWidth = TMP_SEARCH_RANGE_MULT_FACTOR * uiBlkWidth;
+ int searchRangeHeight = TMP_SEARCH_RANGE_MULT_FACTOR * uiBlkHeight;
+ int iMvShift = 0;
int iTemplateSize = uiTemplateSize;
int iBlkWidth = uiBlkWidth;
int iBlkHeight = uiBlkHeight;
if (regionId == 0) //above outside LCU
{
- iHorMax = std::min((iCurrX + SearchRange_Width) << iMvShift, (int)((pcCU->cs->sps->getMaxPicWidthInLumaSamples() - iBlkWidth) << iMvShift));
- iHorMin = std::max((iTemplateSize) << iMvShift, (iCurrX - SearchRange_Width) << iMvShift);
+ iHorMax = std::min((iCurrX + searchRangeWidth) << iMvShift, (int)((pcCU->cs->sps->getMaxPicWidthInLumaSamples() - iBlkWidth) << iMvShift));
+ iHorMin = std::max((iTemplateSize) << iMvShift, (iCurrX - searchRangeWidth) << iMvShift);
iVerMax = (iCurrY - iBlkHeight - offsetLCUY) << iMvShift;
- iVerMin = std::max(((iTemplateSize) << iMvShift), ((iCurrY - SearchRange_Height) << iMvShift));
+ iVerMin = std::max(((iTemplateSize) << iMvShift), ((iCurrY - searchRangeHeight) << iMvShift));
iHorMin = iHorMin - iCurrX;
iHorMax = iHorMax - iCurrX;
@@ -470,7 +467,7 @@ void clipMvIntraConstraint(CodingUnit* pcCU, int regionId, int& iHorMin, int& iH
else if (regionId == 1) //left outside LCU
{
iHorMax = (iCurrX - offsetLCUX - iBlkWidth) << iMvShift;
- iHorMin = std::max((iTemplateSize) << iMvShift, (iCurrX - SearchRange_Width) << iMvShift);
+ iHorMin = std::max((iTemplateSize) << iMvShift, (iCurrX - searchRangeWidth) << iMvShift);
iVerMin = std::max((iTemplateSize) << iMvShift, (iCurrY - iBlkHeight - offsetLCUY) << iMvShift);
iVerMax = (iCurrY) << iMvShift;
@@ -482,7 +479,7 @@ void clipMvIntraConstraint(CodingUnit* pcCU, int regionId, int& iHorMin, int& iH
}
else if (regionId == 2) //left outside LCU (can reach the bottom row of LCU)
{
- iHorMin = std::max((iTemplateSize) << iMvShift, (iCurrX - SearchRange_Width) << iMvShift);
+ iHorMin = std::max((iTemplateSize) << iMvShift, (iCurrX - searchRangeWidth) << iMvShift);
iHorMax = (iCurrX - offsetLCUX - iBlkWidth) << iMvShift;
iVerMin = (iCurrY + 1) << iMvShift;
iVerMax = std::min(pcCU->cs->sps->getMaxPicHeightInLumaSamples() - iBlkHeight, (iCurrY - offsetLCUY + pcCU->cs->sps->getCTUSize() - iBlkHeight) << iMvShift);
@@ -493,10 +490,7 @@ void clipMvIntraConstraint(CodingUnit* pcCU, int regionId, int& iHorMin, int& iH
iVerMin = iVerMin - iCurrY;
}
}
-#endif
-#endif
-#if IDCC_TPM_JEM
TempLibFast::TempLibFast()
{
}
@@ -504,12 +498,10 @@ TempLibFast::TempLibFast()
TempLibFast::~TempLibFast()
{
}
-#endif
-#if IDCC_TPM_JEM
void TempLibFast::initTemplateDiff(unsigned int uiPatchWidth, unsigned int uiPatchHeight, unsigned int uiBlkWidth, unsigned int uiBlkHeight, int bitDepth)
{
- DistType maxValue = ((1 << bitDepth) >> (INIT_THRESHOULD_SHIFTBITS)) * (uiPatchHeight * uiPatchWidth - uiBlkHeight * uiBlkWidth);
+ int maxValue = ((1 << bitDepth) >> (INIT_THRESHOULD_SHIFTBITS)) * (uiPatchHeight * uiPatchWidth - uiBlkHeight * uiBlkWidth);
m_diffMax = maxValue;
{
m_pDiff = maxValue;
@@ -519,8 +511,8 @@ void TempLibFast::initTemplateDiff(unsigned int uiPatchWidth, unsigned int uiPat
void TrQuant::getTargetTemplate(CodingUnit* pcCU, unsigned int uiBlkWidth, unsigned int uiBlkHeight)
{
const ComponentID compID = COMPONENT_Y;
- unsigned int uiPatchWidth = uiBlkWidth + IDCC_TemplateSize;
- unsigned int uiPatchHeight = uiBlkHeight + IDCC_TemplateSize;
+ unsigned int uiPatchWidth = uiBlkWidth + TMP_TEMPLATE_SIZE;
+ unsigned int uiPatchHeight = uiBlkHeight + TMP_TEMPLATE_SIZE;
unsigned int uiTarDepth = floorLog2(std::max(uiBlkHeight, uiBlkWidth)) - 2;
Pel** tarPatch = m_pppTarPatch[uiTarDepth];
CompArea area = pcCU->blocks[compID];
@@ -528,13 +520,11 @@ void TrQuant::getTargetTemplate(CodingUnit* pcCU, unsigned int uiBlkWidth, unsig
unsigned int uiPicStride = pcCU->cs->picture->getRecoBuf(compID).stride;
unsigned int uiY, uiX;
-
-
//fill template
//up-left & up
Pel* tarTemp;
- Pel* pCurrTemp = pCurrStart - IDCC_TemplateSize * uiPicStride - IDCC_TemplateSize;
- for (uiY = 0; uiY < IDCC_TemplateSize; uiY++)
+ Pel* pCurrTemp = pCurrStart - TMP_TEMPLATE_SIZE * uiPicStride - TMP_TEMPLATE_SIZE;
+ for (uiY = 0; uiY < TMP_TEMPLATE_SIZE; uiY++)
{
tarTemp = tarPatch[uiY];
for (uiX = 0; uiX < uiPatchWidth; uiX++)
@@ -544,10 +534,10 @@ void TrQuant::getTargetTemplate(CodingUnit* pcCU, unsigned int uiBlkWidth, unsig
pCurrTemp += uiPicStride;
}
//left
- for (uiY = IDCC_TemplateSize; uiY < uiPatchHeight; uiY++)
+ for (uiY = TMP_TEMPLATE_SIZE; uiY < uiPatchHeight; uiY++)
{
tarTemp = tarPatch[uiY];
- for (uiX = 0; uiX < IDCC_TemplateSize; uiX++)
+ for (uiX = 0; uiX < TMP_TEMPLATE_SIZE; uiX++)
{
tarTemp[uiX] = pCurrTemp[uiX];
}
@@ -559,8 +549,8 @@ void TrQuant::candidateSearchIntra(CodingUnit* pcCU, unsigned int uiBlkWidth, un
{
const ComponentID compID = COMPONENT_Y;
const int channelBitDepth = pcCU->cs->sps->getBitDepth(toChannelType(compID));
- unsigned int uiPatchWidth = uiBlkWidth + IDCC_TemplateSize;
- unsigned int uiPatchHeight = uiBlkHeight + IDCC_TemplateSize;
+ unsigned int uiPatchWidth = uiBlkWidth + TMP_TEMPLATE_SIZE;
+ unsigned int uiPatchHeight = uiBlkHeight + TMP_TEMPLATE_SIZE;
unsigned int uiTarDepth = floorLog2(std::max(uiBlkWidth, uiBlkHeight)) - 2;
Pel** tarPatch = getTargetPatch(uiTarDepth);
//Initialize the library for saving the best candidates
@@ -568,25 +558,29 @@ void TrQuant::candidateSearchIntra(CodingUnit* pcCU, unsigned int uiBlkWidth, un
short setId = 0; //record the reference picture.
searchCandidateFromOnePicIntra(pcCU, tarPatch, uiPatchWidth, uiPatchHeight, setId);
//count collected candidate number
- DistType pDiff = m_tempLibFast.getDiff();
- DistType maxDiff = m_tempLibFast.getDiffMax();
+ int pDiff = m_tempLibFast.getDiff();
+ int maxDiff = m_tempLibFast.getDiffMax();
- if (pDiff < maxDiff)
- m_uiVaildCandiNum = 1;
- else
- m_uiVaildCandiNum = 0;
+ if( pDiff < maxDiff )
+ {
+ m_uiVaildCandiNum = 1;
+ }
+ else
+ {
+ m_uiVaildCandiNum = 0;
+ }
}
void TrQuant::searchCandidateFromOnePicIntra(CodingUnit* pcCU, Pel** tarPatch, unsigned int uiPatchWidth, unsigned int uiPatchHeight, unsigned int setId)
{
const ComponentID compID = COMPONENT_Y;
- unsigned int uiBlkWidth = uiPatchWidth - IDCC_TemplateSize;
- unsigned int uiBlkHeight = uiPatchHeight - IDCC_TemplateSize;
+ unsigned int uiBlkWidth = uiPatchWidth - TMP_TEMPLATE_SIZE;
+ unsigned int uiBlkHeight = uiPatchHeight - TMP_TEMPLATE_SIZE;
int pX = m_tempLibFast.getX();
int pY = m_tempLibFast.getY();
- DistType pDiff = m_tempLibFast.getDiff();
+ int pDiff = m_tempLibFast.getDiff();
short pId = m_tempLibFast.getId();
CompArea area = pcCU->blocks[compID];
int refStride = pcCU->cs->picture->getRecoBuf(compID).stride;
@@ -594,9 +588,7 @@ void TrQuant::searchCandidateFromOnePicIntra(CodingUnit* pcCU, Pel** tarPatch,
Pel* ref = pcCU->cs->picture->getRecoBuf(area).buf;
setRefPicUsed(ref); //facilitate the access of each candidate point
-
setStride(refStride);
-
Mv cTmpMvPred;
cTmpMvPred.setZero();
@@ -614,27 +606,23 @@ void TrQuant::searchCandidateFromOnePicIntra(CodingUnit* pcCU, Pel** tarPatch,
int iYOffset, iXOffset;
- DistType diff;
+ int diff;
Pel* refCurr;
-
-#define REGION_NUM 3
- int mvYMins[REGION_NUM];
- int mvYMaxs[REGION_NUM];
- int mvXMins[REGION_NUM];
- int mvXMaxs[REGION_NUM];
- int regionNum = REGION_NUM;
+ const int regionNum = 3;
+ int mvYMins[regionNum];
+ int mvYMaxs[regionNum];
+ int mvXMins[regionNum];
+ int mvXMaxs[regionNum];
int regionId = 0;
-
//1. check the near pixels within LCU
//above pixels in LCU
- int iTemplateSize = IDCC_TemplateSize;
+ int iTemplateSize = TMP_TEMPLATE_SIZE;
int iBlkWidth = uiBlkWidth;
int iBlkHeight = uiBlkHeight;
regionId = 0;
int iMvShift = 0;
-
int iVerMin = std::max(((iTemplateSize) << iMvShift), (iCurrY - offsetLCUY - iBlkHeight + 1) << iMvShift);
int iVerMax = (iCurrY - iBlkHeight) << iMvShift;
@@ -646,8 +634,6 @@ void TrQuant::searchCandidateFromOnePicIntra(CodingUnit* pcCU, Pel** tarPatch,
mvYMins[regionId] = iVerMin - iCurrY;
mvYMaxs[regionId] = iVerMax - iCurrY;
-
-
//check within CTU pixels
for (regionId = 0; regionId < 1; regionId++)
{
@@ -659,12 +645,13 @@ void TrQuant::searchCandidateFromOnePicIntra(CodingUnit* pcCU, Pel** tarPatch,
{
continue;
}
+
for (iYOffset = mvYMax; iYOffset >= mvYMin; iYOffset--)
{
for (iXOffset = mvXMax; iXOffset >= mvXMin; iXOffset--)
{
refCurr = ref + iYOffset * refStride + iXOffset;
- diff = calcTemplateDiff(refCurr, refStride, tarPatch, uiPatchWidth, uiPatchHeight, pDiff);
+ diff = m_calcTemplateDiff(refCurr, refStride, tarPatch, uiPatchWidth, uiPatchHeight, pDiff);
if (diff < (pDiff))
{
insertNode(diff, iXOffset, iYOffset, pDiff, pX, pY, pId, setId);
@@ -680,8 +667,9 @@ void TrQuant::searchCandidateFromOnePicIntra(CodingUnit* pcCU, Pel** tarPatch,
//2. check the pixels outside CTU
for (regionId = 0; regionId < regionNum; regionId++)
{// this function fills in the range the template matching for pixels outside the current CTU
- clipMvIntraConstraint(pcCU, regionId, mvXMins[regionId], mvXMaxs[regionId], mvYMins[regionId], mvYMaxs[regionId], IDCC_TemplateSize, uiBlkWidth, uiBlkHeight, iCurrY, iCurrX, offsetLCUY, offsetLCUX);
+ clipMvIntraConstraint(pcCU, regionId, mvXMins[regionId], mvXMaxs[regionId], mvYMins[regionId], mvYMaxs[regionId], TMP_TEMPLATE_SIZE, uiBlkWidth, uiBlkHeight, iCurrY, iCurrX, offsetLCUY, offsetLCUX);
}
+
for (regionId = 0; regionId < regionNum; regionId++)
{
int mvYMin = mvYMins[regionId];
@@ -697,11 +685,13 @@ void TrQuant::searchCandidateFromOnePicIntra(CodingUnit* pcCU, Pel** tarPatch,
for (iXOffset = mvXMax; iXOffset >= mvXMin; iXOffset--)
{
refCurr = ref + iYOffset * refStride + iXOffset;
- diff = calcTemplateDiff(refCurr, refStride, tarPatch, uiPatchWidth, uiPatchHeight, pDiff);
+ diff = m_calcTemplateDiff(refCurr, refStride, tarPatch, uiPatchWidth, uiPatchHeight, pDiff);
+
if (diff < (pDiff))
{
insertNode(diff, iXOffset, iYOffset, pDiff, pX, pY, pId, setId);
}
+
if (pDiff == 0)
{
regionId = regionNum;
@@ -709,6 +699,7 @@ void TrQuant::searchCandidateFromOnePicIntra(CodingUnit* pcCU, Pel** tarPatch,
}
}
}
+
m_tempLibFast.m_pX = pX;
m_tempLibFast.m_pY = pY;
m_tempLibFast.m_pDiff = pDiff;
@@ -717,8 +708,8 @@ void TrQuant::searchCandidateFromOnePicIntra(CodingUnit* pcCU, Pel** tarPatch,
bool TrQuant::generateTMPrediction(Pel* piPred, unsigned int uiStride, unsigned int uiBlkWidth, unsigned int uiBlkHeight, int& foundCandiNum)
{
bool bSucceedFlag = true;
- unsigned int uiPatchWidth = uiBlkWidth + IDCC_TemplateSize;
- unsigned int uiPatchHeight = uiBlkHeight + IDCC_TemplateSize;
+ unsigned int uiPatchWidth = uiBlkWidth + TMP_TEMPLATE_SIZE;
+ unsigned int uiPatchHeight = uiBlkHeight + TMP_TEMPLATE_SIZE;
foundCandiNum = m_uiVaildCandiNum;
if (foundCandiNum < 1)
@@ -732,8 +723,8 @@ bool TrQuant::generateTMPrediction(Pel* piPred, unsigned int uiStride, unsigned
int picStride = getStride();
int iOffsetY, iOffsetX;
Pel* refTarget;
- unsigned int uiHeight = uiPatchHeight - IDCC_TemplateSize;
- unsigned int uiWidth = uiPatchWidth - IDCC_TemplateSize;
+ unsigned int uiHeight = uiPatchHeight - TMP_TEMPLATE_SIZE;
+ unsigned int uiWidth = uiPatchWidth - TMP_TEMPLATE_SIZE;
//the data center: we use the prediction block as the center now.
//collect the candidates
@@ -755,128 +746,46 @@ bool TrQuant::generateTMPrediction(Pel* piPred, unsigned int uiStride, unsigned
return bSucceedFlag;
}
-DistType TrQuant::calcTemplateDiff(Pel* ref, unsigned int uiStride, Pel** tarPatch, unsigned int uiPatchWidth, unsigned int uiPatchHeight, DistType iMax)
+int TrQuant::calcTemplateDiff(Pel* ref, unsigned int uiStride, Pel** tarPatch, unsigned int uiPatchWidth, unsigned int uiPatchHeight, int iMax)
{
- DistType iDiffSum = 0;
- int iY;
- Pel* refPatchRow = ref - IDCC_TemplateSize * uiStride - IDCC_TemplateSize;
- Pel* tarPatchRow;
-
- uint32_t uiSum;
- // horizontal difference
- for (iY = 0; iY < IDCC_TemplateSize; iY++)
- {
- tarPatchRow = tarPatch[iY];
- const short* pSrc1 = (const short*)tarPatchRow;
- const short* pSrc2 = (const short*)refPatchRow;
-
- // SIMD difference
- //int iRows = uiPatchHeight;
- int iCols = uiPatchWidth;
- if ((iCols & 7) == 0)
- {
- // Do with step of 8
- __m128i vzero = _mm_setzero_si128();
- __m128i vsum32 = vzero;
- //for (int iY = 0; iY < iRows; iY += iSubStep)
- {
- __m128i vsum16 = vzero;
- for (int iX = 0; iX < iCols; iX += 8)
- {
- __m128i vsrc1 = _mm_loadu_si128((const __m128i*)(&pSrc1[iX]));
- __m128i vsrc2 = _mm_lddqu_si128((const __m128i*)(&pSrc2[iX]));
- vsum16 = _mm_add_epi16(vsum16, _mm_abs_epi16(_mm_sub_epi16(vsrc1, vsrc2)));
- }
- __m128i vsumtemp = _mm_add_epi32(_mm_unpacklo_epi16(vsum16, vzero), _mm_unpackhi_epi16(vsum16, vzero));
- vsum32 = _mm_add_epi32(vsum32, vsumtemp);
- //pSrc1 += iStrideSrc1;
- //pSrc2 += iStrideSrc2;
- }
- vsum32 = _mm_add_epi32(vsum32, _mm_shuffle_epi32(vsum32, 0x4e)); // 01001110
- vsum32 = _mm_add_epi32(vsum32, _mm_shuffle_epi32(vsum32, 0xb1)); // 10110001
- uiSum = _mm_cvtsi128_si32(vsum32);
- }
- else
- {
- // Do with step of 4
- __m128i vzero = _mm_setzero_si128();
- __m128i vsum32 = vzero;
- //for (int iY = 0; iY < iRows; iY += iSubStep)
- {
- __m128i vsum16 = vzero;
- for (int iX = 0; iX < iCols; iX += 4)
- {
- __m128i vsrc1 = _mm_loadl_epi64((const __m128i*) & pSrc1[iX]);
- __m128i vsrc2 = _mm_loadl_epi64((const __m128i*) & pSrc2[iX]);
- vsum16 = _mm_add_epi16(vsum16, _mm_abs_epi16(_mm_sub_epi16(vsrc1, vsrc2)));
- }
- __m128i vsumtemp = _mm_add_epi32(_mm_unpacklo_epi16(vsum16, vzero), _mm_unpackhi_epi16(vsum16, vzero));
- vsum32 = _mm_add_epi32(vsum32, vsumtemp);
- //pSrc1 += iStrideSrc1;
- //pSrc2 += iStrideSrc2;
- }
- vsum32 = _mm_add_epi32(vsum32, _mm_shuffle_epi32(vsum32, 0x4e)); // 01001110
- vsum32 = _mm_add_epi32(vsum32, _mm_shuffle_epi32(vsum32, 0xb1)); // 10110001
- uiSum = _mm_cvtsi128_si32(vsum32);
- }
- iDiffSum += uiSum;
-
- if (iDiffSum > iMax) //for speeding up
- {
- return iDiffSum;
- }
- // update location
- refPatchRow += uiStride;
- }
+ int iDiffSum = 0;
+ Pel* refPatchRow = ref - TMP_TEMPLATE_SIZE * uiStride - TMP_TEMPLATE_SIZE;
+ Pel* tarPatchRow;
- // vertical difference
- int iCols = IDCC_TemplateSize;
- for (iY = IDCC_TemplateSize; iY < uiPatchHeight; iY++)
- {
- tarPatchRow = tarPatch[iY];
- const short* pSrc1 = (const short*)tarPatchRow;
- const short* pSrc2 = (const short*)refPatchRow ;
-
- // SIMD difference
-
- // Do with step of 4
- __m128i vzero = _mm_setzero_si128();
- __m128i vsum32 = vzero;
- //for (int iY = 0; iY < iRows; iY += iSubStep)
- {
- __m128i vsum16 = vzero;
- for (int iX = 0; iX < iCols; iX += 4)
- {
- __m128i vsrc1 = _mm_loadl_epi64((const __m128i*) & pSrc1[iX]);
- __m128i vsrc2 = _mm_loadl_epi64((const __m128i*) & pSrc2[iX]);
- vsum16 = _mm_add_epi16(vsum16, _mm_abs_epi16(_mm_sub_epi16(vsrc1, vsrc2)));
- }
- __m128i vsumtemp = _mm_add_epi32(_mm_unpacklo_epi16(vsum16, vzero), _mm_unpackhi_epi16(vsum16, vzero));
- vsum32 = _mm_add_epi32(vsum32, vsumtemp);
- //pSrc1 += iStrideSrc1;
- //pSrc2 += iStrideSrc2;
- }
- vsum32 = _mm_add_epi32(vsum32, _mm_shuffle_epi32(vsum32, 0x4e)); // 01001110
- vsum32 = _mm_add_epi32(vsum32, _mm_shuffle_epi32(vsum32, 0xb1)); // 10110001
- uiSum = _mm_cvtsi128_si32(vsum32);
+ // horizontal difference
+ for( int iY = 0; iY < TMP_TEMPLATE_SIZE; iY++ )
+ {
+ tarPatchRow = tarPatch[iY];
+ for( int iX = 0; iX < uiPatchWidth; iX++ )
+ {
+ iDiffSum += abs( refPatchRow[iX] - tarPatchRow[iX] );
+ }
+ if( iDiffSum > iMax ) //for speeding up
+ {
+ return iDiffSum;
+ }
+ refPatchRow += uiStride;
+ }
- iDiffSum += uiSum;
+ // vertical difference
+ for( int iY = TMP_TEMPLATE_SIZE; iY < uiPatchHeight; iY++ )
+ {
+ tarPatchRow = tarPatch[iY];
+ for( int iX = 0; iX < TMP_TEMPLATE_SIZE; iX++ )
+ {
+ iDiffSum += abs( refPatchRow[iX] - tarPatchRow[iX] );
+ }
+ if( iDiffSum > iMax ) //for speeding up
+ {
+ return iDiffSum;
+ }
+ refPatchRow += uiStride;
+ }
- if (iDiffSum > iMax) //for speeding up
- {
- return iDiffSum;
- }
- // update location
- refPatchRow += uiStride;
- }
-
- return iDiffSum;
-
+ return iDiffSum;
}
#endif
-
-
uint32_t TrQuant::getLFNSTIntraMode( int wideAngPredMode )
{
uint32_t intraMode;
@@ -930,11 +839,11 @@ void TrQuant::xInvLfnst( const TransformUnit &tu, const ComponentID compID )
{
intraMode = PLANAR_IDX;
}
-#if IDCC_TPM_JEM
- if (PU::isTmp(*tu.cs->getPU(area.pos(), toChannelType(compID)), toChannelType(compID)))
- {
- intraMode = PLANAR_IDX;
- }
+#if JVET_V0130_INTRA_TMP
+ if( PU::isTmp( *tu.cs->getPU( area.pos(), toChannelType( compID ) ), toChannelType( compID ) ) )
+ {
+ intraMode = PLANAR_IDX;
+ }
#endif
CHECK( intraMode >= NUM_INTRA_MODE - 1, "Invalid intra mode" );
@@ -1076,11 +985,11 @@ void TrQuant::xFwdLfnst( const TransformUnit &tu, const ComponentID compID, cons
{
intraMode = PLANAR_IDX;
}
-#if IDCC_TPM_JEM
- if (PU::isTmp(*tu.cs->getPU(area.pos(), toChannelType(compID)), toChannelType(compID)))
- {
- intraMode = PLANAR_IDX;
- }
+#if JVET_V0130_INTRA_TMP
+ if( PU::isTmp( *tu.cs->getPU( area.pos(), toChannelType( compID ) ), toChannelType( compID ) ) )
+ {
+ intraMode = PLANAR_IDX;
+ }
#endif
CHECK( intraMode >= NUM_INTRA_MODE - 1, "Invalid intra mode" );
@@ -1319,8 +1228,8 @@ void TrQuant::getTrTypes(const TransformUnit tu, const ComponentID compID, int &
return;
}
-#if IDCC_TPM_JEM
- if (isImplicitMTS || isISP || tu.cu->TmpFlag)
+#if JVET_V0130_INTRA_TMP
+ if (isImplicitMTS || isISP || tu.cu->tmpFlag)
#else
if (isImplicitMTS || isISP)
#endif
diff --git a/source/Lib/CommonLib/TrQuant.h b/source/Lib/CommonLib/TrQuant.h
index f116d83e256f8ab45ccf75ef0296a4a6ca14bcf1..8556825e4e00241ab087347b8a5609876dfea999 100644
--- a/source/Lib/CommonLib/TrQuant.h
+++ b/source/Lib/CommonLib/TrQuant.h
@@ -57,10 +57,7 @@ typedef void InvTrans(const TCoeff*, TCoeff*, int, int, int, int, const TCoeff,
-#if IDCC_TPM_JEM
-
-
-#define MAX_1DTRANS_LEN (1 << (((USE_MORE_BLOCKSIZE_DEPTH_MAX) + 1) << 1)) ///< 4x4 = 16, 8x8 = 64, 16x16=256, 32x32 = 1024
+#if JVET_V0130_INTRA_TMP
extern unsigned int g_uiDepth2Width[5];
extern unsigned int g_uiDepth2MaxCandiNum[5];
@@ -71,13 +68,13 @@ public:
int m_pY; //offset Y
int m_pXInteger; //offset X for integer pixel search
int m_pYInteger; //offset Y for integer pixel search
- DistType m_pDiffInteger;
+ int m_pDiffInteger;
int getXInteger() { return m_pXInteger; }
int getYInteger() { return m_pYInteger; }
- DistType getDiffInteger() { return m_pDiffInteger; }
+ int getDiffInteger() { return m_pDiffInteger; }
short m_pIdInteger; //frame id
short getIdInteger() { return m_pIdInteger; }
- DistType m_pDiff; //mse
+ int m_pDiff; //mse
short m_pId; //frame id
@@ -86,7 +83,7 @@ public:
//void init();
int getX() { return m_pX; }
int getY() { return m_pY; }
- DistType getDiff() { return m_pDiff; }
+ int getDiff() { return m_pDiff; }
short getId() { return m_pId; }
/*void initDiff(unsigned int uiPatchSize, int bitDepth);
void initDiff(unsigned int uiPatchSize, int bitDepth, int iCandiNumber);*/
@@ -132,13 +129,14 @@ public:
void fwdLfnstNxN( int* src, int* dst, const uint32_t mode, const uint32_t index, const uint32_t size, int zeroOutSize );
void invLfnstNxN( int* src, int* dst, const uint32_t mode, const uint32_t index, const uint32_t size, int zeroOutSize );
#endif
-#if IDCC_TPM_JEM
- DistType calcTemplateDiff(Pel* ref, unsigned int uiStride, Pel** tarPatch, unsigned int uiPatchWidth, unsigned int uiPatchHeight, DistType iMax);
- Pel** getTargetPatch(unsigned int uiDepth) { return m_pppTarPatch[uiDepth]; }
- Pel* getRefPicUsed() { return m_refPicUsed; }
- void setRefPicUsed(Pel* ref) { m_refPicUsed = ref; }
- unsigned int getStride() { return m_uiPicStride; }
- void setStride(unsigned int uiPicStride) { m_uiPicStride = uiPicStride; }
+#if JVET_V0130_INTRA_TMP
+ int ( *m_calcTemplateDiff )(Pel* ref, unsigned int uiStride, Pel** tarPatch, unsigned int uiPatchWidth, unsigned int uiPatchHeight, int iMax);
+ static int calcTemplateDiff(Pel* ref, unsigned int uiStride, Pel** tarPatch, unsigned int uiPatchWidth, unsigned int uiPatchHeight, int iMax);
+ Pel** getTargetPatch(unsigned int uiDepth) { return m_pppTarPatch[uiDepth]; }
+ Pel* getRefPicUsed() { return m_refPicUsed; }
+ void setRefPicUsed(Pel* ref) { m_refPicUsed = ref; }
+ unsigned int getStride() { return m_uiPicStride; }
+ void setStride(unsigned int uiPicStride) { m_uiPicStride = uiPicStride; }
void searchCandidateFromOnePicIntra(CodingUnit* pcCU, Pel** tarPatch, unsigned int uiPatchWidth, unsigned int uiPatchHeight, unsigned int setId);
void candidateSearchIntra(CodingUnit* pcCU, unsigned int uiBlkWidth, unsigned int uiBlkHeight);
@@ -199,14 +197,14 @@ public:
protected:
TCoeff m_tempCoeff[MAX_TB_SIZEY * MAX_TB_SIZEY];
-#if IDCC_TPM_JEM
- int m_uiPartLibSize;
- TempLibFast m_tempLibFast;
- Pel* m_refPicUsed;
- Picture* m_refPicBuf;
+#if JVET_V0130_INTRA_TMP
+ int m_uiPartLibSize;
+ TempLibFast m_tempLibFast;
+ Pel* m_refPicUsed;
+ Picture* m_refPicBuf;
unsigned int m_uiPicStride;
unsigned int m_uiVaildCandiNum;
- Pel*** m_pppTarPatch;
+ Pel*** m_pppTarPatch;
#endif
#if SIGN_PREDICTION
Pel m_tempSignPredResid[SIGN_PRED_MAX_BS * SIGN_PRED_MAX_BS * 2]{0};
@@ -282,7 +280,7 @@ private:
static void fastInverseTransform_SIMD( const TCoeff *coeff, TCoeff *block, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum );
#endif
-#if ENABLE_SIMD_SIGN_PREDICTION || TRANSFORM_SIMD_OPT
+#if ENABLE_SIMD_SIGN_PREDICTION || TRANSFORM_SIMD_OPT || ENABLE_SIMD_TMP
#ifdef TARGET_SIMD_X86
void initTrQuantX86();
template <X86_VEXT vext>
diff --git a/source/Lib/CommonLib/TypeDef.h b/source/Lib/CommonLib/TypeDef.h
index 67363bc713e5b5816e9dff58f580e9e4cea20e5a..9cb01a37576d9d26be8fbad57ed503f2b3f6a32d 100644
--- a/source/Lib/CommonLib/TypeDef.h
+++ b/source/Lib/CommonLib/TypeDef.h
@@ -1,4 +1,4 @@
-/* The copyright in this software is being made available under the BSD
+/* The copyright in this software is being made available under the BSD
* License, included below. This software may be subject to other third party
* and contributor rights, including patent rights, and no such rights are
* granted under this license.
@@ -86,25 +86,6 @@
#define INTRA_RM_SMALL_BLOCK_SIZE_CONSTRAINTS 1 // Enable 2xN and Nx2 block by removing SCIPU constraints
#define CCLM_LATENCY_RESTRICTION_RMV 1 // remove the latency between luma and chroma restriction of CCLM
#define LMS_LINEAR_MODEL 1 // LMS for parameters derivation of CCLM and MMLM mode, Remove constraint in derivation of neighbouring samples
-#define IDCC_TPM_JEM 1 // template matching prediction as implemented in JEM-7.2
-
-#if IDCC_TPM_JEM
-
-#define IDCC_TMP_SIMD 1
-
-#define IDCC_SearchRangeMultFactor 5
-
-#if IDCC_TMP_SIMD
-#define IDCC_TemplateSize 4 // must be multiple of 4 for SIMD
-#else
-#define IDCC_TemplateSize 4
-#endif
-
-#define IDCC_TMP_MaxSize_Depth 6 // should be log2(IDCC_TMP_MaxSize): keep as 6 to avoid any error
-
-typedef int DistType;
-#endif
-
//-- inter
#define CIIP_RM_BLOCK_SIZE_CONSTRAINTS 1 // Remove the 64x64 restriction and enable 8x4/4x8 block for CIIP
@@ -129,6 +110,7 @@ typedef int DistType;
#define SECONDARY_MPM 1 // Primary MPM and Secondary MPM: Add neighbouring modes into MPMs from positions AR, BL, AL, derived modes
#define ENABLE_DIMD 1 // Decoder side intra mode derivation
#define JVET_V0087_DIMD_NO_ISP ENABLE_DIMD // disallow combination of DIMD and ISP
+#define JVET_V0130_INTRA_TMP 1 // JVET-V0130: template matching prediction
// Inter
#define CIIP_PDPC 1 // apply pdpc to megre prediction as a new CIIP mode (CIIP_PDPC) additional to CIIP mode
@@ -167,6 +149,12 @@ typedef int DistType;
#if SIGN_PREDICTION
#define ENABLE_SIMD_SIGN_PREDICTION 1
#endif
+#if JVET_V0130_INTRA_TMP
+#define ENABLE_SIMD_TMP 1
+#endif
+#if JVET_V0094_BILATERAL_FILTER
+#define ENABLE_SIMD_BILATERAL_FILTER 1
+#endif
#endif // tools
diff --git a/source/Lib/CommonLib/Unit.cpp b/source/Lib/CommonLib/Unit.cpp
index c281acf6ffe27554228e07829e54c7becf1f8a15..0d5f7f2a3ca47b9c20ed0a550dc52c9a4388ad19 100644
--- a/source/Lib/CommonLib/Unit.cpp
+++ b/source/Lib/CommonLib/Unit.cpp
@@ -303,8 +303,8 @@ CodingUnit& CodingUnit::operator=( const CodingUnit& other )
smvdMode = other.smvdMode;
ispMode = other.ispMode;
mipFlag = other.mipFlag;
-#if IDCC_TPM_JEM
- TmpFlag = other.TmpFlag;
+#if JVET_V0130_INTRA_TMP
+ tmpFlag = other.tmpFlag;
#endif
#if INTER_LIC
LICFlag = other.LICFlag;
@@ -387,8 +387,8 @@ void CodingUnit::initData()
smvdMode = 0;
ispMode = 0;
mipFlag = false;
-#if IDCC_TPM_JEM
- TmpFlag = false;
+#if JVET_V0130_INTRA_TMP
+ tmpFlag = false;
#endif
#if INTER_LIC
LICFlag = false;
diff --git a/source/Lib/CommonLib/Unit.h b/source/Lib/CommonLib/Unit.h
index e24b8c58bc8be25c0646fcbe72bf6af3eb96bd8a..31fd09ac0aff9313a9fc4a67fee1363dc61d2db2 100644
--- a/source/Lib/CommonLib/Unit.h
+++ b/source/Lib/CommonLib/Unit.h
@@ -334,8 +334,8 @@ struct CodingUnit : public UnitArea
uint8_t BcwIdx;
int8_t refIdxBi[2];
bool mipFlag;
-#if IDCC_TPM_JEM
- bool TmpFlag;
+#if JVET_V0130_INTRA_TMP
+ bool tmpFlag;
#endif
#if INTER_LIC
bool LICFlag;
diff --git a/source/Lib/CommonLib/UnitTools.cpp b/source/Lib/CommonLib/UnitTools.cpp
index 9f6b464441c3aac5f52ae109540b67491cf1415b..ce87e544860d11b5b58c14eb9b268e068afcf705 100644
--- a/source/Lib/CommonLib/UnitTools.cpp
+++ b/source/Lib/CommonLib/UnitTools.cpp
@@ -333,7 +333,7 @@ uint32_t CU::getCtuAddr( const CodingUnit &cu )
{
return getCtuAddr( cu.blocks[cu.chType].lumaPos(), *cu.cs->pcv );
}
-#if IDCC_TPM_JEM
+#if JVET_V0130_INTRA_TMP
Position CU::getCtuXYAddr(const CodingUnit& cu)
{
return Position((cu.blocks[cu.chType].lumaPos().x >> cu.cs->pcv->maxCUWidthLog2) << cu.cs->pcv->maxCUWidthLog2, (cu.blocks[cu.chType].lumaPos().y >> cu.cs->pcv->maxCUHeightLog2) << cu.cs->pcv->maxCUHeightLog2);
@@ -944,10 +944,10 @@ bool PU::isMIP(const PredictionUnit &pu, const ChannelType &chType)
return isDMChromaMIP(pu) && (pu.intraDir[CHANNEL_TYPE_CHROMA] == DM_CHROMA_IDX);
}
}
-#if IDCC_TPM_JEM
+#if JVET_V0130_INTRA_TMP
bool PU::isTmp(const PredictionUnit& pu, const ChannelType& chType)
{
- return (chType == CHANNEL_TYPE_LUMA && pu.cu->TmpFlag);
+ return (chType == CHANNEL_TYPE_LUMA && pu.cu->tmpFlag);
}
#endif
bool PU::isDMChromaMIP(const PredictionUnit &pu)
@@ -961,7 +961,7 @@ bool PU::isDMChromaMIP(const PredictionUnit &pu)
uint32_t PU::getIntraDirLuma( const PredictionUnit &pu )
{
-#if IDCC_TPM_JEM
+#if INTRA_TEMPLATE_MATCHING
if (isMIP(pu) || isTmp(pu))
#else
if (isMIP(pu))
@@ -4984,8 +4984,8 @@ bool CU::isMTSAllowed(const CodingUnit &cu, const ComponentID compID)
mtsAllowed &= cuWidth <= maxSize && cuHeight <= maxSize;
mtsAllowed &= !cu.ispMode;
mtsAllowed &= !cu.sbtInfo;
-#if IDCC_TMP_ImplicitMTS
- mtsAllowed &= !cu.TmpFlag;
+#if JVET_V0130_INTRA_TMP
+ mtsAllowed &= !cu.tmpFlag;
#endif
mtsAllowed &= !(cu.bdpcmMode && cuWidth <= tsMaxSize && cuHeight <= tsMaxSize);
return mtsAllowed;
@@ -5321,8 +5321,8 @@ bool allowLfnstWithMip(const Size& block)
}
return false;
}
-#if IDCC_TPM_JEM
-bool allowLfnstWithTpm()
+#if JVET_V0130_INTRA_TMP
+bool allowLfnstWithTmp()
{
return true;
}
diff --git a/source/Lib/CommonLib/UnitTools.h b/source/Lib/CommonLib/UnitTools.h
index a9c47df9814bcf6a67ce72f5f47f0f8dae446d50..ab2c96231fb0bfa826aa208253437a3eac1656c5 100644
--- a/source/Lib/CommonLib/UnitTools.h
+++ b/source/Lib/CommonLib/UnitTools.h
@@ -70,8 +70,8 @@ namespace CU
bool isSameSubPic (const CodingUnit &cu, const CodingUnit &cu2);
bool isLastSubCUOfCtu (const CodingUnit &cu);
uint32_t getCtuAddr (const CodingUnit &cu);
-#if IDCC_TPM_JEM
- Position getCtuXYAddr(const CodingUnit& cu);
+#if JVET_V0130_INTRA_TMP
+ Position getCtuXYAddr (const CodingUnit& cu);
#endif
int predictQP (const CodingUnit& cu, const int prevQP );
@@ -141,7 +141,7 @@ namespace PU
int getIntraMPMs(const PredictionUnit &pu, unsigned *mpm, const ChannelType &channelType = CHANNEL_TYPE_LUMA);
#endif
bool isMIP (const PredictionUnit &pu, const ChannelType &chType = CHANNEL_TYPE_LUMA);
-#if IDCC_TPM_JEM
+#if JVET_V0130_INTRA_TMP
bool isTmp(const PredictionUnit& pu, const ChannelType& chType = CHANNEL_TYPE_LUMA);
#endif
bool isDMChromaMIP (const PredictionUnit &pu);
@@ -274,8 +274,8 @@ uint32_t getCtuAddr (const Position& pos, const PreCalcValues &pcv);
int getNumModesMip (const Size& block);
int getMipSizeId (const Size& block);
bool allowLfnstWithMip(const Size& block);
-#if IDCC_TPM_JEM
-bool allowLfnstWithTpm();
+#if JVET_V0130_INTRA_TMP
+bool allowLfnstWithTmp();
#endif
template<typename T, size_t N>
diff --git a/source/Lib/CommonLib/dtrace_blockstatistics.cpp b/source/Lib/CommonLib/dtrace_blockstatistics.cpp
index a5e071fbcb35f0572326c2367eaedf59a7de3cd4..2ec1079052779c5c57c9d79f01fa7c9e6bde5b2c 100644
--- a/source/Lib/CommonLib/dtrace_blockstatistics.cpp
+++ b/source/Lib/CommonLib/dtrace_blockstatistics.cpp
@@ -884,8 +884,8 @@ void writeAllData(const CodingStructure& cs, const UnitArea& ctuArea)
if(chType == CHANNEL_TYPE_LUMA)
{
DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, cu, GetBlockStatisticName(BlockStatistic::MIPFlag), cu.mipFlag);
-#if IDCC_TPM_JEM
- DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, cu, GetBlockStatisticName(BlockStatistic::TmpFlag), cu.TmpFlag);
+#if JVET_V0130_INTRA_TMP
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, cu, GetBlockStatisticName(BlockStatistic::TmpFlag), cu.tmpFlag);
#endif
DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, cu, GetBlockStatisticName(BlockStatistic::ISPMode), cu.ispMode);
}
diff --git a/source/Lib/CommonLib/dtrace_blockstatistics.h b/source/Lib/CommonLib/dtrace_blockstatistics.h
index a416227b6dbfe3abf96f6b3f34b7ae77b0fcd397..fe3032dd9323b1abe8443f40833b807eaf6ac2a9 100644
--- a/source/Lib/CommonLib/dtrace_blockstatistics.h
+++ b/source/Lib/CommonLib/dtrace_blockstatistics.h
@@ -78,7 +78,7 @@ enum class BlockStatistic {
Chroma_IntraMode,
MultiRefIdx,
MIPFlag,
-#if IDCC_TPM_JEM
+#if JVET_V0130_INTRA_TMP
TmpFlag,
#endif
ISPMode,
@@ -173,7 +173,7 @@ static const std::map<BlockStatistic, std::tuple<std::string, BlockStatisticType
{ BlockStatistic::JointCbCr, std::tuple<std::string, BlockStatisticType, std::string>{"JointCbCr", BlockStatisticType::Flag, ""}},
{ BlockStatistic::MIPFlag, std::tuple<std::string, BlockStatisticType, std::string>{"MIPFlag", BlockStatisticType::Flag, ""}},
-#if IDCC_TPM_JEM
+#if JVET_V0130_INTRA_TMP
{ BlockStatistic::TmpFlag, std::tuple<std::string, BlockStatisticType, std::string>{"TmpFlag", BlockStatisticType::Flag, ""}},
#endif
{ BlockStatistic::ISPMode, std::tuple<std::string, BlockStatisticType, std::string>{"ISPMode", BlockStatisticType::Integer, "[0, " + std::to_string(NUM_INTRA_SUBPARTITIONS_MODES) + "]"}},
diff --git a/source/Lib/CommonLib/version.h b/source/Lib/CommonLib/version.h
index 250f2cd9fd501badc15831fc0c5ffc13aeb723e1..70ad90a766d6e3ce8d2c18d989ca90082b7ce57a 100644
--- a/source/Lib/CommonLib/version.h
+++ b/source/Lib/CommonLib/version.h
@@ -1,3 +1,4 @@
#if ! defined( VTM_VERSION )
#define VTM_VERSION "10.0"
+#define ECM_VERSION "0.0"
#endif
diff --git a/source/Lib/CommonLib/x86/BilateralFilterX86.h b/source/Lib/CommonLib/x86/BilateralFilterX86.h
new file mode 100644
index 0000000000000000000000000000000000000000..c0ff50f957dc3f99f8e0c257b9402817955fc418
--- /dev/null
+++ b/source/Lib/CommonLib/x86/BilateralFilterX86.h
@@ -0,0 +1,311 @@
+/* The copyright in this software is being made available under the BSD
+ * License, included below. This software may be subject to other third party
+ * and contributor rights, including patent rights, and no such rights are
+ * granted under this license.
+ *
+ * Copyright (c) 2010-2021, ITU/ISO/IEC
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * * Redistributions of source code must retain the above copyright notice,
+ * this list of conditions and the following disclaimer.
+ * * Redistributions in binary form must reproduce the above copyright notice,
+ * this list of conditions and the following disclaimer in the documentation
+ * and/or other materials provided with the distribution.
+ * * Neither the name of the ITU/ISO/IEC nor the names of its contributors may
+ * be used to endorse or promote products derived from this software without
+ * specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS
+ * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
+ * THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#pragma once
+
+#include "CommonDefX86.h"
+#include "../BilateralFilter.h"
+
+#ifdef TARGET_SIMD_X86
+#if defined _MSC_VER
+#include <tmmintrin.h>
+#else
+#include <x86intrin.h>
+#endif
+
+#if ENABLE_SIMD_BILATERAL_FILTER
+template<X86_VEXT vext>
+void BilateralFilter::simdFilterDiamond5x5(uint32_t uiWidth, uint32_t uiHeight, int16_t block[], int16_t blkFilt[], const ClpRng& clpRng, Pel* recPtr, int recStride, int iWidthExtSIMD, int bfac, int bif_round_add, int bif_round_shift, bool isRDO, const char* LUTrowPtr )
+{
+ if( uiWidth < 4 || ( uiWidth < 8 && isRDO ) )
+ {
+ return blockBilateralFilterDiamond5x5(uiWidth, uiHeight, block, blkFilt, clpRng, recPtr, recStride, iWidthExtSIMD, bfac, bif_round_add, bif_round_shift, isRDO, LUTrowPtr );
+ }
+
+ int pad = 2;
+ int padwidth = iWidthExtSIMD;
+
+ __m128i center, left, right, up, down, lu, ld, ru, rd, diffabs, four, fifteen, lut, acc, temp, round_add, clipmin, clipmax, inputVals;
+ __m128i ll, rr, uu, dd;
+
+ four = _mm_set1_epi16(4);
+ fifteen = _mm_set1_epi16(15);
+ round_add = _mm_set1_epi16(bif_round_add);
+ clipmin = _mm_set1_epi16(clpRng.min);
+ clipmax = _mm_set1_epi16(clpRng.max);
+
+ lut = _mm_loadu_si128((__m128i*)(LUTrowPtr));
+ acc = _mm_set1_epi32(0);
+
+ // Copy back parameters
+ Pel *tempBlockPtr = (short*)blkFilt + (((padwidth+4) << 1) + 2);
+ int tempBlockStride = padwidth+4;
+
+ for (int col = 0; col < uiWidth; col += 8)
+ {
+ for (int row = 0; row < uiHeight; row++)
+ {
+ acc = _mm_set1_epi32(0);
+ int16_t *point = &block[(row + pad)*padwidth + pad + col];
+
+ center = _mm_loadu_si128((__m128i*)(point));
+
+ //load neighbours
+ left = _mm_loadu_si128((__m128i*)(point - 1));
+ right = _mm_loadu_si128((__m128i*)(point + 1));
+ up = _mm_loadu_si128((__m128i*)(point - padwidth));
+ down = _mm_loadu_si128((__m128i*)(point + padwidth));
+
+ lu = _mm_loadu_si128((__m128i*)(point - 1 - padwidth));
+ ld = _mm_loadu_si128((__m128i*)(point - 1 + padwidth));
+ ru = _mm_loadu_si128((__m128i*)(point + 1 - padwidth));
+ rd = _mm_loadu_si128((__m128i*)(point + 1 + padwidth));
+
+ ll = _mm_loadu_si128((__m128i*)(point - 2));
+ rr = _mm_loadu_si128((__m128i*)(point + 2));
+ uu = _mm_loadu_si128((__m128i*)(point - 2*padwidth));
+ dd = _mm_loadu_si128((__m128i*)(point + 2*padwidth));
+
+ //calculate diffs
+ left = _mm_sub_epi16(left, center);
+ right = _mm_sub_epi16(right, center);
+ up = _mm_sub_epi16(up, center);
+ down = _mm_sub_epi16(down, center);
+
+ lu = _mm_sub_epi16(lu, center);
+ ld = _mm_sub_epi16(ld, center);
+ ru = _mm_sub_epi16(ru, center);
+ rd = _mm_sub_epi16(rd, center);
+
+ ll = _mm_sub_epi16(ll, center);
+ rr = _mm_sub_epi16(rr, center);
+ uu = _mm_sub_epi16(uu, center);
+ dd = _mm_sub_epi16(dd, center);
+
+ //LEFT!
+ //calculate abs
+ diffabs = _mm_abs_epi16(left); //abs
+ diffabs = _mm_add_epi16(diffabs, four); //+4
+ diffabs = _mm_srai_epi16(diffabs, 3); //>>3
+ diffabs = _mm_min_epi16(diffabs, fifteen); //min(x,15)
+ diffabs = _mm_packus_epi16(diffabs, diffabs); //convert to 8
+ diffabs = _mm_shuffle_epi8(lut, diffabs);//lut
+ diffabs = _mm_cvtepi8_epi16(diffabs);//back to 16-bit
+ diffabs = _mm_sign_epi16(diffabs, left);//fix sign!
+ acc = _mm_add_epi16(diffabs, acc); //add to acc
+ //RIGHT!
+ //calculate abs
+ diffabs = _mm_abs_epi16(right); //abs
+ diffabs = _mm_add_epi16(diffabs, four); //+4
+ diffabs = _mm_srai_epi16(diffabs, 3); //>>3
+ diffabs = _mm_min_epi16(diffabs, fifteen); //min(x,15)
+ diffabs = _mm_packus_epi16(diffabs, diffabs); //convert to 8
+ diffabs = _mm_shuffle_epi8(lut, diffabs);//lut
+ diffabs = _mm_cvtepi8_epi16(diffabs);//back to 16-bit
+ diffabs = _mm_sign_epi16(diffabs, right);//fix sign!
+ acc = _mm_add_epi16(diffabs, acc); //add to acc
+ //UP!
+ //calculate abs
+ diffabs = _mm_abs_epi16(up); //abs
+ diffabs = _mm_add_epi16(diffabs, four); //+4
+ diffabs = _mm_srai_epi16(diffabs, 3); //>>3
+ diffabs = _mm_min_epi16(diffabs, fifteen); //min(x,15)
+ diffabs = _mm_packus_epi16(diffabs, diffabs); //convert to 8
+ diffabs = _mm_shuffle_epi8(lut, diffabs);//lut
+ diffabs = _mm_cvtepi8_epi16(diffabs);//back to 16-bit
+ diffabs = _mm_sign_epi16(diffabs, up);//fix sign!
+ acc = _mm_add_epi16(diffabs, acc); //add to acc
+
+ //DOWN!
+ //calculate abs
+ diffabs = _mm_abs_epi16(down); //abs
+ diffabs = _mm_add_epi16(diffabs, four); //+4
+ diffabs = _mm_srai_epi16(diffabs, 3); //>>3
+ diffabs = _mm_min_epi16(diffabs, fifteen); //min(x,15)
+ diffabs = _mm_packus_epi16(diffabs, diffabs); //convert to 8
+ diffabs = _mm_shuffle_epi8(lut, diffabs);//lut
+ diffabs = _mm_cvtepi8_epi16(diffabs);//back to 16-bit
+ diffabs = _mm_sign_epi16(diffabs, down);//fix sign!
+ acc = _mm_add_epi16(diffabs, acc); //add to acc
+
+ //lu!
+ //calculate abs
+ diffabs = _mm_abs_epi16(lu); //abs
+ diffabs = _mm_add_epi16(diffabs, four); //+4
+ diffabs = _mm_srai_epi16(diffabs, 3); //>>3
+ diffabs = _mm_min_epi16(diffabs, fifteen); //min(x,15)
+ diffabs = _mm_packus_epi16(diffabs, diffabs); //convert to 8
+ diffabs = _mm_shuffle_epi8(lut, diffabs);//lut
+ diffabs = _mm_cvtepi8_epi16(diffabs);//back to 16-bit
+ diffabs = _mm_srai_epi16(diffabs, 1);//diagonal shift!
+ diffabs = _mm_sign_epi16(diffabs, lu);//fix sign!
+ acc = _mm_add_epi16(diffabs, acc); //add to acc
+ //ld!
+ //calculate abs
+ diffabs = _mm_abs_epi16(ld); //abs
+ diffabs = _mm_add_epi16(diffabs, four); //+4
+ diffabs = _mm_srai_epi16(diffabs, 3); //>>3
+ diffabs = _mm_min_epi16(diffabs, fifteen); //min(x,15)
+ diffabs = _mm_packus_epi16(diffabs, diffabs); //convert to 8
+ diffabs = _mm_shuffle_epi8(lut, diffabs);//lut
+ diffabs = _mm_cvtepi8_epi16(diffabs);//back to 16-bit
+ diffabs = _mm_srai_epi16(diffabs, 1);//diagonal shift!
+ diffabs = _mm_sign_epi16(diffabs, ld);//fix sign!
+ acc = _mm_add_epi16(diffabs, acc); //add to acc
+ //ru!
+ //calculate abs
+ diffabs = _mm_abs_epi16(ru); //abs
+ diffabs = _mm_add_epi16(diffabs, four); //+4
+ diffabs = _mm_srai_epi16(diffabs, 3); //>>3
+ diffabs = _mm_min_epi16(diffabs, fifteen); //min(x,15)
+ diffabs = _mm_packus_epi16(diffabs, diffabs); //convert to 8
+ diffabs = _mm_shuffle_epi8(lut, diffabs);//lut
+ diffabs = _mm_cvtepi8_epi16(diffabs);//back to 16-bit
+ diffabs = _mm_srai_epi16(diffabs, 1);//diagonal shift!
+ diffabs = _mm_sign_epi16(diffabs, ru);//fix sign!
+ acc = _mm_add_epi16(diffabs, acc); //add to acc
+ //rd!
+ //calculate abs
+ diffabs = _mm_abs_epi16(rd); //abs
+ diffabs = _mm_add_epi16(diffabs, four); //+4
+ diffabs = _mm_srai_epi16(diffabs, 3); //>>3
+ diffabs = _mm_min_epi16(diffabs, fifteen); //min(x,15)
+ diffabs = _mm_packus_epi16(diffabs, diffabs); //convert to 8
+ diffabs = _mm_shuffle_epi8(lut, diffabs);//lut
+ diffabs = _mm_cvtepi8_epi16(diffabs);//back to 16-bit
+ diffabs = _mm_srai_epi16(diffabs, 1);//diagonal shift!
+ diffabs = _mm_sign_epi16(diffabs, rd);//fix sign!
+ acc = _mm_add_epi16(diffabs, acc); //add to acc
+
+ //ll!
+ //calculate abs
+ diffabs = _mm_abs_epi16(ll); //abs
+ diffabs = _mm_add_epi16(diffabs, four); //+4
+ diffabs = _mm_srai_epi16(diffabs, 3); //>>3
+ diffabs = _mm_min_epi16(diffabs, fifteen); //min(x,15)
+ diffabs = _mm_packus_epi16(diffabs, diffabs); //convert to 8
+ diffabs = _mm_shuffle_epi8(lut, diffabs);//lut
+ diffabs = _mm_cvtepi8_epi16(diffabs);//back to 16-bit
+ diffabs = _mm_srai_epi16(diffabs, 1);//diagonal shift!
+ diffabs = _mm_sign_epi16(diffabs, ll);//fix sign!
+ acc = _mm_add_epi16(diffabs, acc); //add to acc
+ //rr!
+ //calculate abs
+ diffabs = _mm_abs_epi16(rr); //abs
+ diffabs = _mm_add_epi16(diffabs, four); //+4
+ diffabs = _mm_srai_epi16(diffabs, 3); //>>3
+ diffabs = _mm_min_epi16(diffabs, fifteen); //min(x,15)
+ diffabs = _mm_packus_epi16(diffabs, diffabs); //convert to 8
+ diffabs = _mm_shuffle_epi8(lut, diffabs);//lut
+ diffabs = _mm_cvtepi8_epi16(diffabs);//back to 16-bit
+ diffabs = _mm_srai_epi16(diffabs, 1);//diagonal shift!
+ diffabs = _mm_sign_epi16(diffabs, rr);//fix sign!
+ acc = _mm_add_epi16(diffabs, acc); //add to acc
+ //uu!
+ //calculate abs
+ diffabs = _mm_abs_epi16(uu); //abs
+ diffabs = _mm_add_epi16(diffabs, four); //+4
+ diffabs = _mm_srai_epi16(diffabs, 3); //>>3
+ diffabs = _mm_min_epi16(diffabs, fifteen); //min(x,15)
+ diffabs = _mm_packus_epi16(diffabs, diffabs); //convert to 8
+ diffabs = _mm_shuffle_epi8(lut, diffabs);//lut
+ diffabs = _mm_cvtepi8_epi16(diffabs);//back to 16-bit
+ diffabs = _mm_srai_epi16(diffabs, 1);//diagonal shift!
+ diffabs = _mm_sign_epi16(diffabs, uu);//fix sign!
+ acc = _mm_add_epi16(diffabs, acc); //add to acc
+ //dd!
+ //calculate abs
+ diffabs = _mm_abs_epi16(dd); //abs
+ diffabs = _mm_add_epi16(diffabs, four); //+4
+ diffabs = _mm_srai_epi16(diffabs, 3); //>>3
+ diffabs = _mm_min_epi16(diffabs, fifteen); //min(x,15)
+ diffabs = _mm_packus_epi16(diffabs, diffabs); //convert to 8
+ diffabs = _mm_shuffle_epi8(lut, diffabs);//lut
+ diffabs = _mm_cvtepi8_epi16(diffabs);//back to 16-bit
+ diffabs = _mm_srai_epi16(diffabs, 1);//diagonal shift!
+ diffabs = _mm_sign_epi16(diffabs, dd);//fix sign!
+ acc = _mm_add_epi16(diffabs, acc); //add to acc
+
+ if (bfac == 2)
+ {
+ acc = _mm_slli_epi16(acc, 1); // Shift left to get 2*
+ }
+ else if (bfac == 3)
+ {
+ temp = _mm_slli_epi16(acc, 1); // Multiply by two by shifting left
+ acc = _mm_add_epi16(acc, temp); // Add original value to get 3*
+ }
+
+ // Add 16 and shift 5
+ acc = _mm_add_epi16(acc, round_add);
+ acc = _mm_srai_epi16(acc, bif_round_shift);
+
+ // Instead we add our input values to the delta
+ if(isRDO)
+ {
+ acc = _mm_add_epi16(acc, center);
+ }
+ else
+ {
+ int16_t *recpoint = &recPtr[row * recStride + col];
+ inputVals = _mm_loadu_si128((__m128i*)(recpoint));
+ acc = _mm_add_epi16(acc, inputVals);
+ }
+
+ // Clip
+ acc = _mm_max_epi16(acc, clipmin);
+ acc = _mm_min_epi16(acc, clipmax);
+
+ _mm_store_si128((__m128i*)(blkFilt + (row + pad) * (padwidth + 4) + col + pad), acc);
+ }
+ }
+
+ // Copy back from tempbufFilter to recBuf
+ int onerow = uiWidth * sizeof(Pel);
+ for(uint32_t yy = 0; yy < uiHeight; yy++)
+ {
+ std::memcpy(recPtr, tempBlockPtr, onerow);
+ recPtr += recStride;
+ tempBlockPtr += tempBlockStride;
+ }
+}
+
+template <X86_VEXT vext>
+void BilateralFilter::_initBilateralFilterX86()
+{
+ m_bilateralFilterDiamond5x5 = simdFilterDiamond5x5<vext>;
+}
+
+template void BilateralFilter::_initBilateralFilterX86<SIMDX86>();
+#endif
+#endif // TARGET_SIMD_X86
diff --git a/source/Lib/CommonLib/x86/InitX86.cpp b/source/Lib/CommonLib/x86/InitX86.cpp
index 9511379d34cd6fb40cc6a9159316acba912bfa7e..b8e30fbdd77ecd10595385509d689aeff69652ba 100644
--- a/source/Lib/CommonLib/x86/InitX86.cpp
+++ b/source/Lib/CommonLib/x86/InitX86.cpp
@@ -50,9 +50,12 @@
#include "CommonLib/IbcHashMap.h"
-#if ENABLE_SIMD_SIGN_PREDICTION || TRANSFORM_SIMD_OPT
+#if ENABLE_SIMD_SIGN_PREDICTION || TRANSFORM_SIMD_OPT || ENABLE_SIMD_TMP
#include "CommonLib/TrQuant.h"
#endif
+#if ENABLE_SIMD_BILATERAL_FILTER
+#include "CommonLib/BilateralFilter.h"
+#endif
#ifdef TARGET_SIMD_X86
@@ -190,7 +193,7 @@ void IbcHashMap::initIbcHashMapX86()
}
#endif
-#if ENABLE_SIMD_SIGN_PREDICTION || TRANSFORM_SIMD_OPT
+#if ENABLE_SIMD_SIGN_PREDICTION || TRANSFORM_SIMD_OPT || ENABLE_SIMD_TMP
void TrQuant::initTrQuantX86()
{
auto vext = read_x86_extension_flags();
@@ -213,5 +216,28 @@ void TrQuant::initTrQuantX86()
}
#endif
+#if ENABLE_SIMD_BILATERAL_FILTER
+void BilateralFilter::initBilateralFilterX86()
+{
+ auto vext = read_x86_extension_flags();
+ switch( vext )
+ {
+ case AVX512:
+ case AVX2:
+ _initBilateralFilterX86<AVX2>();
+ break;
+ case AVX:
+ _initBilateralFilterX86<AVX>();
+ break;
+ case SSE42:
+ case SSE41:
+ _initBilateralFilterX86<SSE41>();
+ break;
+ default:
+ break;
+ }
+}
+#endif
+
#endif
diff --git a/source/Lib/CommonLib/x86/TrQuantX86.h b/source/Lib/CommonLib/x86/TrQuantX86.h
index 5239521da29f73107d0f9773c23f0da7039316f2..5f7238fc60ea6aa9385a797399af4a7ccb3b524f 100644
--- a/source/Lib/CommonLib/x86/TrQuantX86.h
+++ b/source/Lib/CommonLib/x86/TrQuantX86.h
@@ -410,7 +410,129 @@ uint32_t computeSAD_SIMD( const Pel* ref, const Pel* cur, const int size )
}
#endif
-#if ENABLE_SIMD_SIGN_PREDICTION || TRANSFORM_SIMD_OPT
+#if ENABLE_SIMD_TMP
+template< X86_VEXT vext >
+int calcTemplateDiffSIMD( Pel* ref, unsigned int uiStride, Pel** tarPatch, unsigned int uiPatchWidth, unsigned int uiPatchHeight, int iMax )
+{
+ int iDiffSum = 0;
+ int iY;
+ Pel* refPatchRow = ref - TMP_TEMPLATE_SIZE * uiStride - TMP_TEMPLATE_SIZE;
+ Pel* tarPatchRow;
+ uint32_t uiSum;
+
+ // horizontal difference
+ for( iY = 0; iY < TMP_TEMPLATE_SIZE; iY++ )
+ {
+ tarPatchRow = tarPatch[iY];
+ const short* pSrc1 = ( const short* ) tarPatchRow;
+ const short* pSrc2 = ( const short* ) refPatchRow;
+
+ // SIMD difference
+ //int iRows = uiPatchHeight;
+ int iCols = uiPatchWidth;
+ if( (iCols & 7) == 0 )
+ {
+ // Do with step of 8
+ __m128i vzero = _mm_setzero_si128();
+ __m128i vsum32 = vzero;
+ //for (int iY = 0; iY < iRows; iY += iSubStep)
+ {
+ __m128i vsum16 = vzero;
+ for( int iX = 0; iX < iCols; iX += 8 )
+ {
+ __m128i vsrc1 = _mm_loadu_si128( (const __m128i*)(&pSrc1[iX]) );
+ __m128i vsrc2 = _mm_lddqu_si128( (const __m128i*)(&pSrc2[iX]) );
+ vsum16 = _mm_add_epi16( vsum16, _mm_abs_epi16( _mm_sub_epi16( vsrc1, vsrc2 ) ) );
+ }
+ __m128i vsumtemp = _mm_add_epi32( _mm_unpacklo_epi16( vsum16, vzero ), _mm_unpackhi_epi16( vsum16, vzero ) );
+ vsum32 = _mm_add_epi32( vsum32, vsumtemp );
+ //pSrc1 += iStrideSrc1;
+ //pSrc2 += iStrideSrc2;
+ }
+ vsum32 = _mm_add_epi32( vsum32, _mm_shuffle_epi32( vsum32, 0x4e ) ); // 01001110
+ vsum32 = _mm_add_epi32( vsum32, _mm_shuffle_epi32( vsum32, 0xb1 ) ); // 10110001
+ uiSum = _mm_cvtsi128_si32( vsum32 );
+ }
+ else
+ {
+ // Do with step of 4
+ __m128i vzero = _mm_setzero_si128();
+ __m128i vsum32 = vzero;
+ //for (int iY = 0; iY < iRows; iY += iSubStep)
+ {
+ __m128i vsum16 = vzero;
+ for( int iX = 0; iX < iCols; iX += 4 )
+ {
+ __m128i vsrc1 = _mm_loadl_epi64( (const __m128i*) & pSrc1[iX] );
+ __m128i vsrc2 = _mm_loadl_epi64( (const __m128i*) & pSrc2[iX] );
+ vsum16 = _mm_add_epi16( vsum16, _mm_abs_epi16( _mm_sub_epi16( vsrc1, vsrc2 ) ) );
+ }
+ __m128i vsumtemp = _mm_add_epi32( _mm_unpacklo_epi16( vsum16, vzero ), _mm_unpackhi_epi16( vsum16, vzero ) );
+ vsum32 = _mm_add_epi32( vsum32, vsumtemp );
+ //pSrc1 += iStrideSrc1;
+ //pSrc2 += iStrideSrc2;
+ }
+ vsum32 = _mm_add_epi32( vsum32, _mm_shuffle_epi32( vsum32, 0x4e ) ); // 01001110
+ vsum32 = _mm_add_epi32( vsum32, _mm_shuffle_epi32( vsum32, 0xb1 ) ); // 10110001
+ uiSum = _mm_cvtsi128_si32( vsum32 );
+ }
+ iDiffSum += uiSum;
+
+ if( iDiffSum > iMax ) //for speeding up
+ {
+ return iDiffSum;
+ }
+ // update location
+ refPatchRow += uiStride;
+ }
+
+ // vertical difference
+ int iCols = TMP_TEMPLATE_SIZE;
+
+ for( iY = TMP_TEMPLATE_SIZE; iY < uiPatchHeight; iY++ )
+ {
+ tarPatchRow = tarPatch[iY];
+ const short* pSrc1 = ( const short* ) tarPatchRow;
+ const short* pSrc2 = ( const short* ) refPatchRow;
+
+ // SIMD difference
+
+ // Do with step of 4
+ __m128i vzero = _mm_setzero_si128();
+ __m128i vsum32 = vzero;
+ //for (int iY = 0; iY < iRows; iY += iSubStep)
+ {
+ __m128i vsum16 = vzero;
+ for( int iX = 0; iX < iCols; iX += 4 )
+ {
+ __m128i vsrc1 = _mm_loadl_epi64( (const __m128i*) & pSrc1[iX] );
+ __m128i vsrc2 = _mm_loadl_epi64( (const __m128i*) & pSrc2[iX] );
+ vsum16 = _mm_add_epi16( vsum16, _mm_abs_epi16( _mm_sub_epi16( vsrc1, vsrc2 ) ) );
+ }
+ __m128i vsumtemp = _mm_add_epi32( _mm_unpacklo_epi16( vsum16, vzero ), _mm_unpackhi_epi16( vsum16, vzero ) );
+ vsum32 = _mm_add_epi32( vsum32, vsumtemp );
+ //pSrc1 += iStrideSrc1;
+ //pSrc2 += iStrideSrc2;
+ }
+ vsum32 = _mm_add_epi32( vsum32, _mm_shuffle_epi32( vsum32, 0x4e ) ); // 01001110
+ vsum32 = _mm_add_epi32( vsum32, _mm_shuffle_epi32( vsum32, 0xb1 ) ); // 10110001
+ uiSum = _mm_cvtsi128_si32( vsum32 );
+
+ iDiffSum += uiSum;
+
+ if( iDiffSum > iMax ) //for speeding up
+ {
+ return iDiffSum;
+ }
+ // update location
+ refPatchRow += uiStride;
+ }
+
+ return iDiffSum;
+}
+#endif
+
+#if ENABLE_SIMD_SIGN_PREDICTION || TRANSFORM_SIMD_OPT || ENABLE_SIMD_TMP
template <X86_VEXT vext>
void TrQuant::_initTrQuantX86()
{
@@ -545,6 +667,10 @@ void TrQuant::_initTrQuantX86()
fastInvTrans[2][5] = fastInverseTransform_SIMD<DST7, 64>;
#endif
#endif
+
+#if ENABLE_SIMD_TMP
+ m_calcTemplateDiff = calcTemplateDiffSIMD<vext>;
+#endif
}
template void TrQuant::_initTrQuantX86<SIMDX86>();
diff --git a/source/Lib/CommonLib/x86/avx/BilateralFilter_avx.cpp b/source/Lib/CommonLib/x86/avx/BilateralFilter_avx.cpp
new file mode 100644
index 0000000000000000000000000000000000000000..bb696a1fef690da0fc3fc6830bdcb7189a3a11a3
--- /dev/null
+++ b/source/Lib/CommonLib/x86/avx/BilateralFilter_avx.cpp
@@ -0,0 +1 @@
+#include "../BilateralFilterX86.h"
diff --git a/source/Lib/CommonLib/x86/avx2/BilateralFilter_avx2.cpp b/source/Lib/CommonLib/x86/avx2/BilateralFilter_avx2.cpp
new file mode 100644
index 0000000000000000000000000000000000000000..bb696a1fef690da0fc3fc6830bdcb7189a3a11a3
--- /dev/null
+++ b/source/Lib/CommonLib/x86/avx2/BilateralFilter_avx2.cpp
@@ -0,0 +1 @@
+#include "../BilateralFilterX86.h"
diff --git a/source/Lib/CommonLib/x86/sse41/BilateralFilter_sse41.cpp b/source/Lib/CommonLib/x86/sse41/BilateralFilter_sse41.cpp
new file mode 100644
index 0000000000000000000000000000000000000000..bb696a1fef690da0fc3fc6830bdcb7189a3a11a3
--- /dev/null
+++ b/source/Lib/CommonLib/x86/sse41/BilateralFilter_sse41.cpp
@@ -0,0 +1 @@
+#include "../BilateralFilterX86.h"
diff --git a/source/Lib/CommonLib/x86/sse42/BilateralFilater_sse42.cpp b/source/Lib/CommonLib/x86/sse42/BilateralFilater_sse42.cpp
new file mode 100644
index 0000000000000000000000000000000000000000..bb696a1fef690da0fc3fc6830bdcb7189a3a11a3
--- /dev/null
+++ b/source/Lib/CommonLib/x86/sse42/BilateralFilater_sse42.cpp
@@ -0,0 +1 @@
+#include "../BilateralFilterX86.h"
diff --git a/source/Lib/DecoderLib/CABACReader.cpp b/source/Lib/DecoderLib/CABACReader.cpp
index 10f7f72b23ee698e39a7cea978eaf391eabd11e2..0539150a948c6a7aab9dad25b7c846d7787f8e4a 100644
--- a/source/Lib/DecoderLib/CABACReader.cpp
+++ b/source/Lib/DecoderLib/CABACReader.cpp
@@ -1606,16 +1606,20 @@ void CABACReader::intra_luma_pred_modes( CodingUnit &cu )
cu.firstPU->intraDir[0] = cu.bdpcmMode == 2? VER_IDX : HOR_IDX;
return;
}
-#if IDCC_TPM_JEM
+#if JVET_V0130_INTRA_TMP
int TMP_MaxSize=cu.cs->sps->getIntraTMPMaxSize();
if (cu.lwidth() <= TMP_MaxSize && cu.lheight() <= TMP_MaxSize)
{
- Tmp_Flag(cu);
- if (cu.TmpFlag)
- return;
+ tmp_flag(cu);
+ if( cu.tmpFlag )
+ {
+ return;
+ }
}
else
- cu.TmpFlag = 0;
+ {
+ cu.tmpFlag = 0;
+ }
#endif
mip_flag(cu);
if (cu.mipFlag)
@@ -3862,10 +3866,10 @@ void CABACReader::residual_lfnst_mode( CodingUnit& cu, CUCtx& cuCtx )
int chIdx = cu.isSepTree() && cu.chType == CHANNEL_TYPE_CHROMA ? 1 : 0;
#endif
if ((cu.ispMode && !CU::canUseLfnstWithISP(cu, cu.chType))
-#if IDCC_TPM_JEM
- || (cu.cs->sps->getUseLFNST() && CU::isIntra(cu) && ((cu.mipFlag && !allowLfnstWithMip(cu.firstPU->lumaSize())) || (cu.TmpFlag && !allowLfnstWithTpm())))
+#if JVET_V0130_INTRA_TMP
+ || (cu.cs->sps->getUseLFNST() && CU::isIntra(cu) && ((cu.mipFlag && !allowLfnstWithMip(cu.firstPU->lumaSize())) || (cu.tmpFlag && !allowLfnstWithTmp())))
#else
- || (cu.cs->sps->getUseLFNST() && CU::isIntra(cu) && cu.mipFlag && !allowLfnstWithMip(cu.firstPU->lumaSize()))
+ || (cu.cs->sps->getUseLFNST() && CU::isIntra(cu) && cu.mipFlag && !allowLfnstWithMip(cu.firstPU->lumaSize()))
#endif
#if INTRA_RM_SMALL_BLOCK_SIZE_CONSTRAINTS
|| (CS::isDualITree(*cu.cs) && cu.chType == CHANNEL_TYPE_CHROMA && std::min(cu.blocks[1].width, cu.blocks[1].height) < 4)
@@ -4506,8 +4510,8 @@ unsigned CABACReader::code_unary_fixed( unsigned ctxId, unsigned unary_max, unsi
}
return idx;
}
-#if IDCC_TPM_JEM
-void CABACReader::Tmp_Flag(CodingUnit& cu)
+#if JVET_V0130_INTRA_TMP
+void CABACReader::tmp_flag(CodingUnit& cu)
{
RExt__DECODER_DEBUG_BIT_STATISTICS_CREATE_SET(STATS__CABAC_BITS__OTHER);
@@ -4518,13 +4522,13 @@ void CABACReader::Tmp_Flag(CodingUnit& cu)
if( !cu.cs->sps->getUseIntraTMP() )
{
- cu.TmpFlag = false;
+ cu.tmpFlag = false;
return;
}
unsigned ctxId = DeriveCtx::CtxTmpFlag(cu);
- cu.TmpFlag = m_BinDecoder.decodeBin(Ctx::TmpFlag(ctxId));
- DTRACE(g_trace_ctx, D_SYNTAX, "Tmp_Flag() pos=(%d,%d) mode=%d\n", cu.lumaPos().x, cu.lumaPos().y, cu.TmpFlag ? 1 : 0);
+ cu.tmpFlag = m_BinDecoder.decodeBin(Ctx::TmpFlag(ctxId));
+ DTRACE(g_trace_ctx, D_SYNTAX, "tmp_flag() pos=(%d,%d) mode=%d\n", cu.lumaPos().x, cu.lumaPos().y, cu.tmpFlag ? 1 : 0);
}
#endif
void CABACReader::mip_flag( CodingUnit& cu )
diff --git a/source/Lib/DecoderLib/CABACReader.h b/source/Lib/DecoderLib/CABACReader.h
index 5513be42e0602be439420806bd6e1531a62a82cb..59607c833b83baa691603e354f34bc03f1ea175e 100644
--- a/source/Lib/DecoderLib/CABACReader.h
+++ b/source/Lib/DecoderLib/CABACReader.h
@@ -107,8 +107,8 @@ public:
void adaptive_color_transform(CodingUnit& cu);
void sbt_mode ( CodingUnit& cu );
void end_of_ctu ( CodingUnit& cu, CUCtx& cuCtx );
-#if IDCC_TPM_JEM
- void Tmp_Flag(CodingUnit& cu);
+#if JVET_V0130_INTRA_TMP
+ void tmp_flag ( CodingUnit& cu );
#endif
void mip_flag ( CodingUnit& cu );
void mip_pred_modes ( CodingUnit& cu );
diff --git a/source/Lib/DecoderLib/DecCu.cpp b/source/Lib/DecoderLib/DecCu.cpp
index b20d81b6efc70f8061829531de46d87bc5ba5fec..19f5affab59e9e932206a0c7ec8a21a8304935e8 100644
--- a/source/Lib/DecoderLib/DecCu.cpp
+++ b/source/Lib/DecoderLib/DecCu.cpp
@@ -307,7 +307,7 @@ void DecCu::xIntraRecBlk( TransformUnit& tu, const ComponentID compID )
}
else
{
-#if IDCC_TPM_JEM
+#if JVET_V0130_INTRA_TMP
if (PU::isTmp(pu, chType))
{
int foundCandiNum;
@@ -512,11 +512,11 @@ void DecCu::xIntraRecACTBlk(TransformUnit& tu)
PelBuf piPred = cs.getPredBuf(area);
m_pcIntraPred->initIntraPatternChType(*tu.cu, area);
-#if IDCC_TPM_JEM
+#if JVET_V0130_INTRA_TMP
if (PU::isTmp(pu, chType))
{
int foundCandiNum;
- const unsigned int uiStride = cs.picture->getRecoBuf(COMPONENT_Y).stride;
+ const unsigned int uiStride = cs.picture->getRecoBuf(COMPONENT_Y).stride;
m_pcTrQuant->getTargetTemplate(tu.cu, pu.lwidth(), pu.lheight());
m_pcTrQuant->candidateSearchIntra(tu.cu, pu.lwidth(), pu.lheight());
m_pcTrQuant->generateTMPrediction(piPred.buf, uiStride, pu.lwidth(), pu.lheight(), foundCandiNum);
diff --git a/source/Lib/DecoderLib/VLCReader.cpp b/source/Lib/DecoderLib/VLCReader.cpp
index d8a529556478f0dfcd2abd49c7bc895579a9e39b..b41bffd482821ebc8174ffe69f8abc181fe6215b 100644
--- a/source/Lib/DecoderLib/VLCReader.cpp
+++ b/source/Lib/DecoderLib/VLCReader.cpp
@@ -2260,7 +2260,7 @@ void HLSyntaxReader::parseSPS(SPS* pcSPS)
#if ENABLE_DIMD
READ_FLAG(uiCode, "sps_dimd_enabled_flag"); pcSPS->setUseDimd(uiCode != 0);
#endif
-#if IDCC_TPM_JEM
+#if JVET_V0130_INTRA_TMP
READ_FLAG(uiCode, "sps_intraTMP_enabled_flag"); pcSPS->setUseIntraTMP( uiCode != 0 );
if(pcSPS->getUseIntraTMP())
{
diff --git a/source/Lib/EncoderLib/CABACWriter.cpp b/source/Lib/EncoderLib/CABACWriter.cpp
index 20225abd53984d871c9e6aedd1a193a17597ee7f..391f28a2fbac914005f1b5bf19360e80dc8806a9 100644
--- a/source/Lib/EncoderLib/CABACWriter.cpp
+++ b/source/Lib/EncoderLib/CABACWriter.cpp
@@ -1200,13 +1200,15 @@ void CABACWriter::intra_luma_pred_modes( const CodingUnit& cu )
cu.firstPU->intraDir[0] = cu.bdpcmMode == 2? VER_IDX : HOR_IDX;
return;
}
-#if IDCC_TPM_JEM
+#if JVET_V0130_INTRA_TMP
int TMP_MaxSize=cu.cs->sps->getIntraTMPMaxSize();
if (cu.lwidth() <= TMP_MaxSize && cu.lheight() <= TMP_MaxSize)
{
- Tmp_Flag(cu);
- if (cu.TmpFlag)
- return;
+ tmp_flag(cu);
+ if( cu.tmpFlag )
+ {
+ return;
+ }
}
#endif
mip_flag(cu);
@@ -1389,15 +1391,17 @@ void CABACWriter::intra_luma_pred_mode( const PredictionUnit& pu )
{
if( pu.cu->bdpcmMode ) return;
-#if IDCC_TPM_JEM
+#if JVET_V0130_INTRA_TMP
// check if sufficient search range is available
//bool bCheck = pu.cu->
int TMP_MaxSize=pu.cu->cs->sps->getIntraTMPMaxSize();
if (pu.cu->lwidth() <= TMP_MaxSize && pu.cu->lheight() <= TMP_MaxSize)
{
- Tmp_Flag(*pu.cu);
- if (pu.cu->TmpFlag)
- return;
+ tmp_flag(*pu.cu);
+ if( pu.cu->tmpFlag )
+ {
+ return;
+ }
}
#endif
mip_flag(*pu.cu);
@@ -3592,10 +3596,10 @@ void CABACWriter::residual_lfnst_mode( const CodingUnit& cu, CUCtx& cuCtx )
int chIdx = cu.isSepTree() && cu.chType == CHANNEL_TYPE_CHROMA ? 1 : 0;
#endif
if( ( cu.ispMode && !CU::canUseLfnstWithISP( cu, cu.chType ) ) ||
-#if IDCC_TPM_JEM
- (cu.cs->sps->getUseLFNST() && CU::isIntra(cu) && ((cu.mipFlag && !allowLfnstWithMip(cu.firstPU->lumaSize())) || (cu.TmpFlag && !allowLfnstWithTpm()))) ||
+#if JVET_V0130_INTRA_TMP
+ (cu.cs->sps->getUseLFNST() && CU::isIntra(cu) && ((cu.mipFlag && !allowLfnstWithMip(cu.firstPU->lumaSize())) || (cu.tmpFlag && !allowLfnstWithTmp()))) ||
#else
- (cu.cs->sps->getUseLFNST() && CU::isIntra(cu) && cu.mipFlag && !allowLfnstWithMip(cu.firstPU->lumaSize())) ||
+ (cu.cs->sps->getUseLFNST() && CU::isIntra(cu) && cu.mipFlag && !allowLfnstWithMip(cu.firstPU->lumaSize())) ||
#endif
#if INTRA_RM_SMALL_BLOCK_SIZE_CONSTRAINTS
(CS::isDualITree(*cu.cs) && cu.chType == CHANNEL_TYPE_CHROMA && std::min(cu.blocks[1].width, cu.blocks[1].height) < 4)
@@ -4222,8 +4226,8 @@ void CABACWriter::code_unary_fixed( unsigned symbol, unsigned ctxId, unsigned un
}
}
-#if IDCC_TPM_JEM
-void CABACWriter::Tmp_Flag(const CodingUnit& cu)
+#if JVET_V0130_INTRA_TMP
+void CABACWriter::tmp_flag(const CodingUnit& cu)
{
if (!cu.Y().valid())
{
@@ -4236,8 +4240,8 @@ void CABACWriter::Tmp_Flag(const CodingUnit& cu)
}
unsigned ctxId = DeriveCtx::CtxTmpFlag(cu);
- m_BinEncoder.encodeBin(cu.TmpFlag, Ctx::TmpFlag(ctxId));
- DTRACE(g_trace_ctx, D_SYNTAX, "Tmp_Flag() pos=(%d,%d) mode=%d\n", cu.lumaPos().x, cu.lumaPos().y, cu.TmpFlag ? 1 : 0);
+ m_BinEncoder.encodeBin(cu.tmpFlag, Ctx::TmpFlag(ctxId));
+ DTRACE(g_trace_ctx, D_SYNTAX, "tmp_flag() pos=(%d,%d) mode=%d\n", cu.lumaPos().x, cu.lumaPos().y, cu.tmpFlag ? 1 : 0);
}
#endif
diff --git a/source/Lib/EncoderLib/CABACWriter.h b/source/Lib/EncoderLib/CABACWriter.h
index 6fd5d20e5c5551d20001297e47eae3fdb0fd6643..afbf6ed7c1a280ea67cfcaadaabca11236e8cdb6 100644
--- a/source/Lib/EncoderLib/CABACWriter.h
+++ b/source/Lib/EncoderLib/CABACWriter.h
@@ -116,8 +116,8 @@ public:
void adaptive_color_transform(const CodingUnit& cu);
void sbt_mode ( const CodingUnit& cu );
void end_of_ctu ( const CodingUnit& cu, CUCtx& cuCtx );
-#if IDCC_TPM_JEM
- void Tmp_Flag(const CodingUnit& cu);
+#if JVET_V0130_INTRA_TMP
+ void tmp_flag ( const CodingUnit& cu );
#endif
void mip_flag ( const CodingUnit& cu );
void mip_pred_modes ( const CodingUnit& cu );
diff --git a/source/Lib/EncoderLib/EncCfg.h b/source/Lib/EncoderLib/EncCfg.h
index 55d4aaef581fad47017085d9a45b98f31ef67cf5..a0740d6c0da4a2a4250e539130baa9fa9b919772 100644
--- a/source/Lib/EncoderLib/EncCfg.h
+++ b/source/Lib/EncoderLib/EncCfg.h
@@ -806,9 +806,9 @@ protected:
bool m_alf; ///< Adaptive Loop Filter
-#if IDCC_TPM_JEM
- bool m_IntraTMP; ///< intra Template Matching
- unsigned m_IntraTMP_MaxSize; ///< max CU size for which intra TMP is allowed
+#if JVET_V0130_INTRA_TMP
+ bool m_intraTMP; ///< intra Template Matching
+ unsigned m_intraTmpMaxSize; ///< max CU size for which intra TMP is allowed
#endif
#if JVET_V0094_BILATERAL_FILTER
bool m_BIF;
@@ -1287,11 +1287,11 @@ public:
bool getUseWrapAround () const { return m_wrapAround; }
void setWrapAroundOffset ( unsigned u ) { m_wrapAroundOffset = u; }
unsigned getWrapAroundOffset () const { return m_wrapAroundOffset; }
-#if IDCC_TPM_JEM
- void setUseIntraTMP(bool b) { m_IntraTMP = b; }
- bool getUseIntraTMP() { return m_IntraTMP; }
- void setIntraTMPMaxSize(unsigned n) { m_IntraTMP_MaxSize = n; }
- unsigned getIntraTMPMaxSize() { return m_IntraTMP_MaxSize; }
+#if JVET_V0130_INTRA_TMP
+ void setUseIntraTMP (bool b) { m_intraTMP = b; }
+ bool getUseIntraTMP() const { return m_intraTMP; }
+ void setIntraTMPMaxSize (unsigned n) { m_intraTmpMaxSize = n; }
+ unsigned getIntraTMPMaxSize() const { return m_intraTmpMaxSize; }
#endif
#if JVET_V0094_BILATERAL_FILTER
void setUseBIF ( bool b ) { m_BIF = b; }
diff --git a/source/Lib/EncoderLib/EncCu.cpp b/source/Lib/EncoderLib/EncCu.cpp
index 24f7f15146b2d4f0d3587cf2b2870839a6913383..59dbb6c05e0ec253279e2b27f0db3d0cb8cbef52 100644
--- a/source/Lib/EncoderLib/EncCu.cpp
+++ b/source/Lib/EncoderLib/EncCu.cpp
@@ -2008,8 +2008,8 @@ bool EncCu::xCheckRDCostIntra(CodingStructure *&tempCS, CodingStructure *&bestCS
m_modeCtrl->setISPMode(cu.ispMode);
m_modeCtrl->setISPLfnstIdx(cu.lfnstIdx);
m_modeCtrl->setMIPFlagISPPass(cu.mipFlag);
-#if IDCC_TPM_JEM
- m_modeCtrl->setTPMFlagISPPass(cu.TmpFlag);
+#if JVET_V0130_INTRA_TMP
+ m_modeCtrl->setTPMFlagISPPass(cu.tmpFlag);
#endif
m_modeCtrl->setBestISPIntraModeRelCU(cu.ispMode ? PU::getFinalIntraMode(*cu.firstPU, CHANNEL_TYPE_LUMA) : UINT8_MAX);
m_modeCtrl->setBestDCT2NonISPCostRelCU(m_modeCtrl->getMtsFirstPassNoIspCost());
@@ -3881,8 +3881,8 @@ void EncCu::xCheckRDCostMergeGeo2Nx2N(CodingStructure *&tempCS, CodingStructure
cu.mmvdSkip = false;
cu.skip = false;
cu.mipFlag = false;
-#if IDCC_TPM_JEM
- cu.TmpFlag = false;
+#if JVET_V0130_INTRA_TMP
+ cu.tmpFlag = false;
#endif
cu.bdpcmMode = 0;
@@ -4109,8 +4109,8 @@ void EncCu::xCheckRDCostMergeGeo2Nx2N(CodingStructure *&tempCS, CodingStructure
cu.mmvdSkip = false;
cu.skip = false;
cu.mipFlag = false;
-#if IDCC_TPM_JEM
- cu.TmpFlag = false;
+#if JVET_V0130_INTRA_TMP
+ cu.tmpFlag = false;
#endif
cu.bdpcmMode = 0;
PredictionUnit &pu = tempCS->addPU(cu, pm.chType);
@@ -4808,8 +4808,8 @@ void EncCu::xCheckSATDCostGeoMerge(CodingStructure *&tempCS, CodingUnit &cu, Pre
cu.mmvdSkip = false;
cu.skip = false;
cu.mipFlag = false;
-#if IDCC_TPM_JEM
- cu.TmpFlag = false;
+#if JVET_V0130_INTRA_TMP
+ cu.tmpFlag = false;
#endif
cu.bdpcmMode = 0;
pu.mergeFlag = true;
@@ -7677,8 +7677,6 @@ void EncCu::xCheckRDCostInterMultiHyp2Nx2N(CodingStructure *&tempCS, CodingStruc
std::stable_sort(mhResults.begin(), mhResults.end(), RDCostComp);
// actual testing with "true" RD costs
- if (std::min((int)mhResults.size(), m_pcEncCfg->getAddHypTries()) > 1)
- printf("nbTries=%d\n", std::min((int)mhResults.size(), m_pcEncCfg->getAddHypTries()));
for (int i = 0; i < std::min((int)mhResults.size(), m_pcEncCfg->getAddHypTries()); ++i)
{
tempCS->initStructData(encTestMode.qp);
diff --git a/source/Lib/EncoderLib/EncLib.cpp b/source/Lib/EncoderLib/EncLib.cpp
index d8c7306aed0e4a034ebd6b4e227375c665733f32..b76d81cf34849a886353165abf4603baf1914d66 100644
--- a/source/Lib/EncoderLib/EncLib.cpp
+++ b/source/Lib/EncoderLib/EncLib.cpp
@@ -1429,9 +1429,9 @@ void EncLib::xInitSPS( SPS& sps )
sps.setNumAddHypWeights(m_numAddHypWeights);
sps.setMaxNumAddHypRefFrames(m_maxNumAddHypRefFrames);
#endif
-#if IDCC_TPM_JEM
- sps.setUseIntraTMP(m_IntraTMP);
- sps.setIntraTMPMaxSize(m_IntraTMP_MaxSize);
+#if JVET_V0130_INTRA_TMP
+ sps.setUseIntraTMP(m_intraTMP);
+ sps.setIntraTMPMaxSize(m_intraTmpMaxSize);
#endif
// ADD_NEW_TOOL : (encoder lib) set tool enabling flags and associated parameters here
sps.setUseISP ( m_ISP );
diff --git a/source/Lib/EncoderLib/EncModeCtrl.cpp b/source/Lib/EncoderLib/EncModeCtrl.cpp
index 67b0825473d1807c127628dfe14551828a02688c..007b25ea102de18be0060bb5673c422cff62630a 100644
--- a/source/Lib/EncoderLib/EncModeCtrl.cpp
+++ b/source/Lib/EncoderLib/EncModeCtrl.cpp
@@ -2209,8 +2209,8 @@ bool EncModeCtrlMTnoRQT::tryMode( const EncTestMode& encTestmode, const CodingSt
int bit4 = cuECtx.ispLfnstIdx == 2;
int bit5 = cuECtx.mipFlag;
int bit6 = cuECtx.bestCostIsp < cuECtx.bestNonDCT2Cost * 0.95;
-#if IDCC_TPM_JEM
- int bit7 = cuECtx.TmpFlag;
+#if JVET_V0130_INTRA_TMP
+ int bit7 = cuECtx.tmpFlag;
#endif
int val =
(bit0) |
@@ -2220,8 +2220,8 @@ bool EncModeCtrlMTnoRQT::tryMode( const EncTestMode& encTestmode, const CodingSt
(bit4 << 4) |
(bit5 << 5) |
(bit6 << 6) |
-#if IDCC_TPM_JEM
- (bit7 << 7) |
+#if JVET_V0130_INTRA_TMP
+ (bit7 << 7) |
#endif
( cuECtx.bestPredModeDCT2 << 9 );
relatedCU.ispPredModeVal = val;
diff --git a/source/Lib/EncoderLib/EncModeCtrl.h b/source/Lib/EncoderLib/EncModeCtrl.h
index 5b62615ea0b94bb1cf4ee74eebe77560cf92ad82..528ddb8da068916047865280d04109ada77ef12c 100644
--- a/source/Lib/EncoderLib/EncModeCtrl.h
+++ b/source/Lib/EncoderLib/EncModeCtrl.h
@@ -238,8 +238,8 @@ struct ComprCUCtx
( MAX_DOUBLE )
, bestISPIntraMode
( UINT8_MAX )
-#if IDCC_TPM_JEM
- , TmpFlag(false)
+#if JVET_V0130_INTRA_TMP
+ , tmpFlag (false)
#endif
, mipFlag ( false )
, ispMode ( NOT_INTRA_SUBPARTITIONS )
@@ -286,8 +286,8 @@ struct ComprCUCtx
double bestDCT2NonISPCost;
double bestNonDCT2Cost;
uint8_t bestISPIntraMode;
-#if IDCC_TPM_JEM
- bool TmpFlag;
+#if JVET_V0130_INTRA_TMP
+ bool tmpFlag;
#endif
bool mipFlag;
uint8_t ispMode;
@@ -396,8 +396,8 @@ public:
void setBestNonDCT2Cost ( double val ) { m_ComprCUCtxList.back().bestNonDCT2Cost = val; }
uint8_t getBestISPIntraModeRelCU () const { return m_ComprCUCtxList.back().bestISPIntraMode; }
void setBestISPIntraModeRelCU ( uint8_t val ) { m_ComprCUCtxList.back().bestISPIntraMode = val; }
-#if IDCC_TPM_JEM
- void setTPMFlagISPPass(bool val) { m_ComprCUCtxList.back().TmpFlag = val; }
+#if JVET_V0130_INTRA_TMP
+ void setTPMFlagISPPass (bool val) { m_ComprCUCtxList.back().tmpFlag = val; }
#endif
void setMIPFlagISPPass ( bool val ) { m_ComprCUCtxList.back().mipFlag = val; }
void setISPMode ( uint8_t val ) { m_ComprCUCtxList.back().ispMode = val; }
diff --git a/source/Lib/EncoderLib/IntraSearch.cpp b/source/Lib/EncoderLib/IntraSearch.cpp
index 033ecbab76951161f7ef7acad5885c14694b0540..46474299c2704bb447f3af87fb6398c124173acd 100644
--- a/source/Lib/EncoderLib/IntraSearch.cpp
+++ b/source/Lib/EncoderLib/IntraSearch.cpp
@@ -394,7 +394,7 @@ bool IntraSearch::estIntraPredLumaQT(CodingUnit &cu, Partitioner &partitioner, c
const TempCtx ctxStart ( m_CtxCache, m_CABACEstimator->getCtx() );
const TempCtx ctxStartMipFlag ( m_CtxCache, SubCtx( Ctx::MipFlag, m_CABACEstimator->getCtx() ) );
-#if IDCC_TPM_JEM
+#if JVET_V0130_INTRA_TMP
const TempCtx ctxStartTpmFlag(m_CtxCache, SubCtx(Ctx::TmpFlag, m_CABACEstimator->getCtx()));
#endif
const TempCtx ctxStartIspMode ( m_CtxCache, SubCtx( Ctx::ISPMode, m_CABACEstimator->getCtx() ) );
@@ -494,9 +494,9 @@ bool IntraSearch::estIntraPredLumaQT(CodingUnit &cu, Partitioner &partitioner, c
const bool mipAllowed = sps.getUseMIP() && isLuma(partitioner.chType) && ((cu.lfnstIdx == 0) || allowLfnstWithMip(cu.firstPU->lumaSize()));
const bool testMip = mipAllowed && !(cu.lwidth() > (8 * cu.lheight()) || cu.lheight() > (8 * cu.lwidth()));
const bool supportedMipBlkSize = pu.lwidth() <= MIP_MAX_WIDTH && pu.lheight() <= MIP_MAX_HEIGHT;
-#if IDCC_TPM_JEM
- const bool tpmAllowed = sps.getUseIntraTMP() && isLuma(partitioner.chType) && ((cu.lfnstIdx == 0) || allowLfnstWithTpm());
- const bool testTpm = tpmAllowed && (cu.lwidth() <= sps.getIntraTMPMaxSize() && cu.lheight() <= sps.getIntraTMPMaxSize());
+#if JVET_V0130_INTRA_TMP
+ const bool tpmAllowed = sps.getUseIntraTMP() && isLuma(partitioner.chType) && ((cu.lfnstIdx == 0) || allowLfnstWithTmp());
+ const bool testTpm = tpmAllowed && (cu.lwidth() <= sps.getIntraTMPMaxSize() && cu.lheight() <= sps.getIntraTMPMaxSize());
#endif
static_vector<ModeInfo, FAST_UDI_MAX_RDMODE_NUM> uiRdModeList;
@@ -573,19 +573,20 @@ bool IntraSearch::estIntraPredLumaQT(CodingUnit &cu, Partitioner &partitioner, c
? std::max(numModesForFullRD, floorLog2(std::min(pu.lwidth(), pu.lheight())) - 1)
: numModesForFullRD;
}
-#if IDCC_TPM_JEM
- if (testTpm)
- numModesForFullRD += 1; // testing tpm
- const int numHadCand = (testMip ? 2 : 1) * 3 + testTpm;
+#if JVET_V0130_INTRA_TMP
+ if( testTpm )
+ {
+ numModesForFullRD += 1; // testing tpm
+ }
+ const int numHadCand = (testMip ? 2 : 1) * 3 + testTpm;
+
+ cu.tmpFlag = false;
#else
const int numHadCand = (testMip ? 2 : 1) * 3;
#endif
//*** Derive (regular) candidates using Hadamard
cu.mipFlag = false;
-#if IDCC_TPM_JEM
- cu.TmpFlag = false;
-#endif
//===== init pattern for luma prediction =====
initIntraPatternChType(cu, pu.Y(), true);
@@ -616,8 +617,8 @@ bool IntraSearch::estIntraPredLumaQT(CodingUnit &cu, Partitioner &partitioner, c
minSadHad += std::min(distParamSad.distFunc(distParamSad) * 2, distParamHad.distFunc(distParamHad));
// NB xFracModeBitsIntra will not affect the mode for chroma that may have already been pre-estimated.
-#if IDCC_TPM_JEM
- m_CABACEstimator->getCtx() = SubCtx(Ctx::TmpFlag, ctxStartTpmFlag);
+#if JVET_V0130_INTRA_TMP
+ m_CABACEstimator->getCtx() = SubCtx( Ctx::TmpFlag, ctxStartTpmFlag );
#endif
m_CABACEstimator->getCtx() = SubCtx( Ctx::MipFlag, ctxStartMipFlag );
m_CABACEstimator->getCtx() = SubCtx( Ctx::ISPMode, ctxStartIspMode );
@@ -693,8 +694,8 @@ bool IntraSearch::estIntraPredLumaQT(CodingUnit &cu, Partitioner &partitioner, c
// NB xFracModeBitsIntra will not affect the mode for chroma that may have already been
// pre-estimated.
-#if IDCC_TPM_JEM
- m_CABACEstimator->getCtx() = SubCtx(Ctx::TmpFlag, ctxStartTpmFlag);
+#if JVET_V0130_INTRA_TMP
+ m_CABACEstimator->getCtx() = SubCtx( Ctx::TmpFlag, ctxStartTpmFlag );
#endif
m_CABACEstimator->getCtx() = SubCtx(Ctx::MipFlag, ctxStartMipFlag);
m_CABACEstimator->getCtx() = SubCtx(Ctx::ISPMode, ctxStartIspMode);
@@ -761,8 +762,8 @@ bool IntraSearch::estIntraPredLumaQT(CodingUnit &cu, Partitioner &partitioner, c
std::min(distParamSad.distFunc(distParamSad) * 2, distParamHad.distFunc(distParamHad));
// NB xFracModeBitsIntra will not affect the mode for chroma that may have already been pre-estimated.
-#if IDCC_TPM_JEM
- m_CABACEstimator->getCtx() = SubCtx(Ctx::TmpFlag, ctxStartTpmFlag);
+#if JVET_V0130_INTRA_TMP
+ m_CABACEstimator->getCtx() = SubCtx( Ctx::TmpFlag, ctxStartTpmFlag );
#endif
m_CABACEstimator->getCtx() = SubCtx(Ctx::MipFlag, ctxStartMipFlag);
m_CABACEstimator->getCtx() = SubCtx(Ctx::ISPMode, ctxStartIspMode);
@@ -806,47 +807,41 @@ bool IntraSearch::estIntraPredLumaQT(CodingUnit &cu, Partitioner &partitioner, c
m_dSavedHadListLFNST.resize(3);
LFNSTSaveFlag = false;
}
-#if IDCC_TPM_JEM
- // derive TPM candidate using hadamard
- if (testTpm)
- {
- cu.TmpFlag = true;
- cu.mipFlag = false;
- pu.multiRefIdx = 0;
-
-
-
- int foundCandiNum = 0;
- bool bsuccessfull = 0;
- CodingUnit cu_cpy = cu;
-
- if (isRefTemplateAvailable(cu_cpy, cu_cpy.blocks[COMPONENT_Y]))
- {
- m_pcTrQuant->getTargetTemplate(&cu_cpy, pu.lwidth(), pu.lheight());
- m_pcTrQuant->candidateSearchIntra(&cu_cpy, pu.lwidth(), pu.lheight());
- bsuccessfull = m_pcTrQuant->generateTMPrediction(piPred.buf, piPred.stride, pu.lwidth(), pu.lheight(), foundCandiNum);
- }
- if (bsuccessfull && foundCandiNum >= 1)
- {
-
- Distortion minSadHad =
- std::min(distParamSad.distFunc(distParamSad) * 2, distParamHad.distFunc(distParamHad));
-
- m_CABACEstimator->getCtx() = SubCtx(Ctx::TmpFlag, ctxStartTpmFlag);
-
- uint64_t fracModeBits = xFracModeBitsIntra(pu, 0, CHANNEL_TYPE_LUMA);
-
- double cost = double(minSadHad) + double(fracModeBits) * sqrtLambdaForFirstPass;
- DTRACE(g_trace_ctx, D_INTRA_COST, "IntraTPM: %u, %llu, %f (%d)\n", minSadHad, fracModeBits, cost,
- 0);
-
- updateCandList(ModeInfo(0, 0, 0, NOT_INTRA_SUBPARTITIONS, 0, 1), cost, uiRdModeList,
- CandCostList, numModesForFullRD);
- updateCandList(ModeInfo(0, 0, 0, NOT_INTRA_SUBPARTITIONS, 0, 1),
- 0.8 * double(minSadHad), uiHadModeList, CandHadList, numHadCand);
- }
-
- }
+#if JVET_V0130_INTRA_TMP
+ // derive TPM candidate using hadamard
+ if( testTpm )
+ {
+ cu.tmpFlag = true;
+ cu.mipFlag = false;
+ pu.multiRefIdx = 0;
+
+ int foundCandiNum = 0;
+ bool bsuccessfull = 0;
+ CodingUnit cu_cpy = cu;
+
+ if( isRefTemplateAvailable( cu_cpy, cu_cpy.blocks[COMPONENT_Y] ) )
+ {
+ m_pcTrQuant->getTargetTemplate( &cu_cpy, pu.lwidth(), pu.lheight() );
+ m_pcTrQuant->candidateSearchIntra( &cu_cpy, pu.lwidth(), pu.lheight() );
+ bsuccessfull = m_pcTrQuant->generateTMPrediction( piPred.buf, piPred.stride, pu.lwidth(), pu.lheight(), foundCandiNum );
+ }
+ if( bsuccessfull && foundCandiNum >= 1 )
+ {
+
+ Distortion minSadHad =
+ std::min( distParamSad.distFunc( distParamSad ) * 2, distParamHad.distFunc( distParamHad ) );
+
+ m_CABACEstimator->getCtx() = SubCtx( Ctx::TmpFlag, ctxStartTpmFlag );
+
+ uint64_t fracModeBits = xFracModeBitsIntra( pu, 0, CHANNEL_TYPE_LUMA );
+
+ double cost = double( minSadHad ) + double( fracModeBits ) * sqrtLambdaForFirstPass;
+ DTRACE( g_trace_ctx, D_INTRA_COST, "IntraTPM: %u, %llu, %f (%d)\n", minSadHad, fracModeBits, cost, 0 );
+
+ updateCandList( ModeInfo( 0, 0, 0, NOT_INTRA_SUBPARTITIONS, 0, 1 ), cost, uiRdModeList, CandCostList, numModesForFullRD );
+ updateCandList( ModeInfo( 0, 0, 0, NOT_INTRA_SUBPARTITIONS, 0, 1 ), 0.8 * double( minSadHad ), uiHadModeList, CandHadList, numHadCand );
+ }
+ }
#endif
//*** Derive MIP candidates using Hadamard
if (testMip && !supportedMipBlkSize)
@@ -866,8 +861,8 @@ bool IntraSearch::estIntraPredLumaQT(CodingUnit &cu, Partitioner &partitioner, c
}
else if (testMip)
{
-#if IDCC_TPM_JEM
- cu.TmpFlag = 0;
+#if JVET_V0130_INTRA_TMP
+ cu.tmpFlag = 0;
#endif
cu.mipFlag = true;
pu.multiRefIdx = 0;
@@ -1097,8 +1092,8 @@ bool IntraSearch::estIntraPredLumaQT(CodingUnit &cu, Partitioner &partitioner, c
cs.interHad = 0;
//===== reset context models =====
-#if IDCC_TPM_JEM
- m_CABACEstimator->getCtx() = SubCtx(Ctx::TmpFlag, ctxStartTpmFlag);
+#if JVET_V0130_INTRA_TMP
+ m_CABACEstimator->getCtx() = SubCtx( Ctx::TmpFlag, ctxStartTpmFlag );
#endif
m_CABACEstimator->getCtx() = SubCtx(Ctx::MipFlag, ctxStartMipFlag);
m_CABACEstimator->getCtx() = SubCtx(Ctx::ISPMode, ctxStartIspMode);
@@ -1204,8 +1199,8 @@ bool IntraSearch::estIntraPredLumaQT(CodingUnit &cu, Partitioner &partitioner, c
cu.dimd = true;
}
#endif
-#if IDCC_TPM_JEM
- cu.TmpFlag = uiOrgMode.tpmFlg;
+#if JVET_V0130_INTRA_TMP
+ cu.tmpFlag = uiOrgMode.tmpFlag;
#endif
cu.mipFlag = uiOrgMode.mipFlg;
pu.mipTransposedFlag = uiOrgMode.mipTrFlg;
@@ -1218,10 +1213,10 @@ bool IntraSearch::estIntraPredLumaQT(CodingUnit &cu, Partitioner &partitioner, c
CHECK(cu.ispMode && cu.mipFlag, "Error: combination of ISP and MIP not supported");
CHECK(cu.ispMode && pu.multiRefIdx, "Error: combination of ISP and MRL not supported");
CHECK(cu.ispMode&& cu.colorTransform, "Error: combination of ISP and ACT not supported");
-#if IDCC_TPM_JEM
- CHECK(cu.mipFlag&& cu.TmpFlag, "Error: combination of MIP and TPM not supported");
- CHECK(cu.TmpFlag&& cu.ispMode, "Error: combination of TPM and ISP not supported");
- CHECK(cu.TmpFlag&& pu.multiRefIdx, "Error: combination of TPM and MRL not supported");
+#if JVET_V0130_INTRA_TMP
+ CHECK( cu.mipFlag && cu.tmpFlag, "Error: combination of MIP and TPM not supported" );
+ CHECK( cu.tmpFlag && cu.ispMode, "Error: combination of TPM and ISP not supported" );
+ CHECK( cu.tmpFlag && pu.multiRefIdx, "Error: combination of TPM and MRL not supported" );
#endif
#if ENABLE_DIMD && JVET_V0087_DIMD_NO_ISP
CHECK(cu.ispMode && cu.dimd, "Error: combination of ISP and DIMD not supported");
@@ -1265,14 +1260,14 @@ bool IntraSearch::estIntraPredLumaQT(CodingUnit &cu, Partitioner &partitioner, c
uiBestPUMode.ispMod, mtsCheckRangeFlag, mtsFirstCheckId, mtsLastCheckId, moreProbMTSIdxFirst);
}
}
-#if IDCC_TPM_JEM
- if (!cu.ispMode && !cu.mtsFlag && !cu.lfnstIdx && !cu.bdpcmMode && !pu.multiRefIdx && !cu.mipFlag && !cu.TmpFlag && testISP)
+#if JVET_V0130_INTRA_TMP
+ if( !cu.ispMode && !cu.mtsFlag && !cu.lfnstIdx && !cu.bdpcmMode && !pu.multiRefIdx && !cu.mipFlag && !cu.tmpFlag && testISP )
#else
if (!cu.ispMode && !cu.mtsFlag && !cu.lfnstIdx && !cu.bdpcmMode && !pu.multiRefIdx && !cu.mipFlag && testISP)
#endif
{
-#if IDCC_TPM_JEM
- m_regIntraRDListWithCosts.push_back(ModeInfoWithCost(cu.mipFlag, pu.mipTransposedFlag, pu.multiRefIdx, cu.ispMode, uiOrgMode.modeId, cu.TmpFlag, csTemp->cost));
+#if JVET_V0130_INTRA_TMP
+ m_regIntraRDListWithCosts.push_back( ModeInfoWithCost( cu.mipFlag, pu.mipTransposedFlag, pu.multiRefIdx, cu.ispMode, uiOrgMode.modeId, cu.tmpFlag, csTemp->cost ) );
#else
m_regIntraRDListWithCosts.push_back( ModeInfoWithCost( cu.mipFlag, pu.mipTransposedFlag, pu.multiRefIdx, cu.ispMode, uiOrgMode.modeId, csTemp->cost ) );
#endif
@@ -1290,10 +1285,10 @@ bool IntraSearch::estIntraPredLumaQT(CodingUnit &cu, Partitioner &partitioner, c
{
m_modeCostStore[lfnstIdx][mode] = tmpValidReturn ? csTemp->cost : (MAX_DOUBLE / 2.0); //(MAX_DOUBLE / 2.0) ??
}
-#if IDCC_TPM_JEM
- DTRACE(g_trace_ctx, D_INTRA_COST, "IntraCost T [x=%d,y=%d,w=%d,h=%d] %f (%d,%d,%d,%d,%d,%d,%d) \n", cu.blocks[0].x,
- cu.blocks[0].y, (int)width, (int)height, csTemp->cost, uiOrgMode.modeId, uiOrgMode.ispMod,
- pu.multiRefIdx, cu.TmpFlag, cu.mipFlag, cu.lfnstIdx, cu.mtsFlag);
+#if JVET_V0130_INTRA_TMP
+ DTRACE( g_trace_ctx, D_INTRA_COST, "IntraCost T [x=%d,y=%d,w=%d,h=%d] %f (%d,%d,%d,%d,%d,%d,%d) \n", cu.blocks[0].x,
+ cu.blocks[0].y, ( int ) width, ( int ) height, csTemp->cost, uiOrgMode.modeId, uiOrgMode.ispMod,
+ pu.multiRefIdx, cu.tmpFlag, cu.mipFlag, cu.lfnstIdx, cu.mtsFlag );
#else
DTRACE(g_trace_ctx, D_INTRA_COST, "IntraCost T [x=%d,y=%d,w=%d,h=%d] %f (%d,%d,%d,%d,%d,%d) \n", cu.blocks[0].x,
cu.blocks[0].y, (int) width, (int) height, csTemp->cost, uiOrgMode.modeId, uiOrgMode.ispMod,
@@ -1395,8 +1390,8 @@ bool IntraSearch::estIntraPredLumaQT(CodingUnit &cu, Partitioner &partitioner, c
if( validReturn )
{
//=== update PU data ====
-#if IDCC_TPM_JEM
- cu.TmpFlag = uiBestPUMode.tpmFlg;
+#if JVET_V0130_INTRA_TMP
+ cu.tmpFlag = uiBestPUMode.tmpFlag;
#endif
cu.mipFlag = uiBestPUMode.mipFlg;
pu.mipTransposedFlag = uiBestPUMode.mipTrFlg;
@@ -3352,16 +3347,16 @@ void IntraSearch::xIntraCodingTUBlock(TransformUnit &tu, const ComponentID &comp
}
else
{
-#if IDCC_TPM_JEM
- if (PU::isTmp(pu, chType))
- {
- int foundCandiNum;
- m_pcTrQuant->getTargetTemplate(tu.cu, pu.lwidth(), pu.lheight());
- m_pcTrQuant->candidateSearchIntra(tu.cu, pu.lwidth(), pu.lheight());
- m_pcTrQuant->generateTMPrediction(piPred.buf, piPred.stride, pu.lwidth(), pu.lheight(), foundCandiNum);
- assert(foundCandiNum >= 1);
- }
- else if (PU::isMIP(pu, chType))
+#if JVET_V0130_INTRA_TMP
+ if( PU::isTmp( pu, chType ) )
+ {
+ int foundCandiNum;
+ m_pcTrQuant->getTargetTemplate( tu.cu, pu.lwidth(), pu.lheight() );
+ m_pcTrQuant->candidateSearchIntra( tu.cu, pu.lwidth(), pu.lheight() );
+ m_pcTrQuant->generateTMPrediction( piPred.buf, piPred.stride, pu.lwidth(), pu.lheight(), foundCandiNum );
+ CHECK( foundCandiNum < 1, "" );
+ }
+ else if( PU::isMIP( pu, chType ) )
#else
if( PU::isMIP( pu, chType ) )
#endif
@@ -4586,17 +4581,17 @@ bool IntraSearch::xRecurIntraCodingACTQT(CodingStructure &cs, Partitioner &parti
PelBuf piResi = resiBuf.bufs[compID];
initIntraPatternChType(*tu.cu, area);
-#if IDCC_TPM_JEM
- if (PU::isTmp(pu, chType))
- {
- int foundCandiNum;
- m_pcTrQuant->getTargetTemplate(pu.cu, pu.lwidth(), pu.lheight());
- m_pcTrQuant->candidateSearchIntra(pu.cu, pu.lwidth(), pu.lheight());
- m_pcTrQuant->generateTMPrediction(piPred.buf, piPred.stride, pu.lwidth(), pu.lheight(), foundCandiNum);
- assert(foundCandiNum >= 1);
-
- }
- else if (PU::isMIP(pu, chType))
+#if JVET_V0130_INTRA_TMP
+ if( PU::isTmp( pu, chType ) )
+ {
+ int foundCandiNum;
+ m_pcTrQuant->getTargetTemplate( pu.cu, pu.lwidth(), pu.lheight() );
+ m_pcTrQuant->candidateSearchIntra( pu.cu, pu.lwidth(), pu.lheight() );
+ m_pcTrQuant->generateTMPrediction( piPred.buf, piPred.stride, pu.lwidth(), pu.lheight(), foundCandiNum );
+ CHECK( foundCandiNum < 1, "" );
+
+ }
+ else if( PU::isMIP( pu, chType ) )
#else
if (PU::isMIP(pu, chType))
#endif
diff --git a/source/Lib/EncoderLib/IntraSearch.h b/source/Lib/EncoderLib/IntraSearch.h
index 615408ac82f72010b877b5d9c8161ffd28b0b124..31b813dabccf8007b977a5b4d48c8ed6f2e78a89 100644
--- a/source/Lib/EncoderLib/IntraSearch.h
+++ b/source/Lib/EncoderLib/IntraSearch.h
@@ -221,13 +221,12 @@ private:
int mRefId; // PU::multiRefIdx
uint8_t ispMod; // CU::ispMode
uint32_t modeId; // PU::intraDir[CHANNEL_TYPE_LUMA]
-#if IDCC_TPM_JEM
- bool tpmFlg; // CU::TmpFlag
-#endif
-#if IDCC_TPM_JEM
- ModeInfo() : mipFlg(false), mipTrFlg(false), mRefId(0), ispMod(NOT_INTRA_SUBPARTITIONS), modeId(0), tpmFlg(0) {}
- ModeInfo(const bool mipf, const bool miptf, const int mrid, const uint8_t ispm, const uint32_t mode, const bool tpmf = 0) : mipFlg(mipf), mipTrFlg(miptf), mRefId(mrid), ispMod(ispm), modeId(mode), tpmFlg(tpmf) {}
- bool operator==(const ModeInfo cmp) const { return (mipFlg == cmp.mipFlg && mipTrFlg == cmp.mipTrFlg && mRefId == cmp.mRefId && ispMod == cmp.ispMod && modeId == cmp.modeId && tpmFlg == cmp.tpmFlg); }
+#if JVET_V0130_INTRA_TMP
+ bool tmpFlag; // CU::tmpFlag
+
+ ModeInfo() : mipFlg(false), mipTrFlg(false), mRefId(0), ispMod(NOT_INTRA_SUBPARTITIONS), modeId(0), tmpFlag(0) {}
+ ModeInfo(const bool mipf, const bool miptf, const int mrid, const uint8_t ispm, const uint32_t mode, const bool tpmf = 0) : mipFlg(mipf), mipTrFlg(miptf), mRefId(mrid), ispMod(ispm), modeId(mode), tmpFlag(tpmf) {}
+ bool operator==(const ModeInfo cmp) const { return (mipFlg == cmp.mipFlg && mipTrFlg == cmp.mipTrFlg && mRefId == cmp.mRefId && ispMod == cmp.ispMod && modeId == cmp.modeId && tmpFlag == cmp.tmpFlag); }
#else
ModeInfo() : mipFlg(false), mipTrFlg(false), mRefId(0), ispMod(NOT_INTRA_SUBPARTITIONS), modeId(0) {}
ModeInfo(const bool mipf, const bool miptf, const int mrid, const uint8_t ispm, const uint32_t mode) : mipFlg(mipf), mipTrFlg(miptf), mRefId(mrid), ispMod(ispm), modeId(mode) {}
@@ -238,9 +237,9 @@ private:
{
double rdCost;
ModeInfoWithCost() : ModeInfo(), rdCost(MAX_DOUBLE) {}
-#if IDCC_TPM_JEM
- ModeInfoWithCost(const bool mipf, const bool miptf, const int mrid, const uint8_t ispm, const uint32_t mode, const bool tpmf, double cost) : ModeInfo(mipf, miptf, mrid, ispm, mode, tpmf), rdCost(cost) {}
- bool operator==(const ModeInfoWithCost cmp) const { return (mipFlg == cmp.mipFlg && mipTrFlg == cmp.mipTrFlg && mRefId == cmp.mRefId && ispMod == cmp.ispMod && modeId == cmp.modeId && tpmFlg == cmp.tpmFlg && rdCost == cmp.rdCost); }
+#if JVET_V0130_INTRA_TMP
+ ModeInfoWithCost(const bool mipf, const bool miptf, const int mrid, const uint8_t ispm, const uint32_t mode, const bool tpmf, double cost) : ModeInfo(mipf, miptf, mrid, ispm, mode, tpmf), rdCost(cost) {}
+ bool operator==(const ModeInfoWithCost cmp) const { return (mipFlg == cmp.mipFlg && mipTrFlg == cmp.mipTrFlg && mRefId == cmp.mRefId && ispMod == cmp.ispMod && modeId == cmp.modeId && tmpFlag == cmp.tmpFlag && rdCost == cmp.rdCost); }
#else
ModeInfoWithCost(const bool mipf, const bool miptf, const int mrid, const uint8_t ispm, const uint32_t mode, double cost) : ModeInfo(mipf, miptf, mrid, ispm, mode), rdCost(cost) {}
bool operator==(const ModeInfoWithCost cmp) const { return (mipFlg == cmp.mipFlg && mipTrFlg == cmp.mipTrFlg && mRefId == cmp.mRefId && ispMod == cmp.ispMod && modeId == cmp.modeId && rdCost == cmp.rdCost); }
diff --git a/source/Lib/EncoderLib/VLCWriter.cpp b/source/Lib/EncoderLib/VLCWriter.cpp
index 4d123fa1f371870762637556c002c8b8fdde1232..cec7da12760889e3721451dbd17fe3b50b56380e 100644
--- a/source/Lib/EncoderLib/VLCWriter.cpp
+++ b/source/Lib/EncoderLib/VLCWriter.cpp
@@ -1384,11 +1384,11 @@ void HLSWriter::codeSPS( const SPS* pcSPS )
#if ENABLE_DIMD
WRITE_FLAG( pcSPS->getUseDimd() ? 1 : 0, "sps_dimd_enabled_flag");
#endif
-#if IDCC_TPM_JEM
+#if JVET_V0130_INTRA_TMP
WRITE_FLAG( pcSPS->getUseIntraTMP() ? 1 : 0, "sps_intraTMP_enabled_flag");
if(pcSPS->getUseIntraTMP())
{
- WRITE_UVLC(floorLog2(pcSPS->getIntraTMPMaxSize()), "sps_log2_intra_tmp_max_size");
+ WRITE_UVLC(floorLog2(pcSPS->getIntraTMPMaxSize()), "sps_log2_intra_tmp_max_size");
}
#endif