[update] tensor: add more Opt0 contract func thread

2026-01-22 22:53:20 +01:00
parent 97d881819d
commit d45bb075cf
3 changed files with 512 additions and 9 deletions
@@ -1204,6 +1204,309 @@ TEST(tensorContractnProdOpt0_TYPE_FLOAT2 ){
  free_tensor_TYPE_FLOAT(M1);

 }
+
+
+TEST(tensorContractnProdThreadOpt0_TYPE_FLOAT2 ){
+  dimension *d0=create_dim(3);
+  dimension *d1=create_dim(3);
+#if VALGRIND_
+  d0->perm[0]=5;
+  d0->perm[1]=2; //3;
+  d0->perm[2]=3;
+
+  d1->perm[0]=2;
+  d1->perm[1]=3;//3;
+  d1->perm[2]=8;
+
+#else
+
+
+  d0->perm[0]=335;
+  d0->perm[1]=32; //3;
+  d0->perm[2]=43;
+
+  d1->perm[0]=32;
+  d1->perm[1]=43;//3;
+  d1->perm[2]=244;
+#endif
+
+  updateRankDim(d0);
+  updateRankDim(d1);
+
+
+  tensor_TYPE_FLOAT *M0 = CREATE_TENSOR_TYPE_FLOAT(d0);
+  tensor_TYPE_FLOAT *M1 = CREATE_TENSOR_TYPE_FLOAT(d1);
+
+  LOG("M0->dim->rank = %ld\n",M0->dim->rank);
+  LOG("M1->dim->rank = %ld\n",M1->dim->rank);
+  for(size_t i=0; i<M0->dim->rank;++i) M0->x[i]=i*0.1 +1;
+  for(size_t i=0; i<M1->dim->rank;++i) M1->x[i]=i*0.003 + 2;
+
+//  print_tensor_float(M0,"M0");
+//  print_tensor_float(M1,"M1");
+
+  tensor_TYPE_FLOAT *M=NULL;
+  tensor_TYPE_FLOAT *MnO=NULL;
+
+  tensorContractnProdThreadOpt0_TYPE_FLOAT(&M, M0,M1,2,8);
+//  print_tensor_float(M,"M");
+  tensorContractnProdNotOpt_TYPE_FLOAT(&MnO, M0,M1,2);
+
+
+//  print_tensor_float(MnO,"MnO");
+ 
+  // for(size_t i=0;i<M->dim->rank;++i)
+  //    EXPECT_EQ_TYPE_FLOAT(M->x[i],MnO->x[i]);
+    
+  EXPECT_ARRAY_EQ_TYPE_FLOAT(M->x,M->dim->rank,MnO->x,MnO->dim->rank);
+
+  free_tensor_TYPE_FLOAT(M);
+  free_tensor_TYPE_FLOAT(MnO);
+  free_tensor_TYPE_FLOAT(M0);
+  free_tensor_TYPE_FLOAT(M1);
+
+}
+
+TEST(tensorContractnPro2dThreadOpt0_TYPE_FLOAT2 ){
+  dimension *d0=create_dim(3);
+  dimension *d1=create_dim(3);
+#if VALGRIND_
+  d0->perm[0]=5;
+  d0->perm[1]=2; //3;
+  d0->perm[2]=3;
+
+  d1->perm[0]=2;
+  d1->perm[1]=3;//3;
+  d1->perm[2]=8;
+
+#else
+
+
+  d0->perm[0]=335;
+  d0->perm[1]=32; //3;
+  d0->perm[2]=43;
+
+  d1->perm[0]=32;
+  d1->perm[1]=43;//3;
+  d1->perm[2]=244;
+#endif
+
+  updateRankDim(d0);
+  updateRankDim(d1);
+
+
+  tensor_TYPE_FLOAT *M0 = CREATE_TENSOR_TYPE_FLOAT(d0);
+  tensor_TYPE_FLOAT *M1 = CREATE_TENSOR_TYPE_FLOAT(d1);
+
+  LOG("M0->dim->rank = %ld\n",M0->dim->rank);
+  LOG("M1->dim->rank = %ld\n",M1->dim->rank);
+  for(size_t i=0; i<M0->dim->rank;++i) M0->x[i]=i*0.1 +1;
+  for(size_t i=0; i<M1->dim->rank;++i) M1->x[i]=i*0.003 + 2;
+
+//  print_tensor_float(M0,"M0");
+//  print_tensor_float(M1,"M1");
+
+  tensor_TYPE_FLOAT *M=NULL;
+  tensor_TYPE_FLOAT *MnO=NULL;
+
+  tensorContractnPro2dThreadOpt0_TYPE_FLOAT(&M, M0,M1,2,8);
+//  print_tensor_float(M,"M");
+  tensorContractnProdNotOpt_TYPE_FLOAT(&MnO, M0,M1,2);
+
+
+//  print_tensor_float(MnO,"MnO");
+ 
+  // for(size_t i=0;i<M->dim->rank;++i)
+  //    EXPECT_EQ_TYPE_FLOAT(M->x[i],MnO->x[i]);
+    
+  EXPECT_ARRAY_EQ_TYPE_FLOAT(M->x,M->dim->rank,MnO->x,MnO->dim->rank);
+
+  free_tensor_TYPE_FLOAT(M);
+  free_tensor_TYPE_FLOAT(MnO);
+  free_tensor_TYPE_FLOAT(M0);
+  free_tensor_TYPE_FLOAT(M1);
+
+}
+
+TEST(tensorContractnPro2dThread_TYPE_FLOAT2 ){
+  dimension *d0=create_dim(3);
+  dimension *d1=create_dim(3);
+#if VALGRIND_
+  d0->perm[0]=5;
+  d0->perm[1]=2; //3;
+  d0->perm[2]=3;
+
+  d1->perm[0]=2;
+  d1->perm[1]=3;//3;
+  d1->perm[2]=8;
+
+#else
+
+
+  d0->perm[0]=335;
+  d0->perm[1]=32; //3;
+  d0->perm[2]=43;
+
+  d1->perm[0]=32;
+  d1->perm[1]=43;//3;
+  d1->perm[2]=244;
+#endif
+
+  updateRankDim(d0);
+  updateRankDim(d1);
+
+
+  tensor_TYPE_FLOAT *M0 = CREATE_TENSOR_TYPE_FLOAT(d0);
+  tensor_TYPE_FLOAT *M1 = CREATE_TENSOR_TYPE_FLOAT(d1);
+
+  LOG("M0->dim->rank = %ld\n",M0->dim->rank);
+  LOG("M1->dim->rank = %ld\n",M1->dim->rank);
+  for(size_t i=0; i<M0->dim->rank;++i) M0->x[i]=i*0.1 +1;
+  for(size_t i=0; i<M1->dim->rank;++i) M1->x[i]=i*0.003 + 2;
+
+//  print_tensor_float(M0,"M0");
+//  print_tensor_float(M1,"M1");
+
+  tensor_TYPE_FLOAT *M=NULL;
+  tensor_TYPE_FLOAT *MnO=NULL;
+
+  tensorContractnPro2dThread_TYPE_FLOAT(&M, M0,M1,2,8);
+//  print_tensor_float(M,"M");
+  tensorContractnProdNotOpt_TYPE_FLOAT(&MnO, M0,M1,2);
+
+
+//  print_tensor_float(MnO,"MnO");
+ 
+  // for(size_t i=0;i<M->dim->rank;++i)
+  //    EXPECT_EQ_TYPE_FLOAT(M->x[i],MnO->x[i]);
+    
+  EXPECT_ARRAY_EQ_TYPE_FLOAT(M->x,M->dim->rank,MnO->x,MnO->dim->rank);
+
+  free_tensor_TYPE_FLOAT(M);
+  free_tensor_TYPE_FLOAT(MnO);
+  free_tensor_TYPE_FLOAT(M0);
+  free_tensor_TYPE_FLOAT(M1);
+
+}
+
+
+
+TEST(tensorContractnProdThread_TYPE_FLOAT2 ){
+  dimension *d0=create_dim(3);
+  dimension *d1=create_dim(3);
+#if VALGRIND_
+  d0->perm[0]=5;
+  d0->perm[1]=2; //3;
+  d0->perm[2]=3;
+
+  d1->perm[0]=2;
+  d1->perm[1]=3;//3;
+  d1->perm[2]=8;
+
+#else
+
+
+  d0->perm[0]=335;
+  d0->perm[1]=32; //3;
+  d0->perm[2]=43;
+
+  d1->perm[0]=32;
+  d1->perm[1]=43;//3;
+  d1->perm[2]=244;
+#endif
+
+  updateRankDim(d0);
+  updateRankDim(d1);
+
+
+  tensor_TYPE_FLOAT *M0 = CREATE_TENSOR_TYPE_FLOAT(d0);
+  tensor_TYPE_FLOAT *M1 = CREATE_TENSOR_TYPE_FLOAT(d1);
+
+  LOG("M0->dim->rank = %ld\n",M0->dim->rank);
+  LOG("M1->dim->rank = %ld\n",M1->dim->rank);
+  for(size_t i=0; i<M0->dim->rank;++i) M0->x[i]=i*0.1 +1;
+  for(size_t i=0; i<M1->dim->rank;++i) M1->x[i]=i*0.003 + 2;
+
+//  print_tensor_float(M0,"M0");
+//  print_tensor_float(M1,"M1");
+
+  tensor_TYPE_FLOAT *M=NULL;
+  tensor_TYPE_FLOAT *MnO=NULL;
+
+  tensorContractnProdThread_TYPE_FLOAT(&M, M0,M1,2,8);
+//  print_tensor_float(M,"M");
+  tensorContractnProdNotOpt_TYPE_FLOAT(&MnO, M0,M1,2);
+
+
+//  print_tensor_float(MnO,"MnO");
+ 
+  // for(size_t i=0;i<M->dim->rank;++i)
+  //    EXPECT_EQ_TYPE_FLOAT(M->x[i],MnO->x[i]);
+    
+  EXPECT_ARRAY_EQ_TYPE_FLOAT(M->x,M->dim->rank,MnO->x,MnO->dim->rank);
+
+  free_tensor_TYPE_FLOAT(M);
+  free_tensor_TYPE_FLOAT(MnO);
+  free_tensor_TYPE_FLOAT(M0);
+  free_tensor_TYPE_FLOAT(M1);
+
+}
+TEST(tensorContractnProd_TYPE_DOUBLE_2_1 ){
+  dimension *d0=create_dim(2);
+  dimension *d1=create_dim(1);
+#if VALGRIND_
+  d0->perm[0]=4;
+  d0->perm[1]=2; //3;
+
+  d1->perm[0]=2;
+
+#else
+
+  d0->perm[0]=125;
+  d0->perm[1]=52; //3;
+
+  d1->perm[0]=52;
+#endif
+
+  updateRankDim(d0);
+  updateRankDim(d1);
+
+
+  tensor_TYPE_DOUBLE *M0 = CREATE_TENSOR_TYPE_DOUBLE(d0);
+  tensor_TYPE_DOUBLE *M1 = CREATE_TENSOR_TYPE_DOUBLE(d1);
+
+  LOG("M0->dim->rank = %ld\n",M0->dim->rank);
+  LOG("M1->dim->rank = %ld\n",M1->dim->rank);
+  for(size_t i=0; i<M0->dim->rank;++i) M0->x[i]=i*0.1 +1;
+  for(size_t i=0; i<M1->dim->rank;++i) M1->x[i]=i*0.003 + 2;
+
+  print_tensor_double(M0,"M0");
+  print_tensor_double(M1,"M1");
+
+  tensor_TYPE_DOUBLE *M=NULL;
+  tensor_TYPE_DOUBLE *MnO=NULL;
+
+  tensorContractnProd_TYPE_DOUBLE(&M, M0,M1,1);
+  //print_tensor_double(M,"M");
+  //cl_tensorContractnProd_TYPE_DOUBLE(&MnO, M0,M1,2);
+  tensorContractnProdNotOpt_TYPE_DOUBLE(&MnO, M0,M1,1);
+
+  print_tensor_double(MnO,"MnO");
+ 
+  // for(size_t i=0;i<M->dim->rank;++i)
+  //    EXPECT_EQ_TYPE_DOUBLE(M->x[i],MnO->x[i]);
+    
+  EXPECT_ARRAY_EQ_TYPE_DOUBLE(M->x,M->dim->rank,MnO->x,MnO->dim->rank);
+
+  free_tensor_TYPE_DOUBLE(M);
+  free_tensor_TYPE_DOUBLE(MnO);
+  free_tensor_TYPE_DOUBLE(M0);
+  free_tensor_TYPE_DOUBLE(M1);
+
+}
+
+
+
 TEST(tensorContractnProd_TYPE_DOUBLE_2_1 ){
  dimension *d0=create_dim(2);
  dimension *d1=create_dim(1);