CI: Check if par I/O is compatible with optimizations and parallelization...

CI: Check if par I/O is compatible with optimizations and parallelization (poly, vect, OMP, CSE, unrolling)

.gitlab-ci.yml

@@ -292,32 +292,38 @@ test:Appl_OptFlow3D:
 
 test:ParIO_CheckEquality_2D:
     extends: .test_template_IO
+    tags:
+        - docker
+        - AVX2
     script:
         # 2D_Scalar_CheckEquality_ReadAfterWrite
         - mkdir -p data/2D_Scalar_CheckEquality_ReadAfterWrite
-        - python3 run_test.py ../Compiler/Compiler.jar:../Compiler/lib 2D_Scalar_CheckEquality_ReadAfterWrite IOTest/2D_Scalar_CheckEquality_ReadAfterWrite.knowledge "IOTest/2D_Scalar_CheckEquality_ReadAfterWrite.exa4" "" 2 1 Platform/random.platform output
+        - python3 run_test.py ../Compiler/Compiler.jar:../Compiler/lib 2D_Scalar_CheckEquality_ReadAfterWrite IOTest/2D_Scalar_CheckEquality_ReadAfterWrite.knowledge "IOTest/2D_Scalar_CheckEquality_ReadAfterWrite.exa4" "" 2 1 Platform/anyavx2.platform output
         # 2D_LayoutTrafo_CheckEquality_ReadAfterWrite
         - mkdir -p data/2D_LayoutTrafo_CheckEquality_ReadAfterWrite
-        - python3 run_test.py ../Compiler/Compiler.jar:../Compiler/lib 2D_LayoutTrafo_CheckEquality_ReadAfterWrite IOTest/2D_LayoutTrafo_CheckEquality_ReadAfterWrite.knowledge "IOTest/2D_LayoutTrafo_CheckEquality_ReadAfterWrite.exa4" "" 2 1 Platform/random.platform output
+        - python3 run_test.py ../Compiler/Compiler.jar:../Compiler/lib 2D_LayoutTrafo_CheckEquality_ReadAfterWrite IOTest/2D_LayoutTrafo_CheckEquality_ReadAfterWrite.knowledge "IOTest/2D_LayoutTrafo_CheckEquality_ReadAfterWrite.exa4" "" 2 1 Platform/anyavx2.platform output
         # 2D_Vector_CheckEquality_ReadAfterWrite
         - mkdir -p data/2D_Vector_CheckEquality_ReadAfterWrite
-        - python3 run_test.py ../Compiler/Compiler.jar:../Compiler/lib 2D_Vector_CheckEquality_ReadAfterWrite IOTest/2D_Vector_CheckEquality_ReadAfterWrite.knowledge "IOTest/2D_Vector_CheckEquality_ReadAfterWrite.exa4" "" 4 1 Platform/random.platform output
+        - python3 run_test.py ../Compiler/Compiler.jar:../Compiler/lib 2D_Vector_CheckEquality_ReadAfterWrite IOTest/2D_Vector_CheckEquality_ReadAfterWrite.knowledge "IOTest/2D_Vector_CheckEquality_ReadAfterWrite.exa4" "" 4 1 Platform/anyavx2.platform output
 
 test:ParIO_CheckEquality_3D:
     extends: .test_template_IO
+    tags:
+        - docker
+        - AVX2
     script:
         # 3D_Matrix_CheckEquality_ReadAfterWrite
         - mkdir -p data/3D_Matrix_CheckEquality_ReadAfterWrite
-        - python3 run_test.py ../Compiler/Compiler.jar:../Compiler/lib 3D_Matrix_CheckEquality_ReadAfterWrite IOTest/3D_Matrix_CheckEquality_ReadAfterWrite.knowledge "IOTest/3D_Matrix_CheckEquality_ReadAfterWrite.exa4" "" 4 1 Platform/random.platform output
+        - python3 run_test.py ../Compiler/Compiler.jar:../Compiler/lib 3D_Matrix_CheckEquality_ReadAfterWrite IOTest/3D_Matrix_CheckEquality_ReadAfterWrite.knowledge "IOTest/3D_Matrix_CheckEquality_ReadAfterWrite.exa4" "" 4 1 Platform/anyavx2.platform output
         # 3D_Scalar_CheckEquality_ReadAfterWrite
         - mkdir -p data/3D_Scalar_CheckEquality_ReadAfterWrite
-        - python3 run_test.py ../Compiler/Compiler.jar:../Compiler/lib 3D_Scalar_CheckEquality_ReadAfterWrite IOTest/3D_Scalar_CheckEquality_ReadAfterWrite.knowledge "IOTest/3D_Scalar_CheckEquality_ReadAfterWrite.exa4" "" 8 1 Platform/random.platform output
+        - python3 run_test.py ../Compiler/Compiler.jar:../Compiler/lib 3D_Scalar_CheckEquality_ReadAfterWrite IOTest/3D_Scalar_CheckEquality_ReadAfterWrite.knowledge "IOTest/3D_Scalar_CheckEquality_ReadAfterWrite.exa4" "" 8 1 Platform/anyavx2.platform output
         # 3D_Vector_CheckEquality_ReadAfterWrite
         - mkdir -p data/3D_Vector_CheckEquality_ReadAfterWrite
-        - python3 run_test.py ../Compiler/Compiler.jar:../Compiler/lib 3D_Vector_CheckEquality_ReadAfterWrite IOTest/3D_Vector_CheckEquality_ReadAfterWrite.knowledge "IOTest/3D_Vector_CheckEquality_ReadAfterWrite.exa4" "" 4 1 Platform/random.platform output
+        - python3 run_test.py ../Compiler/Compiler.jar:../Compiler/lib 3D_Vector_CheckEquality_ReadAfterWrite IOTest/3D_Vector_CheckEquality_ReadAfterWrite.knowledge "IOTest/3D_Vector_CheckEquality_ReadAfterWrite.exa4" "" 4 1 Platform/anyavx2.platform output
         # CheckEquality_ReadAfterWrite
         - mkdir -p data/CheckEquality_ReadAfterWrite
-        - python3 run_test.py ../Compiler/Compiler.jar:../Compiler/lib CheckEquality_ReadAfterWrite IOTest/CheckEquality_ReadAfterWrite.knowledge "IOTest/CheckEquality_ReadAfterWrite.exa4" "" 8 1 Platform/random.platform output
+        - python3 run_test.py ../Compiler/Compiler.jar:../Compiler/lib CheckEquality_ReadAfterWrite IOTest/CheckEquality_ReadAfterWrite.knowledge "IOTest/CheckEquality_ReadAfterWrite.exa4" "" 8 1 Platform/anyavx2.platform output
 
 
 ###############################################

Testing/IOTest/2D_LayoutTrafo_CheckEquality_ReadAfterWrite.exa4

@@ -36,7 +36,7 @@ Globals {
 
 Function compareFields@all () : Int {
   // check for equality
-  loop over dest {
+  loop over dest sequentially {
     Var diff : Real = fabs(dest - src)
     if( condition && diff > 0.001 ) {
       print("fields are not equal! src=", src, " dest=", dest)

Testing/IOTest/2D_LayoutTrafo_CheckEquality_ReadAfterWrite.knowledge

@@ -17,9 +17,16 @@ domain_rect_numFragsPerBlock_x = 1
 domain_rect_numFragsPerBlock_y = 2
 domain_rect_numFragsPerBlock_z = 1
 
-omp_enabled                       = false
-omp_numThreads                    = 1
+omp_enabled                       = true
+omp_numThreads                    = 2
 mpi_enabled                       = true
 mpi_numThreads                    = 2
-omp_parallelizeLoopOverFragments  = false
+omp_parallelizeLoopOverFragments  = true
 omp_parallelizeLoopOverDimensions = false
+
+poly_optLevel_fine			= 3
+opt_useAddressPrecalc		= true
+opt_vectorize				= true
+opt_unroll					= 2
+opt_conventionalCSE         = true
+opt_loopCarriedCSE          = true

Testing/IOTest/2D_Scalar_CheckEquality_ReadAfterWrite.exa4

@@ -32,7 +32,7 @@ Globals {
 
 Function compareFields@all () : Int {
   // check for equality
-  loop over dest {
+  loop over dest sequentially {
     Var diff : Real = fabs(dest - src)
     if( condition && diff > 0.001 ) {
       print("fields are not equal! src=", src, " dest=", dest)

Testing/IOTest/2D_Scalar_CheckEquality_ReadAfterWrite.knowledge

@@ -17,9 +17,16 @@ domain_rect_numFragsPerBlock_x = 1
 domain_rect_numFragsPerBlock_y = 2
 domain_rect_numFragsPerBlock_z = 1
 
-omp_enabled                       = false
-omp_numThreads                    = 1
+omp_enabled                       = true
+omp_numThreads                    = 2
 mpi_enabled                       = true
 mpi_numThreads                    = 2
-omp_parallelizeLoopOverFragments  = false
+omp_parallelizeLoopOverFragments  = true
 omp_parallelizeLoopOverDimensions = false
+
+poly_optLevel_fine			= 3
+opt_useAddressPrecalc		= true
+opt_vectorize				= true
+opt_unroll					= 2
+opt_conventionalCSE         = true
+opt_loopCarriedCSE          = true
\ No newline at end of file

Testing/IOTest/2D_Vector_CheckEquality_ReadAfterWrite.exa4

@@ -24,7 +24,7 @@ Field destVec < global, CellVecWithComm, None >@all
 
 Function compareFields@finest ( ) : Int {
   // check for equality
-  loop over destVec {
+  loop over destVec sequentially {
     compare(srcVec, destVec, 0.001)
   }
 

Testing/IOTest/2D_Vector_CheckEquality_ReadAfterWrite.knowledge

@@ -17,9 +17,16 @@ domain_rect_numFragsPerBlock_x = 1
 domain_rect_numFragsPerBlock_y = 1
 domain_rect_numFragsPerBlock_z = 1
 
-omp_enabled                       = false
-omp_numThreads                    = 1
+omp_enabled                       = true
+omp_numThreads                    = 2
 mpi_enabled                       = true
 mpi_numThreads                    = 4
-omp_parallelizeLoopOverFragments  = false
+omp_parallelizeLoopOverFragments  = true
 omp_parallelizeLoopOverDimensions = false
+
+poly_optLevel_fine			= 3
+opt_useAddressPrecalc		= true
+opt_vectorize				= true
+opt_unroll					= 2
+opt_conventionalCSE         = true
+opt_loopCarriedCSE          = true

Testing/IOTest/3D_Matrix_CheckEquality_ReadAfterWrite.exa4

@@ -24,7 +24,7 @@ Field destMat < global, CellMatWithComm, None >@all
 
 Function compareFields@finest ( ) : Int {
   // check for equality
-  loop over destMat {
+  loop over destMat sequentially {
     compare(srcMat, destMat, 0.001)
   }
 

Testing/IOTest/3D_Matrix_CheckEquality_ReadAfterWrite.knowledge

@@ -17,9 +17,16 @@ domain_rect_numFragsPerBlock_x = 1
 domain_rect_numFragsPerBlock_y = 1
 domain_rect_numFragsPerBlock_z = 1
 
-omp_enabled                       = false
-omp_numThreads                    = 1
+omp_enabled                       = true
+omp_numThreads                    = 10
 mpi_enabled                       = true
 mpi_numThreads                    = 4
 omp_parallelizeLoopOverFragments  = false
-omp_parallelizeLoopOverDimensions = false
+omp_parallelizeLoopOverDimensions = true
+
+poly_optLevel_fine			= 3
+opt_useAddressPrecalc		= true
+opt_vectorize				= true
+opt_unroll					= 4
+opt_conventionalCSE         = true
+opt_loopCarriedCSE          = false

Testing/IOTest/3D_Scalar_CheckEquality_ReadAfterWrite.exa4

@@ -24,7 +24,7 @@ Field dest< global, NodeWithComm, None >@all
 
 Function compareFields@all ( ) : Int {
   // check for equality
-  loop over dest {
+  loop over dest sequentially {
     Var diff : Real = fabs(dest - src)
     if( diff > 0.001 ) {
       print("fields are not equal! src=", src, " dest=", dest)

Testing/IOTest/3D_Scalar_CheckEquality_ReadAfterWrite.knowledge

@@ -17,9 +17,16 @@ domain_rect_numFragsPerBlock_x = 2
 domain_rect_numFragsPerBlock_y = 1
 domain_rect_numFragsPerBlock_z = 1
 
-omp_enabled                       = false
-omp_numThreads                    = 1
+omp_enabled                       = true
+omp_numThreads                    = 2
 mpi_enabled                       = true
 mpi_numThreads                    = 8
 omp_parallelizeLoopOverFragments  = false
-omp_parallelizeLoopOverDimensions = false
+omp_parallelizeLoopOverDimensions = true
+
+poly_optLevel_fine			= 3
+opt_useAddressPrecalc		= true
+opt_vectorize				= true
+opt_unroll					= 2
+opt_conventionalCSE         = true
+opt_loopCarriedCSE          = true

Testing/IOTest/3D_Vector_CheckEquality_ReadAfterWrite.exa4

@@ -24,7 +24,7 @@ Field destVec < global, CellVecWithComm, None >@all
 
 Function compareFields@finest ( ) : Int {
   // check for equality
-  loop over destVec {
+  loop over destVec sequentially {
     compare(srcVec, destVec, 0.001)
   }
 

Testing/IOTest/3D_Vector_CheckEquality_ReadAfterWrite.knowledge

@@ -17,10 +17,17 @@ domain_rect_numFragsPerBlock_x = 1
 domain_rect_numFragsPerBlock_y = 1
 domain_rect_numFragsPerBlock_z = 1
 
-omp_enabled                       = false
-omp_numThreads                    = 1
+omp_enabled                       = true
+omp_numThreads                    = 2
 mpi_enabled                       = true
 mpi_numThreads                    = 4
-omp_parallelizeLoopOverFragments  = false
+omp_parallelizeLoopOverFragments  = true
 omp_parallelizeLoopOverDimensions = false
 
+poly_optLevel_fine			= 3
+opt_useAddressPrecalc		= true
+opt_vectorize				= true
+opt_unroll					= 2
+opt_conventionalCSE         = true
+opt_loopCarriedCSE          = true
+

Testing/IOTest/CheckEquality_ReadAfterWrite.exa4

@@ -34,7 +34,7 @@ Field dest< global, NodeWithComm, 0.0 >@all
 
 Function compareFields@all : Int {
     // check for equality
-    loop over dest {
+    loop over dest sequentially {
         Var diff : Real = fabs(dest - src)
         if( diff > 0.001 ) {
             print("fields are not equal! src=", src, " dest=", dest)

Testing/IOTest/CheckEquality_ReadAfterWrite.knowledge

@@ -18,9 +18,16 @@ domain_rect_numFragsPerBlock_x = 2
 domain_rect_numFragsPerBlock_y = 1
 domain_rect_numFragsPerBlock_z = 1
 
-omp_enabled                       = false
-omp_numThreads                    = 1
+omp_enabled                       = true
+omp_numThreads                    = 2
 mpi_enabled                       = true
 mpi_numThreads                    = 8
 omp_parallelizeLoopOverFragments  = false
-omp_parallelizeLoopOverDimensions = false
+omp_parallelizeLoopOverDimensions = true
+
+poly_optLevel_fine			= 3
+opt_useAddressPrecalc		= true
+opt_vectorize				= true
+opt_unroll					= 2
+opt_conventionalCSE         = true
+opt_loopCarriedCSE          = true