Merge pull request #3800 from Tongdongq/move-gpu-files

Move gpu files

CMakeLists.txt

@@ -28,7 +28,7 @@ option(OPM_ENABLE_PYTHON_TESTS "Enable tests for the python bindings?" ON)
 option(ENABLE_FPGA "Enable FPGA kernels integration?" OFF)
 option(USE_CHOW_PATEL_ILU "Use the iterative ILU by Chow and Patel?" OFF)
 option(USE_CHOW_PATEL_ILU_GPU "Run iterative ILU decomposition on GPU? Requires USE_CHOW_PATEL_ILU" OFF)
-option(USE_CHOW_PATEL_ILU_GPU_PARALLEL "Try to use more parallelism on the GPU during the iterative ILU decomposition? Requires USE_CHOW_PATEL_ILU_GPU_PARALLEL" OFF)
+option(USE_CHOW_PATEL_ILU_GPU_PARALLEL "Try to use more parallelism on the GPU during the iterative ILU decomposition? Requires USE_CHOW_PATEL_ILU_GPU" OFF)
 
 if(SIBLING_SEARCH AND NOT opm-common_DIR)
   # guess the sibling dir

CMakeLists_files.cmake

@@ -93,22 +93,22 @@ list (APPEND MAIN_SOURCE_FILES
   )
 
 if(CUDA_FOUND)
-  list (APPEND MAIN_SOURCE_FILES opm/simulators/linalg/bda/cusparseSolverBackend.cu)
-  list (APPEND MAIN_SOURCE_FILES opm/simulators/linalg/bda/cuWellContributions.cu)
+  list (APPEND MAIN_SOURCE_FILES opm/simulators/linalg/bda/cuda/cusparseSolverBackend.cu)
+  list (APPEND MAIN_SOURCE_FILES opm/simulators/linalg/bda/cuda/cuWellContributions.cu)
 endif()
 if(OPENCL_FOUND)
   list (APPEND MAIN_SOURCE_FILES opm/simulators/linalg/bda/BlockedMatrix.cpp)
-  list (APPEND MAIN_SOURCE_FILES opm/simulators/linalg/bda/BILU0.cpp)
+  list (APPEND MAIN_SOURCE_FILES opm/simulators/linalg/bda/opencl/BILU0.cpp)
   list (APPEND MAIN_SOURCE_FILES opm/simulators/linalg/bda/Reorder.cpp)
-  list (APPEND MAIN_SOURCE_FILES opm/simulators/linalg/bda/ChowPatelIlu.cpp)
-  list (APPEND MAIN_SOURCE_FILES opm/simulators/linalg/bda/BISAI.cpp)
-  list (APPEND MAIN_SOURCE_FILES opm/simulators/linalg/bda/CPR.cpp)
-  list (APPEND MAIN_SOURCE_FILES opm/simulators/linalg/bda/opencl.cpp)
-  list (APPEND MAIN_SOURCE_FILES opm/simulators/linalg/bda/openclKernels.cpp)
-  list (APPEND MAIN_SOURCE_FILES opm/simulators/linalg/bda/OpenclMatrix.cpp)
+  list (APPEND MAIN_SOURCE_FILES opm/simulators/linalg/bda/opencl/ChowPatelIlu.cpp)
+  list (APPEND MAIN_SOURCE_FILES opm/simulators/linalg/bda/opencl/BISAI.cpp)
+  list (APPEND MAIN_SOURCE_FILES opm/simulators/linalg/bda/opencl/CPR.cpp)
+  list (APPEND MAIN_SOURCE_FILES opm/simulators/linalg/bda/opencl/opencl.cpp)
+  list (APPEND MAIN_SOURCE_FILES opm/simulators/linalg/bda/opencl/openclKernels.cpp)
+  list (APPEND MAIN_SOURCE_FILES opm/simulators/linalg/bda/opencl/OpenclMatrix.cpp)
   list (APPEND MAIN_SOURCE_FILES opm/simulators/linalg/bda/opencl/Preconditioner.cpp)
-  list (APPEND MAIN_SOURCE_FILES opm/simulators/linalg/bda/openclSolverBackend.cpp)
-  list (APPEND MAIN_SOURCE_FILES opm/simulators/linalg/bda/openclWellContributions.cpp)
+  list (APPEND MAIN_SOURCE_FILES opm/simulators/linalg/bda/opencl/openclSolverBackend.cpp)
+  list (APPEND MAIN_SOURCE_FILES opm/simulators/linalg/bda/opencl/openclWellContributions.cpp)
 endif()
 if(CUDA_FOUND OR OPENCL_FOUND OR HAVE_FPGA OR HAVE_AMGCL)
   list (APPEND MAIN_SOURCE_FILES opm/simulators/linalg/bda/WellContributions.cpp)
@@ -124,7 +124,7 @@ endif()
 if(HAVE_AMGCL)
   list (APPEND MAIN_SOURCE_FILES opm/simulators/linalg/bda/amgclSolverBackend.cpp)
   if(CUDA_FOUND)
-    list (APPEND MAIN_SOURCE_FILES opm/simulators/linalg/bda/amgclSolverBackend.cu)
+    list (APPEND MAIN_SOURCE_FILES opm/simulators/linalg/bda/cuda/amgclSolverBackend.cu)
   endif()
 endif()
 if(MPI_FOUND)
@@ -254,24 +254,24 @@ list (APPEND PUBLIC_HEADER_FILES
   opm/simulators/linalg/bda/BdaBridge.hpp
   opm/simulators/linalg/bda/BdaResult.hpp
   opm/simulators/linalg/bda/BdaSolver.hpp
-  opm/simulators/linalg/bda/BILU0.hpp
+  opm/simulators/linalg/bda/opencl/BILU0.hpp
   opm/simulators/linalg/bda/BlockedMatrix.hpp
-  opm/simulators/linalg/bda/CPR.hpp
-  opm/simulators/linalg/bda/cuda_header.hpp
-  opm/simulators/linalg/bda/cusparseSolverBackend.hpp
-  opm/simulators/linalg/bda/ChowPatelIlu.hpp
-  opm/simulators/linalg/bda/BISAI.hpp
+  opm/simulators/linalg/bda/opencl/CPR.hpp
+  opm/simulators/linalg/bda/cuda/cuda_header.hpp
+  opm/simulators/linalg/bda/cuda/cusparseSolverBackend.hpp
+  opm/simulators/linalg/bda/opencl/ChowPatelIlu.hpp
+  opm/simulators/linalg/bda/opencl/BISAI.hpp
   opm/simulators/linalg/bda/FPGABILU0.hpp
   opm/simulators/linalg/bda/FPGASolverBackend.hpp
   opm/simulators/linalg/bda/FPGAUtils.hpp
   opm/simulators/linalg/bda/Reorder.hpp
   opm/simulators/linalg/bda/ILUReorder.hpp
-  opm/simulators/linalg/bda/opencl.hpp
-  opm/simulators/linalg/bda/openclKernels.hpp
-  opm/simulators/linalg/bda/OpenclMatrix.hpp
+  opm/simulators/linalg/bda/opencl/opencl.hpp
+  opm/simulators/linalg/bda/opencl/openclKernels.hpp
+  opm/simulators/linalg/bda/opencl/OpenclMatrix.hpp
   opm/simulators/linalg/bda/opencl/Preconditioner.hpp
-  opm/simulators/linalg/bda/openclSolverBackend.hpp
-  opm/simulators/linalg/bda/openclWellContributions.hpp
+  opm/simulators/linalg/bda/opencl/openclSolverBackend.hpp
+  opm/simulators/linalg/bda/opencl/openclWellContributions.hpp
   opm/simulators/linalg/bda/Matrix.hpp
   opm/simulators/linalg/bda/MultisegmentWellContribution.hpp
   opm/simulators/linalg/bda/WellContributions.hpp

opencl-source-provider.cmake

@@ -14,7 +14,7 @@ endif()
 set(CL_SRC_FILE ${PROJECT_BINARY_DIR}/clSources.cpp)
 file(WRITE ${CL_SRC_FILE} "// This file is auto-generated. Do not edit!\n\n")
 file(APPEND ${CL_SRC_FILE} "#include <config.h>\n\n")
-file(APPEND ${CL_SRC_FILE} "#include <${BDA_DIR}/openclKernels.hpp>\n\n")
+file(APPEND ${CL_SRC_FILE} "#include <${BDA_DIR}/opencl/openclKernels.hpp>\n\n")
 file(APPEND ${CL_SRC_FILE} "namespace Opm\{\n\n")
 file(APPEND ${CL_SRC_FILE} "namespace Accelerator\{\n\n")
 

opm/simulators/linalg/bda/BdaBridge.cpp

@@ -30,12 +30,12 @@
 #include <opm/simulators/linalg/bda/BdaResult.hpp>
 
 #if HAVE_CUDA
-#include <opm/simulators/linalg/bda/cusparseSolverBackend.hpp>
+#include <opm/simulators/linalg/bda/cuda/cusparseSolverBackend.hpp>
 #endif
 
 #if HAVE_OPENCL
-#include <opm/simulators/linalg/bda/openclSolverBackend.hpp>
-#include <opm/simulators/linalg/bda/openclWellContributions.hpp>
+#include <opm/simulators/linalg/bda/opencl/openclSolverBackend.hpp>
+#include <opm/simulators/linalg/bda/opencl/openclWellContributions.hpp>
 #endif
 
 #if HAVE_FPGA

opm/simulators/linalg/bda/WellContributions.cpp

@@ -26,11 +26,11 @@
 #include <opm/simulators/linalg/bda/WellContributions.hpp>
 
 #ifdef HAVE_OPENCL
-#include <opm/simulators/linalg/bda/openclWellContributions.hpp>
+#include <opm/simulators/linalg/bda/opencl/openclWellContributions.hpp>
 #endif
 
 #ifdef HAVE_CUDA
-#include <opm/simulators/linalg/bda/cuWellContributions.hpp>
+#include <opm/simulators/linalg/bda/cuda/cuWellContributions.hpp>
 #endif
 
 namespace Opm

opm/simulators/linalg/bda/cuda/cuWellContributions.cu

@@ -22,9 +22,9 @@
 #include <cstdlib>
 #include <cstring>
 
-#include "opm/simulators/linalg/bda/cuWellContributions.hpp"
+#include "opm/simulators/linalg/bda/cuda/cuWellContributions.hpp"
 
-#include "opm/simulators/linalg/bda/cuda_header.hpp"
+#include "opm/simulators/linalg/bda/cuda/cuda_header.hpp"
 #include <cuda_runtime.h>
 
 #include <opm/common/OpmLog/OpmLog.hpp>

opm/simulators/linalg/bda/cuda/cusparseSolverBackend.cu

@@ -25,10 +25,10 @@
 #include <opm/common/OpmLog/OpmLog.hpp>
 #include <dune/common/timer.hh>
 
-#include <opm/simulators/linalg/bda/cusparseSolverBackend.hpp>
-#include <opm/simulators/linalg/bda/cuWellContributions.hpp>
+#include <opm/simulators/linalg/bda/cuda/cusparseSolverBackend.hpp>
+#include <opm/simulators/linalg/bda/cuda/cuWellContributions.hpp>
 #include <opm/simulators/linalg/bda/BdaResult.hpp>
-#include <opm/simulators/linalg/bda/cuda_header.hpp>
+#include <opm/simulators/linalg/bda/cuda/cuda_header.hpp>
 
 #include "cublas_v2.h"
 #include "cusparse_v2.h"

opm/simulators/linalg/bda/opencl/BILU0.cpp

@@ -25,8 +25,9 @@
 #include <dune/common/timer.hh>
 
 #include <opm/simulators/linalg/bda/BdaSolver.hpp>
-#include <opm/simulators/linalg/bda/BILU0.hpp>
-#include <opm/simulators/linalg/bda/ChowPatelIlu.hpp>
+#include <opm/simulators/linalg/bda/opencl/BILU0.hpp>
+#include <opm/simulators/linalg/bda/opencl/ChowPatelIlu.hpp>
+#include <opm/simulators/linalg/bda/opencl/openclKernels.hpp>
 #include <opm/simulators/linalg/bda/Reorder.hpp>
 
 
@@ -111,8 +112,8 @@ bool BILU0<block_size>::analyze_matrix(BlockedMatrix *mat)
     invDiagVals.resize(mat->Nb * bs * bs);
 
 #if CHOW_PATEL
-    Lmat = std::make_unique<BlockedMatrix>(mat->Nb, (mat->nnzbs - mat->Nb) / 2);
-    Umat = std::make_unique<BlockedMatrix>(mat->Nb, (mat->nnzbs - mat->Nb) / 2);
+    Lmat = std::make_unique<BlockedMatrix>(mat->Nb, (mat->nnzbs - mat->Nb) / 2, bs);
+    Umat = std::make_unique<BlockedMatrix>(mat->Nb, (mat->nnzbs - mat->Nb) / 2, bs);
 #endif
 
     s.invDiagVals = cl::Buffer(*context, CL_MEM_READ_WRITE, sizeof(double) * bs * bs * mat->Nb);
@@ -181,7 +182,7 @@ bool BILU0<block_size>::create_preconditioner(BlockedMatrix *mat)
     }
 
 #if CHOW_PATEL
-    chowPatelIlu.decomposition(queue, context,
+    chowPatelIlu.decomposition(queue.get(), context.get(),
                                LUmat.get(), Lmat.get(), Umat.get(),
                                invDiagVals.data(), diagIndex,
                                s.diagIndex, s.invDiagVals,

opm/simulators/linalg/bda/opencl/BILU0.hpp

@@ -25,10 +25,9 @@
 #include <opm/simulators/linalg/bda/BlockedMatrix.hpp>
 #include <opm/simulators/linalg/bda/ILUReorder.hpp>
 
-#include <opm/simulators/linalg/bda/opencl.hpp>
-#include <opm/simulators/linalg/bda/openclKernels.hpp>
+#include <opm/simulators/linalg/bda/opencl/opencl.hpp>
 #include <opm/simulators/linalg/bda/opencl/Preconditioner.hpp>
-#include <opm/simulators/linalg/bda/ChowPatelIlu.hpp>
+#include <opm/simulators/linalg/bda/opencl/ChowPatelIlu.hpp>
 
 
 namespace Opm
@@ -122,7 +121,11 @@ public:
 
     std::pair<cl::Buffer, cl::Buffer> get_preconditioner_data()
     {
+#if CHOW_PATEL
+        return std::make_pair(s.Lvals, s.invDiagVals); // send dummy, BISAI is disabled when ChowPatel is selected
+#else
         return std::make_pair(s.LUvals, s.invDiagVals);
+#endif
     }
 
 };

opm/simulators/linalg/bda/opencl/BISAI.cpp

@@ -26,9 +26,12 @@
 #include <dune/common/timer.hh>
 
 #include <opm/simulators/linalg/bda/BdaSolver.hpp>
-#include <opm/simulators/linalg/bda/BILU0.hpp>
-#include <opm/simulators/linalg/bda/BISAI.hpp>
+#include <opm/simulators/linalg/bda/opencl/opencl.hpp>
+#include <opm/simulators/linalg/bda/opencl/BILU0.hpp>
+#include <opm/simulators/linalg/bda/opencl/BISAI.hpp>
+#include <opm/simulators/linalg/bda/opencl/openclKernels.hpp>
 #include <opm/simulators/linalg/bda/Reorder.hpp>
+#include <opm/simulators/linalg/bda/opencl/ChowPatelIlu.hpp> // disable BISAI if ChowPatel is selected
 
 namespace Opm
 {
@@ -42,6 +45,9 @@ template <unsigned int block_size>
 BISAI<block_size>::BISAI(ILUReorder opencl_ilu_reorder_, int verbosity_) :
     Preconditioner<block_size>(verbosity_)
 {
+#if CHOW_PATEL
+    OPM_THROW(std::logic_error, "Error --linsolver=isai cannot be used if ChowPatelIlu is used, probably defined by CMake\n");
+#endif
     bilu0 = std::make_unique<BILU0<block_size> >(opencl_ilu_reorder_, verbosity_);
 }
 

opm/simulators/linalg/bda/opencl/BISAI.hpp

@@ -22,10 +22,8 @@
 
 #include <mutex>
 
-#include <opm/simulators/linalg/bda/BILU0.hpp>
-#include <opm/simulators/linalg/bda/opencl.hpp>
-#include <opm/simulators/linalg/bda/openclKernels.hpp>
-#include <opm/simulators/linalg/bda/openclSolverBackend.hpp>
+#include <opm/simulators/linalg/bda/opencl/opencl.hpp>
+#include <opm/simulators/linalg/bda/opencl/BILU0.hpp>
 #include <opm/simulators/linalg/bda/opencl/Preconditioner.hpp>
 
 namespace Opm
@@ -33,9 +31,6 @@ namespace Opm
 namespace Accelerator
 {
 
-template <unsigned int block_size>
-class openclSolverBackend;
-
 class BlockedMatrix;
 
 /// This class implements a Blocked version of the Incomplete Sparse Approximate Inverse (ISAI) preconditioner.

opm/simulators/linalg/bda/opencl/CPR.cpp

@@ -30,8 +30,9 @@
 
 #include <opm/simulators/linalg/bda/BdaBridge.hpp>
 #include <opm/simulators/linalg/bda/BlockedMatrix.hpp>
-#include <opm/simulators/linalg/bda/CPR.hpp>
-#include <opm/simulators/linalg/bda/OpenclMatrix.hpp>
+#include <opm/simulators/linalg/bda/opencl/CPR.hpp>
+#include <opm/simulators/linalg/bda/opencl/OpenclMatrix.hpp>
+#include <opm/simulators/linalg/bda/opencl/openclKernels.hpp>
 
 
 namespace Opm

opm/simulators/linalg/bda/opencl/CPR.hpp

@@ -30,25 +30,20 @@
 #endif
 #include <dune/istl/umfpack.hh>
 
-#include <opm/simulators/linalg/bda/BILU0.hpp>
-#include <opm/simulators/linalg/bda/opencl.hpp>
+#include <opm/simulators/linalg/bda/opencl/opencl.hpp>
+#include <opm/simulators/linalg/bda/opencl/BILU0.hpp>
 #include <opm/simulators/linalg/bda/Matrix.hpp>
-#include <opm/simulators/linalg/bda/OpenclMatrix.hpp>
+#include <opm/simulators/linalg/bda/opencl/OpenclMatrix.hpp>
 #include <opm/simulators/linalg/bda/ILUReorder.hpp>
 #include <opm/simulators/linalg/bda/opencl/Preconditioner.hpp>
 
-#include <opm/simulators/linalg/bda/openclKernels.hpp>
-#include <opm/simulators/linalg/bda/ChowPatelIlu.hpp>
-#include <opm/simulators/linalg/bda/openclSolverBackend.hpp>
+#include <opm/simulators/linalg/bda/opencl/openclSolverBackend.hpp>
 
 namespace Opm
 {
 namespace Accelerator
 {
 
-template <unsigned int block_size>
-class openclSolverBackend;
-
 class BlockedMatrix;
 
 /// This class implements a Constrained Pressure Residual (CPR) preconditioner

opm/simulators/linalg/bda/opencl/ChowPatelIlu.cpp

@@ -24,7 +24,8 @@
 #include <dune/common/timer.hh>
 #include <opm/simulators/linalg/MatrixBlock.hpp>
 
-#include <opm/simulators/linalg/bda/ChowPatelIlu.hpp>
+#include <opm/simulators/linalg/bda/BlockedMatrix.hpp>
+#include <opm/simulators/linalg/bda/opencl/ChowPatelIlu.hpp>
 
 #if CHOW_PATEL
 
@@ -532,7 +533,7 @@ void ChowPatelIlu<block_size>::decomposition(
     Timer t_total, t_preprocessing;
 
     // Ut is actually BSC format
-    std::unique_ptr<BlockedMatrix<bs> > Ut = std::make_unique<BlockedMatrix<bs> >(Nb, (nnzbs + Nb) / 2);
+    std::unique_ptr<BlockedMatrix> Ut = std::make_unique<BlockedMatrix>(Nb, (nnzbs + Nb) / 2, bs);
 
     Lmat->rowPointers[0] = 0;
     for (int i = 0; i < Nb+1; i++) {
@@ -989,7 +990,7 @@ void ChowPatelIlu<block_size>::gpu_decomposition(
 #define INSTANTIATE_BDA_FUNCTIONS(n)                                          \
 template void ChowPatelIlu<n>::decomposition(                                 \
     cl::CommandQueue *queue, cl::Context *context,                            \
-    BlockedMatrix<n> *LUmat, BlockedMatrix<n> *Lmat, BlockedMatrix<n> *Umat,  \
+    BlockedMatrix *LUmat, BlockedMatrix *Lmat, BlockedMatrix *Umat,  \
     double *invDiagVals, std::vector<int>& diagIndex,                         \
     cl::Buffer& d_diagIndex, cl::Buffer& d_invDiagVals,                       \
     cl::Buffer& d_Lvals, cl::Buffer& d_Lcols, cl::Buffer& d_Lrows,            \

opm/simulators/linalg/bda/opencl/ChowPatelIlu.hpp

@@ -23,8 +23,7 @@
 
 #include <mutex>
 
-#include <opm/simulators/linalg/bda/opencl.hpp>
-#include <opm/simulators/linalg/bda/BlockedMatrix.hpp>
+#include <opm/simulators/linalg/bda/opencl/opencl.hpp>
 
 // Variables CHOW_PATEL, CHOW_PATEL_GPU and CHOW_PATEL_GPU_PARALLEL are set by CMake
 // Pass -DUSE_CHOW_PATEL_ILU=1 to cmake to define CHOW_PATEL and use the iterative ILU decomposition
@@ -47,6 +46,7 @@ namespace Opm
 namespace Accelerator
 {
 
+class BlockedMatrix;
 
 // This class implements a blocked version on GPU of the Fine-Grained Parallel ILU (FGPILU) by Chow and Patel 2015:
 //     FINE-GRAINED PARALLEL INCOMPLETE LU FACTORIZATION, E. Chow and A. Patel, SIAM 2015, https://doi.org/10.1137/140968896

opm/simulators/linalg/bda/opencl/OpenclMatrix.cpp

@@ -22,7 +22,7 @@
 #include <opm/common/OpmLog/OpmLog.hpp>
 #include <opm/common/ErrorMacros.hpp>
 
-#include <opm/simulators/linalg/bda/OpenclMatrix.hpp>
+#include <opm/simulators/linalg/bda/opencl/OpenclMatrix.hpp>
 #include <opm/simulators/linalg/bda/BlockedMatrix.hpp>
 #include <opm/simulators/linalg/bda/Matrix.hpp>
 

opm/simulators/linalg/bda/opencl/OpenclMatrix.hpp

@@ -22,7 +22,7 @@
 
 #include <vector>
 
-#include <opm/simulators/linalg/bda/opencl.hpp>
+#include <opm/simulators/linalg/bda/opencl/opencl.hpp>
 
 namespace Opm
 {

opm/simulators/linalg/bda/opencl/Preconditioner.cpp

@@ -22,9 +22,9 @@
 
 #include <opm/common/ErrorMacros.hpp>
 
-#include <opm/simulators/linalg/bda/BILU0.hpp>
-#include <opm/simulators/linalg/bda/BISAI.hpp>
-#include <opm/simulators/linalg/bda/CPR.hpp>
+#include <opm/simulators/linalg/bda/opencl/BILU0.hpp>
+#include <opm/simulators/linalg/bda/opencl/BISAI.hpp>
+#include <opm/simulators/linalg/bda/opencl/CPR.hpp>
 #include <opm/simulators/linalg/bda/opencl/Preconditioner.hpp>
 
 namespace Opm

opm/simulators/linalg/bda/opencl/Preconditioner.hpp

@@ -20,7 +20,7 @@
 #ifndef OPM_PRECONDITIONER_HEADER_INCLUDED
 #define OPM_PRECONDITIONER_HEADER_INCLUDED
 
-#include <opm/simulators/linalg/bda/opencl.hpp>
+#include <opm/simulators/linalg/bda/opencl/opencl.hpp>
 #include <opm/simulators/linalg/bda/ILUReorder.hpp>
 
 namespace Opm

opm/simulators/linalg/bda/opencl/opencl.cpp

@@ -18,7 +18,7 @@
 */
 
 #include <config.h>
-#include <opm/simulators/linalg/bda/opencl.hpp>
+#include <opm/simulators/linalg/bda/opencl/opencl.hpp>
 #include <string>
 
 namespace Opm

opm/simulators/linalg/bda/opencl/openclKernels.cpp

@@ -25,8 +25,8 @@
 #include <opm/common/ErrorMacros.hpp>
 #include <dune/common/timer.hh>
 
-#include <opm/simulators/linalg/bda/openclKernels.hpp>
-#include <opm/simulators/linalg/bda/ChowPatelIlu.hpp>  // defines CHOW_PATEL
+#include <opm/simulators/linalg/bda/opencl/openclKernels.hpp>
+#include <opm/simulators/linalg/bda/opencl/ChowPatelIlu.hpp>  // defines CHOW_PATEL
 
 namespace Opm
 {

opm/simulators/linalg/bda/opencl/openclKernels.hpp

@@ -23,7 +23,7 @@
 #include <string>
 #include <memory>
 
-#include <opm/simulators/linalg/bda/opencl.hpp>
+#include <opm/simulators/linalg/bda/opencl/opencl.hpp>
 
 namespace Opm
 {

opm/simulators/linalg/bda/opencl/openclSolverBackend.cpp

@@ -25,8 +25,10 @@
 #include <opm/common/ErrorMacros.hpp>
 #include <dune/common/timer.hh>
 
-#include <opm/simulators/linalg/bda/openclSolverBackend.hpp>
-#include <opm/simulators/linalg/bda/openclWellContributions.hpp>
+#include <opm/simulators/linalg/bda/opencl/opencl.hpp>
+#include <opm/simulators/linalg/bda/opencl/openclKernels.hpp>
+#include <opm/simulators/linalg/bda/opencl/openclSolverBackend.hpp>
+#include <opm/simulators/linalg/bda/opencl/openclWellContributions.hpp>
 
 #include <opm/simulators/linalg/bda/BdaResult.hpp>
 #include <opm/simulators/linalg/bda/Reorder.hpp>

opm/simulators/linalg/bda/opencl/openclSolverBackend.hpp

@@ -20,26 +20,19 @@
 #ifndef OPM_OPENCLSOLVER_BACKEND_HEADER_INCLUDED
 #define OPM_OPENCLSOLVER_BACKEND_HEADER_INCLUDED
 
-#include <opm/simulators/linalg/bda/opencl.hpp>
-#include <opm/simulators/linalg/bda/openclKernels.hpp>
+#include <opm/simulators/linalg/bda/opencl/opencl.hpp>
 #include <opm/simulators/linalg/bda/BdaResult.hpp>
 #include <opm/simulators/linalg/bda/BdaSolver.hpp>
 #include <opm/simulators/linalg/bda/ILUReorder.hpp>
 #include <opm/simulators/linalg/bda/WellContributions.hpp>
-#include <opm/simulators/linalg/bda/BILU0.hpp>
 
 #include <opm/simulators/linalg/bda/opencl/Preconditioner.hpp>
 
-#include <tuple>
-
 namespace Opm
 {
 namespace Accelerator
 {
 
-template <unsigned int block_size>
-class CPR;
-
 /// This class implements a opencl-based ilu0-bicgstab solver on GPU
 template <unsigned int block_size>
 class openclSolverBackend : public BdaSolver<block_size>
@@ -183,9 +176,9 @@ public:
     /// \param[in] tolerance                  required relative tolerance for openclSolver
     /// \param[in] platformID                 the OpenCL platform to be used
     /// \param[in] deviceID                   the device to be used
-    /// \param[in] opencl_ilu_reorder         select either level_scheduling or graph_coloring, see BILU0.hpp for explanation
+    /// \param[in] opencl_ilu_reorder         select either level_scheduling or graph_coloring, see Reorder.hpp for explanation
     /// \param[in] linsolver                  indicating the preconditioner, equal to the --linsolver cmdline argument
-    ///                                       only ilu0 and cpr_quasiimpes are supported
+    ///                                       only ilu0, cpr_quasiimpes and isai are supported
     openclSolverBackend(int linear_solver_verbosity, int maxit, double tolerance, unsigned int platformID, unsigned int deviceID,
         ILUReorder opencl_ilu_reorder, std::string linsolver);
 

opm/simulators/linalg/bda/opencl/openclWellContributions.cpp

@@ -19,7 +19,7 @@
 
 #include <config.h> // CMake
 
-#include <opm/simulators/linalg/bda/openclWellContributions.hpp>
+#include <opm/simulators/linalg/bda/opencl/openclWellContributions.hpp>
 
 #include <cstdlib>
 #include <cstring>

opm/simulators/linalg/bda/opencl/openclWellContributions.hpp

@@ -22,8 +22,8 @@
 
 #include <opm/simulators/linalg/bda/WellContributions.hpp>
 
-#include <opm/simulators/linalg/bda/opencl.hpp>
-#include <opm/simulators/linalg/bda/openclKernels.hpp>
+#include <opm/simulators/linalg/bda/opencl/opencl.hpp>
+#include <opm/simulators/linalg/bda/opencl/openclKernels.hpp>
 
 #include <memory>
 #include <vector>

tests/test_csrToCscOffsetMap.cpp

@@ -10,7 +10,7 @@
 #include <dune/istl/bcrsmatrix.hh>
 #include <dune/istl/matrixmarket.hh>
 
-#include <opm/simulators/linalg/bda/BISAI.hpp>
+#include <opm/simulators/linalg/bda/opencl/BISAI.hpp>
 
 BOOST_AUTO_TEST_CASE(testcsrtocscoffsetmap){
 

tests/test_solvetransposed3x3.cpp

@@ -21,6 +21,7 @@
 
 #define BOOST_TEST_MODULE SolveTransposed3x3
 #include <boost/test/unit_test.hpp>
+#include <boost/version.hpp>
 
 #if BOOST_VERSION / 100000 == 1 && BOOST_VERSION / 100 % 1000 < 71
 #include <boost/test/floating_point_comparison.hpp>
@@ -30,7 +31,7 @@
 
 #include <dune/istl/bcrsmatrix.hh>
 
-#include <opm/simulators/linalg/bda/CPR.hpp>
+#include <opm/simulators/linalg/bda/opencl/CPR.hpp>
 
 BOOST_AUTO_TEST_CASE(testsolvetransposed3x3)
 {