-
Notifications
You must be signed in to change notification settings - Fork 5
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
* work in progress * work in progress * test first step * typo * changed template * changed template * debug * debug * added matrixParam to AddForcingTermsKernelDriver * changed header * debug * debug * debug * test * passing vector_matrix directly * debug * implemented simple class * add std::vector * debug * ahhh * passing 3 matrix data as a group * deleted const * debug * debug * debug * debug * debug * added matrix data in a group for AddJacobianFunction * debug * store d_rate_constants in struct pointer * using cudaMallocManaged * using cudaMallocManaged * using cudaMallocManaged * device strcut * get all pointers into struct for AddJacobian * debug * debug * passed? * passing struct of processSet to kernelDriver * clean up * removed header * group processSet data members into struct * minor fix * minor fix * get rid of micm from struct * debug * debug * minor fix * minor fix * minor fix * clean up * debug * test * clean up * clean up * clean up * clean up * test * test out for storing pointer to variable * store reused device pointer to variable in kernel * store reused device pointer to variable in kernel * work in progress - 8/31 * added struct for sparseMatrix * added struct for sparseMatrix * sparse matrix in struct * compiling test * compiling test * test run for decompose * test run for decompose function * edited cmake * minor fix * test * test for passing std::vector to .cu file * testing for thrust * moved all vectors * test * cleanup * minor fix * cleanup * added header file for vector * minor fix * debug * chnage ngrids in test * checking on cpu code * bug * trying for smaller thrust vector * vector reallocation failed? * small bug * test memory allocation * clean up vectors * added const * debug * no more vector * added const * removed header * added cmake * edited header file * debug * debug * change dot operator * testing * test * small bug * test * testing * testing cpu code * changed kernel config * changed kernel config * changed kernel config * checking pair reference in GPU * debug * debug * bug * transfer data from device to data * transfer data * get rid of i++ * test * bug * bug * bug * inside the kernel * debug with print statement * debug print * modified test * print * print statement * print * test * debug * debug * printing aik * debug * get rid of ++ inside [] * debug * changed * changed * changed * test * test * print test * testing for one grid * testing for one grid * print A sparseMatrix * print * print A outside of kernel * printing A with pointer * printing A with pointer * print A * check * print lf * print lf * test * added synchronization * try to print with hd * test * check * check * test * print * test * checking niLU second * test * check * test * debug * test * test * added more print * added more print * test * test * more debug * more print * test * more print * more print * more print * more print * more print * more print * more print * more print * test * more print * test * more print * more print * more print * more print * more print * test * test * test * print * print * print * pass? cleanup * changed to expect_near to expect_eq * changed back to expect_near * used constructor from base class * eliminated dervied class constructor * added underscore to all struct members * added more underscore * added more underscore * added more underscore * more modification * added time measurement * GPU test against CPU * added time measurement * added time measurement * vector * addSparseMatrix template * test * typo * added std to vector * minor change * incremental change made to processSet * modified cuda_process_set.cuh * changed to device struct
- Loading branch information
Showing
13 changed files
with
636 additions
and
196 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -1,17 +1,21 @@ | ||
| // Copyright (C) 2023 National Center for Atmospheric Research, | ||
| // | ||
| // SPDX-License-Identifier: Apache-2.0 | ||
| #pragma once | ||
| #include <micm/util/cuda_param.hpp> | ||
| #include <micm/util/cuda_param.hpp> | ||
|
|
||
| namespace micm | ||
| { | ||
| namespace cuda | ||
| { | ||
| std::chrono::nanoseconds AddForcingTermsKernelDriver( | ||
| CUDAMatrixParam& matrixParam, | ||
| CUDAProcessSetParam& processSet); | ||
| CudaMatrixParam& matrixParam, | ||
| CudaProcessSetParam& processSet); | ||
|
|
||
| std::chrono::nanoseconds AddJacobianTermsKernelDriver( | ||
| CUDAMatrixParam& matrixParam, | ||
| CUDASparseMatrixParam& sparseMatrix, | ||
| CUDAProcessSetParam& processSet); | ||
| CudaMatrixParam& matrixParam, | ||
| CudaSparseMatrixParam& sparseMatrix, | ||
| CudaProcessSetParam& processSet); | ||
| } // namespace cuda | ||
| } // namespace micm |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,15 @@ | ||
| // Copyright (C) 2023 National Center for Atmospheric Research, | ||
| // | ||
| // SPDX-License-Identifier: Apache-2.0 | ||
| #pragma once | ||
| #include <micm/util/cuda_param.hpp> | ||
| #include <vector> | ||
| namespace micm | ||
| { | ||
| namespace cuda | ||
| { | ||
| std::chrono::nanoseconds DecomposeKernelDriver( | ||
| CudaSparseMatrixParam& sparseMatrix, | ||
| CudaSolverParam& solver); | ||
| } // namespace cuda | ||
| } // namespace micm |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,70 @@ | ||
| // Copyright (C) 2023 National Center for Atmospheric Research, | ||
| // | ||
| // SPDX-License-Identifier: Apache-2.0 | ||
| #pragma once | ||
| #include<micm/solver/lu_decomposition.hpp> | ||
| #include<micm/util/cuda_param.hpp> | ||
| #include <stdexcept> | ||
| #include <chrono> | ||
| #ifdef USE_CUDA | ||
| #include <micm/solver/cuda_lu_decomposition.cuh> | ||
| #endif | ||
|
|
||
| #ifdef USE_CUDA | ||
| namespace micm{ | ||
| class CudaLuDecomposition: public LuDecomposition{ | ||
| public: | ||
| CudaLuDecomposition(){}; | ||
|
|
||
| template<typename T, template<class> typename SparseMatrixPolicy> | ||
| requires VectorizableSparse<SparseMatrixPolicy<T>> | ||
| std::chrono::nanoseconds Decompose( | ||
| const SparseMatrixPolicy<T>&A, | ||
| SparseMatrixPolicy<T>& L, | ||
| SparseMatrixPolicy<T>& U) const; | ||
| }; | ||
|
|
||
| template<typename T, template<class> class SparseMatrixPolicy> | ||
| requires(VectorizableSparse<SparseMatrixPolicy<T>>) | ||
| std::chrono::nanoseconds CudaLuDecomposition::Decompose( | ||
| const SparseMatrixPolicy<T>& A, | ||
| SparseMatrixPolicy<T>& L, | ||
| SparseMatrixPolicy<T>& U) const | ||
| { | ||
| CudaSparseMatrixParam sparseMatrix; | ||
| sparseMatrix.A_ = A.AsVector().data(); | ||
| sparseMatrix.A_size_ = A.AsVector().size(); | ||
| sparseMatrix.L_ = L.AsVector().data(); | ||
| sparseMatrix.L_size_ = L.AsVector().size(); | ||
| sparseMatrix.U_ = U.AsVector().data(); | ||
| sparseMatrix.U_size_ = U.AsVector().size(); | ||
| sparseMatrix.n_grids_ = A.size(); | ||
|
|
||
| CudaSolverParam solver; | ||
| solver.do_aik_ = do_aik_.data(); | ||
| solver.do_aik_size_ = do_aik_.size(); | ||
| solver.aik_ = aik_.data(); | ||
| solver.aik_size_ = aik_.size(); | ||
| solver.do_aki_ = do_aki_.data(); | ||
| solver.do_aki_size_ = do_aki_.size(); | ||
| solver.aki_ = aki_.data(); | ||
| solver.aki_size_ = aki_.size(); | ||
| solver.uii_ = uii_.data(); | ||
| solver.uii_size_ = uii_.size(); | ||
|
|
||
| solver.niLU_ = niLU_.data(); | ||
| solver.niLU_size_ = niLU_.size(); | ||
| solver.uik_nkj_ = uik_nkj_.data(); | ||
| solver.uik_nkj_size_ = uik_nkj_.size(); | ||
| solver.lij_ujk_ = lij_ujk_.data(); | ||
| solver.lij_ujk_size_ = lij_ujk_.size(); | ||
| solver.lki_nkj_ = lki_nkj_.data(); | ||
| solver.lki_nkj_size_ = lki_nkj_.size(); | ||
| solver.lkj_uji_ = lkj_uji_.data(); | ||
| solver.lkj_uji_size_ = lkj_uji_.size(); | ||
|
|
||
| //calling kernelSetup function | ||
| return micm::cuda::DecomposeKernelDriver(sparseMatrix, solver); | ||
| } | ||
| }//end micm | ||
| #endif |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -1,33 +1,63 @@ | ||
| #include<vector> | ||
|
|
||
| #ifndef CUDA_PARAM_HPP | ||
| #define CUDA_PARAM_HPP | ||
| // member data of class CUDAProcessSet grouped in struct passing to kernel driver function | ||
| struct CUDAProcessSetParam | ||
| { | ||
| const size_t* number_of_reactants; | ||
| const size_t* reactant_ids; | ||
| size_t reactant_ids_size; | ||
| const size_t* number_of_products; | ||
| const size_t* product_ids; | ||
| size_t product_ids_size; | ||
| const double* yields; | ||
| size_t yields_size; | ||
| const size_t* jacobian_flat_ids; | ||
| size_t jacobian_flat_ids_size; | ||
| }; | ||
| // different matrix data grouped in struct passing to kernel driver function | ||
| struct CUDAMatrixParam | ||
| { | ||
| const double* rate_constants; | ||
| const double* state_variables; | ||
| double* forcing; | ||
| size_t n_grids; | ||
| size_t n_reactions; | ||
| size_t n_species; | ||
| }; | ||
| // sparseMatrix data grouped in struct passing to kernel driver function | ||
| struct CUDASparseMatrixParam | ||
| { | ||
| double* jacobian; | ||
| size_t jacobian_size; | ||
| //member data of class CudaProcessSet grouped in struct passing to kernel driver function | ||
| const size_t BLOCK_SIZE = 320; | ||
| struct CudaProcessSetParam{ | ||
| const size_t* number_of_reactants_; | ||
| const size_t* reactant_ids_; | ||
| size_t reactant_ids_size_; | ||
| const size_t* number_of_products_; | ||
| const size_t* product_ids_; | ||
| size_t product_ids_size_; | ||
| const double* yields_; | ||
| size_t yields_size_; | ||
| const size_t* jacobian_flat_ids_; | ||
| size_t jacobian_flat_ids_size_; | ||
| }; | ||
|
|
||
| struct CudaSolverParam{ | ||
| const std::pair<size_t, size_t>* niLU_; | ||
| size_t niLU_size_; | ||
| const char* do_aik_; | ||
| size_t do_aik_size_; | ||
| const size_t* aik_; | ||
| size_t aik_size_; | ||
| const std::pair<size_t, size_t>* uik_nkj_; | ||
| size_t uik_nkj_size_; | ||
| const std::pair<size_t, size_t>* lij_ujk_; | ||
| size_t lij_ujk_size_; | ||
| const char* do_aki_; | ||
| size_t do_aki_size_; | ||
| const size_t* aki_; | ||
| size_t aki_size_; | ||
| const std::pair<size_t, size_t>* lki_nkj_; | ||
| size_t lki_nkj_size_; | ||
| const std::pair<size_t, size_t>* lkj_uji_; | ||
| size_t lkj_uji_size_; | ||
| const size_t* uii_; | ||
| size_t uii_size_; | ||
| }; | ||
| //different matrix data grouped in struct passing to kernel driver function | ||
| struct CudaMatrixParam{ | ||
| const double* rate_constants_; | ||
| const double* state_variables_; | ||
| double* forcing_; | ||
| size_t n_grids_; | ||
| size_t n_reactions_; | ||
| size_t n_species_; | ||
| }; | ||
| //sparseMatrix data grouped in struct passing to kernel driver function | ||
| struct CudaSparseMatrixParam{ | ||
| double* jacobian_; | ||
| size_t jacobian_size_; | ||
| const double* A_; | ||
| size_t A_size_; | ||
| double* L_; | ||
| size_t L_size_; | ||
| double* U_; | ||
| size_t U_size_; | ||
| size_t n_grids_; | ||
| }; | ||
| #endif |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
|
|
@@ -86,4 +86,5 @@ if(ENABLE_CUDA) | |
| ) | ||
| endif() | ||
|
|
||
| add_subdirectory(process) | ||
| add_subdirectory(process) | ||
| add_subdirectory(solver) | ||
Oops, something went wrong.