You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Both TPLs and GEOS can be compiled on Maple GPUs with clang 17.0.4 .
However, only flow only simulations (e.g., SPE 11) can be running there.
Even with serial run, any case (tested five tutorial examples) involving mechanical deformation failed at the first time step with following error:
terminate called after throwing an instance of 'umpire::runtime_error'
what(): ! Umpire runtime_error [/shared/data1/Users/j0551570/Compilation/Develop_250324/thirdPartyLibs/build-maple-llvm-release/chai/src/chai/src/tpl/umpire/src/umpire/alloc/CudaMallocAllocator.hpp:62]: cudaFree( ptr = 0xfffa2e000000 ) failed with error: an illegal memory access was encountered
I tried to use --trace-data-migration to trace host-device data migration for running Kirsch problem, which is a simple mechanical example with linear elasticity.
Allocated 49.0 KB to the DEVICE: LvArray::Array<long long, 1, camp::int_seq<long, 0l>, int, LvArray::ChaiBuffer> Problem/domain/MeshBodies/mesh1/meshLevels/Level0/nodeManager/mechanicsSolver_totalDisplacement_dofIndex Free memory on device: 94.0 GB
Moved 49.0 KB to the DEVICE: LvArray::Array<long long, 1, camp::int_seq<long, 0l>, int, LvArray::ChaiBuffer> Problem/domain/MeshBodies/mesh1/meshLevels/Level0/nodeManager/mechanicsSolver_totalDisplacement_dofIndex
Allocated 147.0 KB to the DEVICE: LvArray::Array<double, 2, camp::int_seq<long, 1l, 0l>, int, LvArray::ChaiBuffer> Free memory on device: 94.0 GB
Moved 147.0 KB to the DEVICE: LvArray::Array<double, 2, camp::int_seq<long, 1l, 0l>, int, LvArray::ChaiBuffer>
Allocated 147.0 KB to the DEVICE: LvArray::Array<double, 2, camp::int_seq<long, 1l, 0l>, int, LvArray::ChaiBuffer> Problem/domain/MeshBodies/mesh1/meshLevels/Level0/nodeManager/totalDisplacement Free memory on device: 94.0 GB
Moved 147.0 KB to the DEVICE: LvArray::Array<double, 2, camp::int_seq<long, 1l, 0l>, int, LvArray::ChaiBuffer> Problem/domain/MeshBodies/mesh1/meshLevels/Level0/nodeManager/totalDisplacement
Allocated 250.0 KB to the DEVICE: LvArray::Array<double, 2, camp::int_seq<long, 0l, 1l>, int, LvArray::ChaiBuffer> Problem/domain/MeshBodies/mesh1/meshLevels/Level0/ElementRegions/elementRegionsGroup/Omega/elementSubRegions/cb1/rock_density Free memory on device: 94.0 GB
Moved 250.0 KB to the DEVICE: LvArray::Array<double, 2, camp::int_seq<long, 0l, 1l>, int, LvArray::ChaiBuffer> Problem/domain/MeshBodies/mesh1/meshLevels/Level0/ElementRegions/elementRegionsGroup/Omega/elementSubRegions/cb1/rock_density
Freed 8.0 B to the HOST : LvArray::Array<int, 1, camp::int_seq<long, 0l>, int, LvArray::ChaiBuffer> Free memory on device: 0.0 B
Freed 24.5 KB to the HOST : LvArray::ArrayOfArrays<int, int, LvArray::ChaiBuffer> Problem/domain/MeshBodies/mesh1/meshLevels/Level0/nodeManager/elemList/m_sizes Free memory on device: 0.0 B
Freed 24.5 KB to the HOST : LvArray::ArrayOfArrays<int, int, LvArray::ChaiBuffer> Problem/domain/MeshBodies/mesh1/meshLevels/Level0/nodeManager/elemList/m_offsets Free memory on device: 0.0 B
Freed 392.1 KB to the HOST : LvArray::ArrayOfArrays<int, int, LvArray::ChaiBuffer> Problem/domain/MeshBodies/mesh1/meshLevels/Level0/nodeManager/elemList/m_values Free memory on device: 0.0 B
Freed 24.5 KB to the HOST : LvArray::ArrayOfArrays<int, int, LvArray::ChaiBuffer> Problem/domain/MeshBodies/mesh1/meshLevels/Level0/nodeManager/elemSubRegionList/m_sizes Free memory on device: 0.0 B
Freed 24.5 KB to the HOST : LvArray::ArrayOfArrays<int, int, LvArray::ChaiBuffer> Problem/domain/MeshBodies/mesh1/meshLevels/Level0/nodeManager/elemSubRegionList/m_offsets Free memory on device: 0.0 B
Freed 392.1 KB to the HOST : LvArray::ArrayOfArrays<int, int, LvArray::ChaiBuffer> Problem/domain/MeshBodies/mesh1/meshLevels/Level0/nodeManager/elemSubRegionList/m_values Free memory on device: 0.0 B
Freed 24.5 KB to the HOST : LvArray::ArrayOfArrays<int, int, LvArray::ChaiBuffer> Problem/domain/MeshBodies/mesh1/meshLevels/Level0/nodeManager/elemRegionList/m_sizes Free memory on device: 0.0 B
Freed 24.5 KB to the HOST : LvArray::ArrayOfArrays<int, int, LvArray::ChaiBuffer> Problem/domain/MeshBodies/mesh1/meshLevels/Level0/nodeManager/elemRegionList/m_offsets Free memory on device: 0.0 B
Freed 392.1 KB to the HOST : LvArray::ArrayOfArrays<int, int, LvArray::ChaiBuffer> Problem/domain/MeshBodies/mesh1/meshLevels/Level0/nodeManager/elemRegionList/m_values Free memory on device: 0.0 B
Freed 24.5 KB to the HOST : LvArray::ArrayOfArrays<int, int, LvArray::ChaiBuffer> /m_sizes Free memory on device: 0.0 B
Freed 24.5 KB to the HOST : LvArray::ArrayOfArrays<int, int, LvArray::ChaiBuffer> /m_offsets Free memory on device: 0.0 B
Freed 417.6 KB to the HOST : LvArray::ArrayOfArrays<int, int, LvArray::ChaiBuffer> /m_values Free memory on device: 0.0 B
Freed 24.5 KB to the HOST : LvArray::ArrayOfArrays<int, int, LvArray::ChaiBuffer> /m_sizes Free memory on device: 0.0 B
Freed 24.5 KB to the HOST : LvArray::ArrayOfArrays<int, int, LvArray::ChaiBuffer> /m_offsets Free memory on device: 0.0 B
Freed 324.6 KB to the HOST : LvArray::ArrayOfArrays<int, int, LvArray::ChaiBuffer> /m_values Free memory on device: 0.0 B
Freed 147.0 KB to the HOST : LvArray::Array<double, 2, camp::int_seq<long, 1l, 0l>, int, LvArray::ChaiBuffer> Free memory on device: 0.0 B
Freed 147.0 KB to the DEVICE: LvArray::Array<double, 2, camp::int_seq<long, 1l, 0l>, int, LvArray::ChaiBuffer> Free memory on device: 0.0 B
Issue came after this step
Problem/domain/MeshBodies/mesh1/meshLevels/Level0/ElementRegions/elementRegionsGroup/Omega/elementSubRegions/cb1/rock_density
Freed 8.0 B to the HOST : LvArray::Array<int, 1, camp::int_seq<long, 0l>, int, LvArray::ChaiBuffer> Free memory on device: 0.0 B
What is the possible cause of this memory access issue with CUDA? and how to fix it?
The text was updated successfully, but these errors were encountered:
Both TPLs and GEOS can be compiled on Maple GPUs with
clang 17.0.4
.However, only flow only simulations (e.g., SPE 11) can be running there.
Even with serial run, any case (tested five tutorial examples) involving mechanical deformation failed at the first time step with following error:
I tried to use
--trace-data-migration
to trace host-device data migration for running Kirsch problem, which is a simple mechanical example with linear elasticity.Issue came after this step
What is the possible cause of this memory access issue with CUDA? and how to fix it?
The text was updated successfully, but these errors were encountered: