Skip to content

Fix: BPCG results not stable when using CUDA AWare MPI #5976

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Mar 9, 2025
Merged

Fix: BPCG results not stable when using CUDA AWare MPI #5976

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 5 additions & 0 deletions source/module_base/para_gemm.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -256,6 +256,11 @@ void PGemmCN<T, Device>::multiply_col(const T alpha, const T* A, const T* B, con
int m = colA_loc[ip];
int size = m * LDA;
MPI_Status status;
#ifdef __CUDA_MPI
// If the memory is not set to zero, it may cause the result to be wrong when using CUDA Aware MPI
// I am not sure if it is due to CUDA Aware MPI or not
base_device::memory::set_memory_op<T, Device>()(Atmp_device, 0, size);
#endif
Parallel_Common::recv_dev<T, Device>(Atmp_device, size, ip, 0, col_world, &status, A_tmp_.data());
MPI_Wait(&requests[ip], &status);
ModuleBase::gemm_op<T, Device>()('C',
Expand Down
5 changes: 5 additions & 0 deletions source/module_hsolver/para_linear_transform.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -120,6 +120,11 @@ void PLinearTransform<T, Device>::act(const T alpha, const T* A, const T* U, con

int size = LDA * ncolA_ip;
MPI_Status status;
#ifdef __CUDA_MPI
// If the memory is not set to zero, it may cause the result to be wrong when using CUDA Aware MPI
// I am not sure if it is due to CUDA Aware MPI or not
base_device::memory::set_memory_op<T, Device>()(Atmp_device, 0, size);
#endif
Parallel_Common::recv_dev<T, Device>(Atmp_device, size, ip, 0, col_world, &status, A_tmp_.data());
ModuleBase::gemm_op<T, Device>()('N',
'N',
Expand Down
Loading