Skip to content

Concurrent halt fix on barrier #1810

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

Open
akifcorduk wants to merge 6 commits into
base: main
Choose a base branch
from
Open

Concurrent halt fix on barrier #1810

Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
e004040
revert presolve tolerance
akifcorduk Aug 20, 2026
063458e
Merge branch 'main' of github.com:NVIDIA/cuopt
akifcorduk Aug 21, 2026
222405c
Merge branch 'main' of github.com:NVIDIA/cuopt
akifcorduk Aug 24, 2026
65b1a7c
fix: synchronize cuDSS before concurrent halt
akifcorduk Aug 27, 2026
23d8085
Merge branch 'main' into concurrent_halt_fix
chris-maes Aug 27, 2026
8455505
fix: preserve cuDSS status during concurrent halt
akifcorduk Aug 27, 2026
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
31 changes: 25 additions & 6 deletions cpp/src/barrier/sparse_cholesky.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -483,6 +483,8 @@ class sparse_cholesky_cudss_t : public sparse_cholesky_base_t<i_t, f_t> {
status =
cudssExecute(handle, CUDSS_PHASE_REORDERING, solverConfig, solverData, A, cudss_x, cudss_b);
if (settings_.concurrent_halt != nullptr && *settings_.concurrent_halt == 1) {
RAFT_CUDA_TRY(cudaStreamSynchronize(stream));
handle_ptr_->get_stream().synchronize();
return CONCURRENT_HALT_RETURN;
}
if (status != CUDSS_STATUS_SUCCESS) {
Expand All @@ -499,6 +501,8 @@ class sparse_cholesky_cudss_t : public sparse_cholesky_base_t<i_t, f_t> {
status = cudssExecute(
handle, CUDSS_PHASE_SYMBOLIC_FACTORIZATION, solverConfig, solverData, A, cudss_x, cudss_b);
if (settings_.concurrent_halt != nullptr && *settings_.concurrent_halt == 1) {
RAFT_CUDA_TRY(cudaStreamSynchronize(stream));
handle_ptr_->get_stream().synchronize();
return CONCURRENT_HALT_RETURN;
}
if (status != CUDSS_STATUS_SUCCESS) {
Expand Down Expand Up @@ -556,6 +560,8 @@ class sparse_cholesky_cudss_t : public sparse_cholesky_base_t<i_t, f_t> {
status = cudssExecute(
handle, CUDSS_PHASE_FACTORIZATION, solverConfig, solverData, A, cudss_x, cudss_b);
if (settings_.concurrent_halt != nullptr && *settings_.concurrent_halt == 1) {
RAFT_CUDA_TRY(cudaStreamSynchronize(stream));
handle_ptr_->get_stream().synchronize();
return CONCURRENT_HALT_RETURN;
}
if (status != CUDSS_STATUS_SUCCESS) {
Expand All @@ -572,6 +578,8 @@ class sparse_cholesky_cudss_t : public sparse_cholesky_base_t<i_t, f_t> {

f_t numeric_time = toc(start_numeric);
if (settings_.concurrent_halt != nullptr && *settings_.concurrent_halt == 1) {
RAFT_CUDA_TRY(cudaStreamSynchronize(stream));
handle_ptr_->get_stream().synchronize();
return CONCURRENT_HALT_RETURN;
}

Expand Down Expand Up @@ -694,15 +702,22 @@ class sparse_cholesky_cudss_t : public sparse_cholesky_base_t<i_t, f_t> {
return CONCURRENT_HALT_RETURN;
}
f_t start_analysis = tic();
CUDSS_CALL_AND_CHECK(
cudssExecute(handle, CUDSS_PHASE_REORDERING, solverConfig, solverData, A, cudss_x, cudss_b),
status,
"cudssExecute for reordering");

f_t reorder_time = toc(start_analysis);
status =
cudssExecute(handle, CUDSS_PHASE_REORDERING, solverConfig, solverData, A, cudss_x, cudss_b);
if (settings_.concurrent_halt != nullptr && *settings_.concurrent_halt == 1) {
RAFT_CUDA_TRY(cudaStreamSynchronize(stream));
handle_ptr_->get_stream().synchronize();
Comment thread
coderabbitai[bot] marked this conversation as resolved.
return CONCURRENT_HALT_RETURN;
}
if (status != CUDSS_STATUS_SUCCESS) {
settings_.log.printf(
"FAILED: CUDSS call ended unsuccessfully with status = %d, details: cuDSSExecute for "
"reordering\n",
status);
return -1;
}

f_t reorder_time = toc(start_analysis);

f_t start_symbolic = tic();

Expand Down Expand Up @@ -767,6 +782,8 @@ class sparse_cholesky_cudss_t : public sparse_cholesky_base_t<i_t, f_t> {

f_t numeric_time = toc(start_numeric);
if (settings_.concurrent_halt != nullptr && *settings_.concurrent_halt == 1) {
RAFT_CUDA_TRY(cudaStreamSynchronize(stream));
handle_ptr_->get_stream().synchronize();
return CONCURRENT_HALT_RETURN;
}

Expand Down Expand Up @@ -832,6 +849,8 @@ class sparse_cholesky_cudss_t : public sparse_cholesky_base_t<i_t, f_t> {

status = cudssExecute(handle, CUDSS_PHASE_SOLVE, solverConfig, solverData, A, cudss_x, cudss_b);
if (settings_.concurrent_halt != nullptr && *settings_.concurrent_halt == 1) {
RAFT_CUDA_TRY(cudaStreamSynchronize(stream));
handle_ptr_->get_stream().synchronize();
return CONCURRENT_HALT_RETURN;
}
if (status != CUDSS_STATUS_SUCCESS) {
Expand Down
Loading