Skip to content

Harden CUDA error checking across the codebase #1994

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
tianleiwu merged 8 commits into from
Mar 1, 2026
Merged

Harden CUDA error checking across the codebase #1994

Show file tree
Hide file tree
Changes from 2 commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
2ea16f1
Initial plan
Copilot Feb 27, 2026
f58bc38
Fix cascading CUDA errors in TopK benchmark by using cudaGetLastError…
Copilot Feb 27, 2026
cd5667c
Changes before error encountered
Copilot Feb 27, 2026
126ab89
Fix cooperative kernel benchmark fallback and stale CUDA error handling
Copilot Feb 27, 2026
764de80
refine
tianleiwu Feb 27, 2026
d2e7de0
Add CUDA error checking to all kernel launches and API calls
tianleiwu Feb 27, 2026
5b0e366
refine
tianleiwu Feb 27, 2026
ce8bab9
Address PR review feedback: FillInput check, redundant check, hybrid_…
tianleiwu Feb 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
14 changes: 0 additions & 14 deletions src/cuda/cuda_common.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -129,19 +129,6 @@ class CudaError : public std::runtime_error {
} \
} while (0)

#ifdef NDEBUG
#define CUDA_CHECK_LAUNCH() \
do { \
cudaError_t err = cudaPeekAtLastError(); \
if (err != cudaSuccess) { \
std::stringstream ss; \
ss << "CUDA launch error in " << __func__ << " at " \
<< __FILE__ << ":" << __LINE__ << " - " \
<< cudaGetErrorString(err); \
throw Generators::CudaError(ss.str(), err); \
Comment thread
tianleiwu marked this conversation as resolved.
} \
} while (0)
#else
#define CUDA_CHECK_LAUNCH() \
do { \
cudaError_t err = cudaGetLastError(); \
Expand All @@ -153,6 +140,5 @@ class CudaError : public std::runtime_error {
throw Generators::CudaError(ss.str(), err); \
} \
} while (0)
#endif

} // namespace Generators
1 change: 1 addition & 0 deletions src/cuda/cuda_topk_benchmark.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -79,6 +79,7 @@ static const char* TopkAlgoToString(TopkAlgo algo) {
best_algo = algo_enum; \
} \
} catch (const Generators::CudaError& e) { \
cudaGetLastError(); \
std::cerr << "Benchmarking failed for " << TopkAlgoToString(algo_enum) \
<< " kernel with k=" << k << ", batch_size=" << batch_size \
<< ", vocab_size=" << vocab_size << ". Error: " << e.what() << std::endl; \
Expand Down
Loading