Skip to content

[Tutorial] Fix race condition in 2SM TMEM alloc for Cute Blackwell Tutorial 04/05 #3316

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
pchen7e2 wants to merge 1 commit into
base: main
Choose a base branch
from
+6 −0
Open

[Tutorial] Fix race condition in 2SM TMEM alloc for Cute Blackwell Tutorial 04/05 #3316

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
3 changes: 3 additions & 0 deletions examples/cute/tutorial/blackwell/04_mma_tma_2sm_sm100.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -243,6 +243,9 @@ gemm_device(ATensor mA, // (Gemm_M, Gemm_K)
using TmemAllocator = cute::TMEM::Allocator2Sm;
TmemAllocator tmem_allocator{};

// Make sure all CTAs in Cluster are alive. 2SM mode TMEM alloc requires the leader and peer CTA have both
// been launched. We need this explicit cluster sync here to guarantee that.
cute::cluster_sync();
if (elect_one_warp) {
tmem_allocator.allocate(TmemAllocator::Sm100TmemCapacityColumns, &shared_storage.tmem_base_ptr);
}
Expand Down
3 changes: 3 additions & 0 deletions examples/cute/tutorial/blackwell/05_mma_tma_epi_sm100.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -256,6 +256,9 @@ gemm_device(ATensor mA, // (Gemm_M, Gemm_K)
using TmemAllocator = cute::TMEM::Allocator2Sm;
TmemAllocator tmem_allocator{};

// Make sure all CTAs in Cluster are alive. 2SM mode TMEM alloc requires the leader and peer CTA have both
// been launched. We need this explicit cluster sync here to guarantee that.
cute::cluster_sync();
if (elect_one_warp) {
tmem_allocator.allocate(TmemAllocator::Sm100TmemCapacityColumns, &shared_storage.tmem_base_ptr);
}
Expand Down