Skip to content

Commit

Permalink
Update cpp/include/raft/matrix/detail/select_radix.cuh
Browse files Browse the repository at this point in the history
Fix small typos in the comments
  • Loading branch information
achirkin authored Nov 9, 2023
1 parent baceae1 commit ceb07c3
Showing 1 changed file with 3 additions and 3 deletions.
6 changes: 3 additions & 3 deletions cpp/include/raft/matrix/detail/select_radix.cuh
Original file line number Diff line number Diff line change
Expand Up @@ -681,10 +681,10 @@ int calc_chunk_size(int batch_size, IdxT len, int sm_cnt, Kernel kernel, bool on
RAFT_CUDA_TRY(
cudaOccupancyMaxActiveBlocksPerMultiprocessor(&active_blocks, kernel, BlockSize, 0));

// The chunk size is chose that there is enough workload to fully utilize GPU.
// The chunk size is chosen so that there is enough workload to fully utilize GPU.
// One full wave contains (sm_cnt * active_blocks) blocks, and 10 waves is an empirically safe
// estimation of enough workload. It also counteracts imbalance if some blocks run slow than
// others.
// estimation of enough workload. It also counteracts imbalance if some blocks run slower
// than others.
constexpr int num_waves = 10;
int chunk_size;
if (one_block) {
Expand Down

0 comments on commit ceb07c3

Please sign in to comment.