leavelet/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

Add residual support for shmem staging iterator used in back-to-back GEMM fusion. This allows support of problem_size_0_n that is not multiple of 32. (#590)

Browse Source 
Co-authored-by: Haicheng Wu <haichengw@nvidia.com>
This commit is contained in:
Haicheng Wu
2022-08-15 11:19:24 -04:00
committed by GitHub
parent e66bfcb1f8
commit 497b499d9d
8 changed files with 462 additions and 35 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
examples/13_two_tensor_op_fusion/kernel/b2b_gemm.h
View File

@@ -341,7 +341,7 @@ struct B2bGemm {
OutputOp0 output_op_0(params.output_op_0);
// Construct thread-scoped matrix multiply
B2bMma b2bMma(shared_storage.main_loop, thread_idx, warp_idx, lane_idx);
B2bMma b2bMma(shared_storage.main_loop, thread_idx, warp_idx, lane_idx, params.problem_size_0.n());
typename B2bMma::FragmentC0 src_accum;
typename B2bMma::FragmentC1 accumulators;