Fix the slow cpu reference batched gemm kernels. (#388)

* fix the performance of the batched gemm verification * fix tabs

Fix the slow cpu reference batched gemm kernels. (#388)
* fix the performance of the batched gemm verification * fix tabs
9061d39b · Illia Silin · GitHub · 1e5b59df · 9061d39b
Unverified Commit 9061d39b authored Aug 29, 2022 by Illia Silin Committed by GitHub Aug 29, 2022
Show whitespace changes
Inline Side-by-side

Showing with 2 additions and 2 deletions

library/include/ck/library/reference_tensor_operation/cpu/reference_batched_gemm.hpp ...reference_tensor_operation/cpu/reference_batched_gemm.hpp +2 -2

No files found.
--- a/library/include/ck/library/reference_tensor_operation/cpu/reference_batched_gemm.hpp
+++ b/library/include/ck/library/reference_tensor_operation/cpu/reference_batched_gemm.hpp
@@ -83,8 +83,8 @@ struct ReferenceBatchedGemm : public device::BaseOperator
            make_ParallelTensorFunctor(f_gmk_gkn_gmn,
                                       arg.c_g_m_n_.mDesc.GetLengths()[0],
                                       arg.c_g_m_n_.mDesc.GetLengths()[1],
-                                       arg.c_g_m_n_.mDesc.GetLengths()[2])();
-
+                                       arg.c_g_m_n_.mDesc.GetLengths()[2])(
+                std::thread::hardware_concurrency());
            return 0;
        }