[Mlir-commits] [mlir] 97678ce - [mlir][sparse][gpu] remove zero init memset

[Aart Bik]

Author: Aart Bik
Date: 2023-07-13T18:22:21-07:00
New Revision: 97678cec1b9de5023e1eb5c9804cb137351eb21e

URL: https://github.com/llvm/llvm-project/commit/97678cec1b9de5023e1eb5c9804cb137351eb21e
DIFF: https://github.com/llvm/llvm-project/commit/97678cec1b9de5023e1eb5c9804cb137351eb21e.diff

LOG: [mlir][sparse][gpu] remove zero init memset

avoids quite a big memory fill for each setup

Reviewed By: K-Wu

Differential Revision: https://reviews.llvm.org/D155251

Added: 
    

Modified: 
    mlir/lib/ExecutionEngine/CudaRuntimeWrappers.cpp

Removed: 
    


################################################################################
diff  --git a/mlir/lib/ExecutionEngine/CudaRuntimeWrappers.cpp b/mlir/lib/ExecutionEngine/CudaRuntimeWrappers.cpp
index b5dbc5144ddd7a..ec822125eda456 100644
--- a/mlir/lib/ExecutionEngine/CudaRuntimeWrappers.cpp
+++ b/mlir/lib/ExecutionEngine/CudaRuntimeWrappers.cpp
@@ -528,8 +528,6 @@ mgpuCreateCuSparseLtDnMat(void *dh, intptr_t rows, intptr_t cols, void *values,
                           int32_t dtp, CUstream /*stream*/) {
   assert(cusparseLt_initiated && "client did not call mgpuCreateSparseLtEnv()");
   auto dnmat_handle = reinterpret_cast<cusparseLtDnMatHandleAndData *>(dh);
-  // CusparseLt expects the descriptors to be zero-initialized.
-  memset(dnmat_handle, 0, sizeof(cusparseLtDnMatHandleAndData));
   dnmat_handle->values = values;
   auto dTp = static_cast<cudaDataType_t>(dtp);
   // Assume row-major when deciding lda.
@@ -550,8 +548,6 @@ mgpuCusparseLtCreate2To4SpMat(void *sh, intptr_t rows, intptr_t cols,
                               void *values, int32_t dtp, CUstream /*stream*/) {
   assert(cusparseLt_initiated && "client did not call mgpuCreateSparseLtEnv()");
   auto spmat_handle = reinterpret_cast<cusparseLtSpMatHandleAndData *>(sh);
-  // CusparseLt expects the descriptors to be zero-initialized.
-  memset(spmat_handle, 0, sizeof(cusparseLtSpMatHandleAndData));
   spmat_handle->values = values;
   auto dTp = static_cast<cudaDataType_t>(dtp);
   // Assume row-major when deciding lda.