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

Cutlass 1.3 Release (#42)

Browse Source 
CUTLASS 1.3 Release
- Efficient GEMM kernel targeting Volta Tensor Cores via mma.sync instruction added in CUDA 10.1.
This commit is contained in:
Andrew Kerr
2019-03-20 10:49:17 -07:00
committed by GitHub
parent 19a9d64e3c
commit 877bdcace6
256 changed files with 16930 additions and 802 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
examples/06_splitK_gemm/CMakeLists.txt
View File

@ -1,4 +1,4 @@
# Copyright (c) 2017-2018, NVIDIA CORPORATION. All rights reserved.
# Copyright (c) 2017-2019, NVIDIA CORPORATION. All rights reserved.
#
# Redistribution and use in source and binary forms, with or without modification, are permitted
# provided that the following conditions are met:

8
examples/06_splitK_gemm/splitK_gemm.cu
View File

@ -1,5 +1,5 @@
/***************************************************************************************************
* Copyright (c) 2017-2018, NVIDIA CORPORATION. All rights reserved.
* Copyright (c) 2017-2019, NVIDIA CORPORATION. All rights reserved.
*
* Redistribution and use in source and binary forms, with or without modification, are permitted
* provided that the following conditions are met:
@ -84,11 +84,7 @@ cudaError_t cutlass_splitK_sgemm_nn(float const *A,
typename deviceGemm::Params deviceGemmParams(m, n, k);
// query if workspace is needed. the workspace size is sizeof(accumulateType) * M * N * splits_count
int workspace_size = deviceGemmParams.required_workspace_memory_in_byte();
if (workspace_size <= 0) {
std::cerr << "splitK workspace_size is smaller than 0" << std::endl;
return cudaErrorInvalidValue;
}
size_t workspace_size = deviceGemmParams.required_workspace_memory_in_byte();
// allocate workspace memory
float *workspace_ptr;