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

CUTLASS 3.0.0 (#786)

Browse Source 
* CUTLASS 3.0.0
This commit is contained in:
Vijay Thakkar
2023-01-23 17:55:28 -08:00
committed by GitHub
parent 66d9cddc83
commit 277bd6e537
377 changed files with 76396 additions and 1186 deletions
Download Patch File Download Diff File Expand all files Collapse all files

38
include/cutlass/device_kernel.h
View File

@@ -34,7 +34,24 @@
#pragma once
#include "cutlass/cutlass.h"
// __grid_constant__ was introduced in CUDA 11.7.
#if ((__CUDACC_VER_MAJOR__ >= 12) || ((__CUDACC_VER_MAJOR__ == 11) && (__CUDACC_VER_MINOR__ >= 7)))
# define CUTLASS_GRID_CONSTANT_SUPPORTED
#endif
// __grid_constant__ can be enabled only on SM70+
#if defined(CUTLASS_GRID_CONSTANT_SUPPORTED) && defined(__CUDA_ARCH__) && (__CUDA_ARCH__ >= 700)
# define CUTLASS_GRID_CONSTANT_ENABLED
#endif
#if ! defined(CUTLASS_GRID_CONSTANT)
# if defined(CUTLASS_GRID_CONSTANT_ENABLED)
# define CUTLASS_GRID_CONSTANT __grid_constant__
# else
# define CUTLASS_GRID_CONSTANT
# endif
#endif
////////////////////////////////////////////////////////////////////////////////
namespace cutlass {
@@ -75,5 +92,22 @@ void Kernel2(typename Operator::Params params) {
////////////////////////////////////////////////////////////////////////////////
} /// namespace cutlass
//
// 3.0 specific launch
//
////////////////////////////////////////////////////////////////////////////////
/// Generic CUTLASS kernel template.
template <typename Operator>
__global__ __launch_bounds__(Operator::MaxThreadsPerBlock, Operator::MinBlocksPerMultiprocessor)
void device_kernel(CUTLASS_GRID_CONSTANT typename Operator::Params const params)
{
// Dynamic shared memory base pointer
extern __shared__ char smem[];
Operator op;
op(params, smem);
}
////////////////////////////////////////////////////////////////////////////////
} /// namespace cutlass