RunCUDASettings.h File Reference

Contains namespace with constants and functions to get CUDA device properties as well as default and test parallel parameters to use in CUDA implementation run optimization. More...

#include <vector>
#include <array>
#include <set>
#include "RunEval/RunData.h"
#include <cuda_fp16.h>

Include dependency graph for RunCUDASettings.h:

This graph shows which files directly or indirectly include this file:

Go to the source code of this file.

Namespaces
	run_cuda
	Namespace with constants and functions to get CUDA device properties as well as default and test parallel parameters to use in CUDA implementation run optimization.

Typedefs
using	halftype = half

Functions
RunData	run_cuda::retrieveDeviceProperties (int num_device)

Variables
constexpr std::string_view	run_cuda::kOptimizeCUDADesc {"CUDA"}
constexpr std::string_view	run_cuda::kCUDAVersionHeader {"Cuda version"}
constexpr std::string_view	run_cuda::kCUDARuntimeHeader {"Cuda Runtime Version"}
constexpr std::array< unsigned int, 2 >	run_cuda::kParallelParamsDefault {{32, 4}}
	Default thread block dimensions (which is what parallel parameters corresponds to in CUDA implementation) More...
const std::set< std::array< unsigned int, 2 > >	run_cuda::kParallelParameterAltOptions
	Parallel parameter alternative options to run to retrieve optimized parallel parameters in CUDA implementation Parallel parameter corresponds to thread block dimensions in CUDA implementation OK to include default parallel parameters in alternative options but not required. More...

Detailed Description

Contains namespace with constants and functions to get CUDA device properties as well as default and test parallel parameters to use in CUDA implementation run optimization.

Author

Scott Grauer-Gray

Copyright

Copyright (c) 2024

Definition in file RunCUDASettings.h.

Typedef Documentation

halftype

using halftype = half

Definition at line 43 of file RunCUDASettings.h.