RunCUDASettings.h

Go to the documentation of this file.

/*
Copyright (C) 2024 Scott Grauer-Gray
 
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2 of the License, or
(at your option) any later version.
 
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
GNU General Public License for more details.
 
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
*/
 
#ifndef RUN_CUDA_SETTINGS_H_
#define RUN_CUDA_SETTINGS_H_
 
#include <vector>
#include <array>
#include <set>
#include "RunEval/RunData.h"
 
//set data type used for half-precision with CUDA
#if defined(USE_BFLOAT16_FOR_HALF_PRECISION)
#include <cuda_bf16.h>
using halftype = __nv_bfloat16;
#else
#include <cuda_fp16.h>
using halftype = half;
#endif //USE_BFLOAT16_FOR_HALF_PRECISION
 
namespace run_cuda {
 
constexpr std::string_view kOptimizeCUDADesc{"CUDA"};
constexpr std::string_view kCUDAVersionHeader{"Cuda version"};
constexpr std::string_view kCUDARuntimeHeader{"Cuda Runtime Version"};
 
inline RunData retrieveDeviceProperties(int num_device)
{
  cudaDeviceProp prop;
  std::array<int, 2> cuda_version_driver_runtime;
  cudaGetDeviceProperties(&prop, num_device);
  cudaDriverGetVersion(&(cuda_version_driver_runtime[0]));
  cudaRuntimeGetVersion(&(cuda_version_driver_runtime[1]));
 
  RunData run_data;
  run_data.AddDataWHeader(
    "Device " + std::to_string(num_device),
    std::string(prop.name) + " with " + 
      std::to_string(prop.multiProcessorCount) +
      " multiprocessors");
  run_data.AddDataWHeader(
    std::string(kCUDAVersionHeader),
    std::to_string(cuda_version_driver_runtime[0]));
  run_data.AddDataWHeader(
    std::string(kCUDARuntimeHeader),
    std::to_string(cuda_version_driver_runtime[1]));
  return run_data;
}
 
constexpr std::array<unsigned int, 2> kParallelParamsDefault{{32, 4}};
 
#if defined(DEFAULT_PARALLEL_PARAMS_ONLY)
 
const std::set<std::array<unsigned int, 2>> kParallelParameterAltOptions{};
 
#elif defined(LIMITED_ALT_PARALLEL_PARAMS)
 
const std::set<std::array<unsigned int, 2>> kParallelParameterAltOptions{
  {16, 1}, {32, 4}, {64, 4}};
 
#else
 
const std::set<std::array<unsigned int, 2>> kParallelParameterAltOptions{
  {16, 1}, {32, 1}, {32, 2}, {32, 3}, {32, 4}, {32, 5},
  {32, 6}, {32, 8}, {64, 1}, {64, 2}, {64, 3}, {64, 4},
  {128, 1}, {128, 2}, {256, 1}, {32, 10}, {32, 12}, {32, 14}, {32, 16},
  {64, 5}, {64, 6}, {64, 7}, {64, 8}, {128, 3}, {128, 4}, {256, 2}};
 
#endif
 
};
 
#endif /* RUN_CUDA_SETTINGS_H_ */