Program Listing for File gpu_utils.hpp

↰ Return to documentation for file (include/ripple/arch/gpu_utils.hpp)

#ifndef RIPPLE_ARCH_GPU_UTILS_HPP
#define RIPPLE_ARCH_GPU_UTILS_HPP

#include <ripple/utility/portability.hpp>

namespace ripple::gpu {

inline auto set_device(uint32_t device_id) noexcept -> void {
  ripple_check_cuda_result(cudaSetDevice(device_id));
}

inline auto create_stream(GpuStream* stream) noexcept -> void {
  ripple_check_cuda_result(cudaStreamCreate(stream));
}

inline auto create_nonblocking_stream(GpuStream* stream) noexcept -> void {
  ripple_check_cuda_result(
    cudaStreamCreateWithFlags(stream, cudaStreamNonBlocking));
}

inline auto destroy_stream(GpuStream stream) noexcept -> void {
  ripple_check_cuda_result(cudaStreamDestroy(stream));
}

inline auto create_event(GpuStream stream) noexcept -> GpuEvent {
  GpuEvent event = nullptr;
  ripple_check_cuda_result(cudaEventCreate(&event));
  ripple_check_cuda_result(cudaEventRecord(event, stream));
  return event;
}

inline auto record_event(GpuEvent event, GpuStream stream) noexcept -> void {
  ripple_check_cuda_result(cudaEventRecord(event, stream));
}

inline auto destroy_event(GpuEvent event) noexcept -> void {
  ripple_check_cuda_result(cudaEventDestroy(event));
}

inline auto synchronize_stream(GpuStream stream) noexcept -> void {
  ripple_check_cuda_result(cudaStreamSynchronize(stream));
}

inline auto
synchronize_stream(GpuStream stream, GpuEvent event) noexcept -> void {
  ripple_check_cuda_result(cudaStreamWaitEvent(stream, event));
}

inline auto synchronize_device() noexcept -> void {
  ripple_check_cuda_result(cudaDeviceSynchronize());
}

inline auto check_last_error() noexcept -> void {
  ripple_check_cuda_result(cudaGetLastError());
}

} // namespace ripple::gpu

#endif // RIPPLE_ARCH_GPU_UTILS_HPP