62 lines
1.8 KiB
Plaintext
62 lines
1.8 KiB
Plaintext
// Basic cuda functions callable from C/C++ code
|
|
#include <cuda.h>
|
|
#include <stdio.h>
|
|
|
|
extern "C" int ScaLBL_SetDevice(int rank){
|
|
int n_devices;
|
|
//int local_rank = atoi(getenv("MV2_COMM_WORLD_LOCAL_RANK"));
|
|
cudaGetDeviceCount(&n_devices);
|
|
//int device = local_rank % n_devices;
|
|
int device = rank % n_devices;
|
|
cudaSetDevice(device);
|
|
return device;
|
|
}
|
|
|
|
extern "C" void ScaLBL_AllocateDeviceMemory(void** address, size_t size){
|
|
cudaMalloc(address,size);
|
|
cudaError_t err = cudaGetLastError();
|
|
if (cudaSuccess != err){
|
|
printf("Error in cudaMalloc: %s \n",cudaGetErrorString(err));
|
|
}
|
|
}
|
|
|
|
extern "C" void ScaLBL_FreeDeviceMemory(void* pointer){
|
|
cudaFree(pointer);
|
|
}
|
|
|
|
extern "C" void ScaLBL_CopyToDevice(void* dest, const void* source, size_t size){
|
|
cudaMemcpy(dest,source,size,cudaMemcpyHostToDevice);
|
|
cudaError_t err = cudaGetLastError();
|
|
if (cudaSuccess != err){
|
|
printf("Error in cudaMemcpy (host->device): %s \n",cudaGetErrorString(err));
|
|
}
|
|
}
|
|
|
|
extern "C" void ScaLBL_AllocateZeroCopy(void** address, size_t size){
|
|
cudaMallocHost(address,size);
|
|
//cudaMalloc(address,size);
|
|
cudaError_t err = cudaGetLastError();
|
|
if (cudaSuccess != err){
|
|
printf("Error in cudaMallocHost: %s \n",cudaGetErrorString(err));
|
|
}
|
|
}
|
|
|
|
extern "C" void ScaLBL_CopyToZeroCopy(void* dest, const void* source, size_t size){
|
|
//cudaMemcpy(dest,source,size,cudaMemcpyHostToDevice);
|
|
//cudaError_t err = cudaGetLastError();
|
|
memcpy(dest, source, size);
|
|
|
|
}
|
|
|
|
extern "C" void ScaLBL_CopyToHost(void* dest, const void* source, size_t size){
|
|
cudaMemcpy(dest,source,size,cudaMemcpyDeviceToHost);
|
|
cudaError_t err = cudaGetLastError();
|
|
if (cudaSuccess != err){
|
|
printf("Error in cudaMemcpy (device->host): %s \n",cudaGetErrorString(err));
|
|
}
|
|
}
|
|
|
|
extern "C" void ScaLBL_DeviceBarrier(){
|
|
cudaDeviceSynchronize();
|
|
}
|