aparler/PyNorch
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Makefile cuda

Browse source
This commit is contained in:
lucasdelimanogueira 2024-04-29 12:50:48 -03:00
parent 83b7b89e7f
commit 35c028dbeb
2 changed files with 5 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

2
build/Makefile
View file

@ -19,7 +19,7 @@ CU_OBJS = $(patsubst $(SRCDIR)/%.cu, $(BUILDDIR)/%.cu.o, $(CU_SRCS))
# CUDA flags and libraries
CUDAFLAGS = -arch=sm_75
CUDALIBS = -lcudart
CUDALIBS = -L/usr/local/cuda-10.2/lib64/ -lcudart -lcuda
# Rule to build the target
$(TARGET): $(OBJS) $(CU_OBJS)

6
csrc/tensor.cpp
View file

@ -2,6 +2,7 @@
#include <stdlib.h>
#include <string.h>
#include <math.h>
#include <cuda_runtime_api.h>
#include "tensor.h"
#include "cuda.h"
#define THREADS_PER_BLOCK 128
@ -76,14 +77,15 @@ extern "C" {
void to_device(Tensor* tensor, char* device) {
printf("Sending tensor to device: %s\n", device);
#ifdef __CUDACC__
if ((strcmp(device, "cuda") == 0) && (strcmp(tensor->device, "cpu") == 0)) {
float* data_tmp;
cudaMalloc((void **)&data_tmp, tensor->size * sizeof(float));
cudaMemcpy(data_tmp, tensor->data, tensor->size * sizeof(float), cudaMemcpyHostToDevice);
free(tensor_data);
free(tensor->data);
tensor->data = data_tmp;
tensor->device = device;
}