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

remove atomicAdd test

Browse source
This commit is contained in:
lucasdelimanogueira 2024-05-23 00:38:33 -03:00
parent 192a50f773
commit c02170731f
2 changed files with 2 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

2
norch/csrc/cuda.cu
View file

@ -152,7 +152,7 @@ __global__ void sum_tensor_cuda_kernel_axis(float* data, float* result_data, int
for (int i = 0; i < strides[target_axis]; ++i) {
int index = offset + i * strides[target_axis + 1];
atomicAdd(&result_data[outer_index * inner_size + inner_index], data[index]);
//atomicAdd(&result_data[outer_index * inner_size + inner_index], data[index]);
}
}
}

2
norch/csrc/tensor.cpp
View file

@ -217,7 +217,7 @@ extern "C" {
int target_axis_stride = tensor->strides[axis];
int outer_size = tensor->size / target_axis_stride;
cudaMalloc((void**)&result_data, (outer_size) * sizeof(float));
cudaMalloc((void**)&result_data, outer_size * sizeof(float));
}
sum_tensor_cuda(tensor, result_data, axis);