From c02170731ffb8087e701392ba97270c6b4cec4e8 Mon Sep 17 00:00:00 2001 From: lucasdelimanogueira Date: Thu, 23 May 2024 00:38:33 -0300 Subject: [PATCH] remove atomicAdd test --- norch/csrc/cuda.cu | 2 +- norch/csrc/tensor.cpp | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/norch/csrc/cuda.cu b/norch/csrc/cuda.cu index 6908780..fdf2cb5 100644 --- a/norch/csrc/cuda.cu +++ b/norch/csrc/cuda.cu @@ -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]); } } } diff --git a/norch/csrc/tensor.cpp b/norch/csrc/tensor.cpp index 221ec62..129a747 100644 --- a/norch/csrc/tensor.cpp +++ b/norch/csrc/tensor.cpp @@ -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);