Skip to content

GitLab

Unverified Commit e395a79c authored by Nicolas Hug's avatar Nicolas Hug Committed by GitHub
Browse files

Fix DeformConvTester::test_backward_cuda by setting threads per block to 512 (#3942)

parent 2cc83594
Hide whitespace changes
Inline Side-by-side
Showing with 1 addition and 4 deletions
torchvision/csrc/ops/cuda/deform_conv2d_kernel.cu
View file @ e395a79c
...@@ -85,10 +85,7 @@ inline unsigned int GET_THREADS() { ...@@ -85,10 +85,7 @@ inline unsigned int GET_THREADS() {
#ifdef __HIP_PLATFORM_HCC__ #ifdef __HIP_PLATFORM_HCC__
return 256; return 256;
#endif #endif
if (at::cuda::getCurrentDeviceProperties()->major >= 6) { return 512;
return 1024;
}
return 512;
} }
inline unsigned int GET_BLOCKS( inline unsigned int GET_BLOCKS(
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment