Skip to content

GitLab

Commit a640c63b authored by ashishfarmer's avatar ashishfarmer
Browse files

fix launch bounds for cleanup

parent f2fcce58
Hide whitespace changes
Inline Side-by-side
Showing with 10 additions and 2 deletions
csrc/multi_tensor_l2norm_kernel.cu
View file @ a640c63b
...@@ -194,7 +194,11 @@ struct MaxNormFunctor ...@@ -194,7 +194,11 @@ struct MaxNormFunctor
}; };
__global__ void cleanup( __global__ void
#ifdef __HIP_PLATFORM_HCC__
__launch_bounds__(1024)
#endif
cleanup(
float* output, float* output,
float* output_per_tensor, float* output_per_tensor,
float* ret, float* ret,
...@@ -231,7 +235,11 @@ __global__ void cleanup( ...@@ -231,7 +235,11 @@ __global__ void cleanup(
} }
} }
__global__ void cleanup_v2( __global__ void
#ifdef __HIP_PLATFORM_HCC__
__launch_bounds__(1024)
#endif
cleanup_v2(
float* output, float* output,
float* output_per_tensor, float* output_per_tensor,
float* ret, float* ret,
......
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