sxjscience commented on a change in pull request #14935: [OP] Accelerate GPU
version of LayerNorm(axis=-1)
URL: https://github.com/apache/incubator-mxnet/pull/14935#discussion_r285342114
##########
File path: src/operator/nn/layer_norm.cu
##########
@@ -24,9 +24,662 @@
*/
#include "./layer_norm-inl.h"
+using namespace mshadow::cuda;
+
namespace mxnet {
namespace op {
+template <typename DType>
+__device__ __forceinline__ DType WARP_SHFL(DType value, int src_lane,
Review comment:
Because it's warp-level primitives and I feel it's clear to set it as
capital-letters. If this is inconsistent, I could change it to be `warp_shfl`
and `warp_shfl_xor`.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
