[GitHub] eric-haibin-lin commented on a change in pull request #11179: [MXNET-404] elemwise_add/sub between rsp and rsp on GPU

[GitBox <gi...@apache.org>]

eric-haibin-lin commented on a change in pull request #11179: [MXNET-404] elemwise_add/sub between rsp and rsp on GPU
URL: https://github.com/apache/incubator-mxnet/pull/11179#discussion_r193908123
 
 

 ##########
 File path: src/operator/tensor/elemwise_binary_op_basic.cu
 ##########
 @@ -22,12 +22,141 @@
  * \file elemwise_binary_scalar_op.cu
  * \brief GPU Implementation of unary function.
  */
+#include <cub/cub.cuh>
 #include "./elemwise_binary_op.h"
 #include "./elemwise_binary_op-inl.h"
 
 namespace mxnet {
 namespace op {
 
+template<typename OP>
+struct RspElemwiseKernel {
+  template<typename DType, typename IType>
+  static MSHADOW_XINLINE void Map(int i, DType* out, const IType* lookup_table,
+                                  const DType* data, const IType* indices,
+                                  const nnvm::dim_t nz_rows, const nnvm::dim_t num_cols) {
+    if (i < nz_rows * num_cols) {
+      const nnvm::dim_t row = i / num_cols;
+      const nnvm::dim_t col = i % num_cols;
+      const nnvm::dim_t out_row = lookup_table[indices[row]] - 1;
+      const nnvm::dim_t out_idx = out_row * num_cols + col;
+      out[out_idx] = OP::Map(out[out_idx], data[i]);
+    }
+  }
+};
+
+template<typename OP>
+void ElemwiseBinaryOp::RspRspOp(mshadow::Stream<gpu> *s,
 
 Review comment:
   do we have unit test for write inplace?

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
users@infra.apache.org


With regards,
Apache Git Services