remove inline

Author: ChaiBapchya

src/operator/tensor/matrix_op-inl.h

@@ -263,7 +263,7 @@ struct TransposeParam : public dmlc::Parameter<TransposeParam> {
 };
 
 template<typename DType, typename xpu>
-inline void Transpose2D(const DType *in, DType *out, index_t row, index_t col);
+void Transpose2D(const DType *in, DType *out, index_t row, index_t col);
 
 template<typename xpu>
 void TransposeImpl(RunContext ctx,

src/operator/tensor/matrix_op.cc

@@ -313,7 +313,7 @@ inline static bool TransposeStorageType(const nnvm::NodeAttrs& attrs,
    * \param col shape of dim 1 of input
    */
   template<typename DType, typename cpu>
-  inline void Transpose2D(const DType *in, DType *out, index_t row, index_t col) {
+  void Transpose2D(const DType *in, DType *out, index_t row, index_t col) {
     // ensure cache line hits and prevent cache miss for any configuration
     // L1 cache size to be utilized = 32kb = 2^15
     // Largest size of a single unit of any dtype <= 8 byte = 2^3