[Misc] Replace /*! with /**. (#4823)

* replace

* blabla

* balbla

* blabla
Co-authored-by: Steve <ubuntu@ip-172-31-34-29.ap-northeast-1.compute.internal>

src/array/cpu/gather_mm.h

-/*!
+/**
  *  Copyright (c) 2022 by Contributors
  * @file array/cpu/gather_mm.h
  * @brief GATHER_MM CPU kernel function header.
@@ -45,7 +45,7 @@ void matmul(
   }
 }
 
-/*!
+/**
  * @brief CPU kernel of Gather_mm. The input matrix A is expected to be
  *        sorted according to relation type.
  * @param A The input dense matrix of dimension m x k

src/array/cpu/negative_sampling.cc

-/*!
+/**
  *  Copyright (c) 2021 by Contributors
  * @file array/cpu/negative_sampling.cc
  * @brief Uniform negative sampling on CSR.

src/array/cpu/rowwise_pick.h

-/*!
+/**
  *  Copyright (c) 2020 by Contributors
  * @file array/cpu/rowwise_pick.h
  * @brief Template implementation for rowwise pick operators.

src/array/cpu/rowwise_sampling.cc

-/*!
+/**
  *  Copyright (c) 2020 by Contributors
  * @file array/cpu/rowwise_sampling.cc
  * @brief rowwise sampling

src/array/cpu/rowwise_topk.cc

-/*!
+/**
  *  Copyright (c) 2020 by Contributors
  * @file array/cpu/rowwise_topk.cc
  * @brief rowwise topk

src/array/cpu/sddmm.cc

-/*!
+/**
  *  Copyright (c) 2020 by Contributors
  * @file aten/cpu/sddmm.cc
  * @brief SDDMM C APIs and definitions.
@@ -42,7 +42,7 @@ namespace aten {
   } while (0)
 
 
-/*! @brief Generalized SDDMM on Csr format. */
+/** @brief Generalized SDDMM on Csr format. */
 template <int XPU, typename IdType, typename DType>
 void SDDMMCsr(const std::string& op,
               const BcastOff& bcast,
@@ -59,7 +59,7 @@ void SDDMMCsr(const std::string& op,
   });
 }
 
-/*! @brief Generalized SDDMM on Csr format with Heterograph support. */
+/** @brief Generalized SDDMM on Csr format with Heterograph support. */
 template <int XPU, typename IdType, typename DType>
 void SDDMMCsrHetero(const std::string& op,
               const BcastOff& bcast,
@@ -131,7 +131,7 @@ template void SDDMMCsrHetero<kDGLCPU, int64_t, double>(
     const std::vector<dgl_type_t>& in_eid,
     const std::vector<dgl_type_t>& out_eid);
 
-/*! @brief Generalized SDDMM on Coo format. */
+/** @brief Generalized SDDMM on Coo format. */
 template <int XPU, typename IdType, typename DType>
 void SDDMMCoo(const std::string& op,
               const BcastOff& bcast,
@@ -148,7 +148,7 @@ void SDDMMCoo(const std::string& op,
   });
 }
 
-/*! @brief Generalized SDDMM on Coo format with Heterograph support. */
+/** @brief Generalized SDDMM on Coo format with Heterograph support. */
 template <int XPU, typename IdType, typename DType>
 void SDDMMCooHetero(const std::string& op,
               const BcastOff& bcast,

src/array/cpu/sddmm.h

-/*!
+/**
  *  Copyright (c) 2020 by Contributors
  * @file array/cpu/sddmm.h
  * @brief SDDMM CPU kernel function header.
@@ -16,7 +16,7 @@ namespace dgl {
 namespace aten {
 namespace cpu {
 
-/*!
+/**
  * @brief CPU kernel of g-SDDMM on Csr format.
  * @param bcast Broadcast information.
  * @param csr The Csr matrix.
@@ -68,7 +68,7 @@ void SDDMMCsr(
   });
 }
 
-/*!
+/**
  * @brief CPU kernel of g-SDDMM on Coo format.
  * @param bcast Broadcast information.
  * @param coo The COO matrix.

src/array/cpu/segment_reduce.cc

-/*!
+/**
  *  Copyright (c) 2020 by Contributors
  * @file kernel/cpu/segment_reduce.cc
  * @brief Segment reduce C APIs and definitions.
@@ -11,7 +11,7 @@
 namespace dgl {
 namespace aten {
 
-/*! @brief Segment Reduce operator. */
+/** @brief Segment Reduce operator. */
 template <int XPU, typename IdType, typename DType>
 void SegmentReduce(
     const std::string& op,
@@ -34,7 +34,7 @@ void SegmentReduce(
   }
 }
 
-/*! @brief Scatter Add.*/
+/** @brief Scatter Add.*/
 template <int XPU, typename IdType, typename DType>
 void ScatterAdd(NDArray feat,
                 NDArray idx,
@@ -42,7 +42,7 @@ void ScatterAdd(NDArray feat,
   cpu::ScatterAdd<IdType, DType>(feat, idx, out);
 }
 
-/*! @brief Update gradients for reduce operator max/min on heterogeneous graph.*/
+/** @brief Update gradients for reduce operator max/min on heterogeneous graph.*/
 template <int XPU, typename IdType, typename DType>
 void UpdateGradMinMax_hetero(const HeteroGraphPtr& g,
                 const std::string& op,
@@ -53,7 +53,7 @@ void UpdateGradMinMax_hetero(const HeteroGraphPtr& g,
   cpu::UpdateGradMinMax_hetero<IdType, DType>(g, op, feat, idx, idx_etype, out);
 }
 
-/*! @brief Backward function of segment cmp.*/
+/** @brief Backward function of segment cmp.*/
 template <int XPU, typename IdType, typename DType>
 void BackwardSegmentCmp(
     NDArray feat,

src/array/cpu/segment_reduce.h

-/*!
+/**
  *  Copyright (c) 2020 by Contributors
  * @file array/cpu/spmm.h
  * @brief Segment reduce kernel function header.
@@ -17,7 +17,7 @@ namespace dgl {
 namespace aten {
 namespace cpu {
 
-/*!
+/**
  * @brief CPU kernel of segment sum.
  * @param feat The input tensor.
  * @param offsets The offset tensor storing the ranges of segments.
@@ -42,7 +42,7 @@ void SegmentSum(NDArray feat, NDArray offsets, NDArray out) {
   });
 }
 
-/*!
+/**
  * @brief CPU kernel of segment min/max.
  * @param feat The input tensor.
  * @param offsets The offset tensor storing the ranges of segments.
@@ -76,7 +76,7 @@ void SegmentCmp(NDArray feat, NDArray offsets, NDArray out, NDArray arg) {
   });
 }
 
-/*!
+/**
  * @brief CPU kernel of Scatter Add (on first dimension) operator.
  * @note math equation: out[idx[i], *] += feat[i, *]
  * @param feat The input tensor.
@@ -101,7 +101,7 @@ void ScatterAdd(NDArray feat, NDArray idx, NDArray out) {
   }
 }
 
-/*!
+/**
  * @brief CPU kernel to update gradients for reduce op max/min
  * @param graph The input heterogeneous graph.
  * @param op The binary operator, could be `copy_u`, `copy_e'.
@@ -159,7 +159,7 @@ void UpdateGradMinMax_hetero(
   }
 }
 
-/*!
+/**
  * @brief CPU kernel of backward phase of segment min/max.
  * @note math equation: out[arg[i, k], k] = feat[i, k]
  * @param feat The input tensor.

src/array/cpu/spmat_op_impl_coo.cc

-/*!
+/**
  *  Copyright (c) 2019 by Contributors
  * @file array/cpu/spmat_op_impl.cc
  * @brief CPU implementation of COO sparse matrix operators
@@ -22,7 +22,7 @@ using runtime::parallel_for;
 namespace aten {
 namespace impl {
 
-/*
+/**
  * TODO(BarclayII):
  * For row-major sorted COOs, we have faster implementation with binary search,
  * sorted search, etc.  Later we should benchmark how much we can gain with
@@ -624,7 +624,7 @@ CSRMatrix UnSortedDenseCOOToCSR(const COOMatrix &coo) {
 
 }  // namespace
 
-/*
+/**
 Implementation and Complexity details. N: num_nodes, NNZ: num_edges, P:
 num_threads.
   1. If row is sorted in COO, SortedCOOToCSR<> is applied. Time: O(NNZ/P).

src/array/cpu/spmat_op_impl_csr.cc

-/*!
+/**
  *  Copyright (c) 2019 by Contributors
  * @file array/cpu/spmat_op_impl_csr.cc
  * @brief CSR matrix operator CPU implementation

src/array/cpu/spmm.cc

-/*!
+/**
  *  Copyright (c) 2020 by Contributors
  * @file kernel/cpu/spmm.cc
  * @brief SPMM C APIs and definitions.
@@ -9,7 +9,7 @@
 namespace dgl {
 namespace aten {
 
-/*! @brief Generalized SpMM on Csr format. */
+/** @brief Generalized SpMM on Csr format. */
 template <int XPU, typename IdType, typename DType>
 void SpMMCsr(const std::string& op, const std::string& reduce,
              const BcastOff& bcast,
@@ -41,7 +41,7 @@ void SpMMCsr(const std::string& op, const std::string& reduce,
   }
 }
 
-/*! @brief Generalized SpMM on Csr format. */
+/** @brief Generalized SpMM on Csr format. */
 template <int XPU, typename IdType, typename DType>
 void SpMMCsrHetero(const std::string& op, const std::string& reduce,
              const BcastOff& bcast,
@@ -159,7 +159,7 @@ template void SpMMCsrHetero<kDGLCPU, int64_t, double>(
     const std::vector<dgl_type_t>& ufeat_node_tids,
     const std::vector<dgl_type_t>& out_node_tids);
 
-/*! @brief Edge_softmax_csr forward op on Csr format. */
+/** @brief Edge_softmax_csr forward op on Csr format. */
 template <int XPU, typename IdType, typename DType>
 void Edge_softmax_csr_forward(const std::string& op,
              const BcastOff& bcast,
@@ -172,7 +172,7 @@ void Edge_softmax_csr_forward(const std::string& op,
   });
 }
 
-/*! @brief Edge_softmax_csr backward op on Csr format. */
+/** @brief Edge_softmax_csr backward op on Csr format. */
 template <int XPU, typename IdType, typename DType>
 void Edge_softmax_csr_backward(const std::string& op,
              const BcastOff& bcast,
@@ -219,7 +219,7 @@ template void Edge_softmax_csr_backward<kDGLCPU, int64_t, double>(
     const BcastOff& bcast, const CSRMatrix& csr,
     NDArray ufeat, NDArray efeat, NDArray out);
 
-/*! @brief Generalized SpMM on Coo format. */
+/** @brief Generalized SpMM on Coo format. */
 template <int XPU, typename IdType, typename DType>
 void SpMMCoo(const std::string& op, const std::string& reduce,
              const BcastOff& bcast,

src/array/cpu/spmm.h

-/*!
+/**
  *  Copyright (c) 2020 by Contributors
  * @file array/cpu/spmm.h
  * @brief SPMM CPU kernel function header.
@@ -32,7 +32,7 @@ namespace cpu {
 
 #if !defined(_WIN32)
 #ifdef USE_AVX
-/*!
+/**
  * @brief CPU kernel of SpMM on Csr format using Xbyak.
  * @param cpu_spec JIT'ed kernel
  * @param bcast Broadcast information.
@@ -69,7 +69,7 @@ void SpMMSumCsrXbyak(
 #endif  // USE_AVX
 #endif  // _WIN32
 
-/*!
+/**
  * @brief Naive CPU kernel of SpMM on Csr format.
  * @param cpu_spec JIT'ed kernel
  * @param bcast Broadcast information.
@@ -110,7 +110,7 @@ void SpMMSumCsrNaive(
   });
 }
 
-/*!
+/**
  * @brief CPU kernel of SpMM on Csr format.
  * @param bcast Broadcast information.
  * @param csr The Csr matrix.
@@ -176,7 +176,7 @@ void SpMMSumCsr(
 #endif  // _WIN32
 }
 
-/*!
+/**
  * @brief CPU kernel of SpMM on Coo format.
  * @param bcast Broadcast information.
  * @param coo The Coo matrix.
@@ -225,7 +225,7 @@ void SpMMSumCoo(
   }
 }
 
-/*!
+/**
  * @brief CPU kernel of SpMM-Min/Max on Csr format.
  * @param bcast Broadcast information.
  * @param csr The Csr matrix.
@@ -322,7 +322,7 @@ void SpMMCmpCsr(
 #endif  // _WIN32
 }
 
-/*!
+/**
  * @brief CPU kernel of SpMM-Min/Max on Csr format.
  * @param bcast Broadcast information.
  * @param csr The Csr matrix.
@@ -418,7 +418,7 @@ void SpMMCmpCsrHetero(
   });
 }
 
-/*!
+/**
  * @brief CPU kernel of SpMM-Min/Max on Coo format.
  * @param bcast Broadcast information.
  * @param coo The Coo matrix.
@@ -484,7 +484,7 @@ void SpMMCmpCoo(
   }
 }
 
-/*!
+/**
  * @brief CPU kernel of Edge_softmax_csr_forward on Csr format.
  * @param bcast Broadcast information.
  * @param csr The Csr matrix.
@@ -532,7 +532,7 @@ void Edge_softmax_csr_forward(
   });
 }
 
-/*!
+/**
  * @brief CPU kernel of Edge_softmax_csr_backward on Csr format.
  * @param bcast Broadcast information.
  * @param csr The Csr matrix.

src/array/cpu/spmm_binary_ops.h

-/*!
+/**
  *  Copyright (c) 2020 by Contributors
  * @file array/cpu/spmm_binary_ops.h
  * @brief SPMM CPU Binary ops.

src/array/cpu/spmm_blocking_libxsmm.h

-/*!
+/**
  *  Copyright (c) 2021 Intel Corporation
  * @file array/cpu/spmm.h
  * @brief SPMM CPU kernel function header.
@@ -48,7 +48,7 @@ int32_t GetLLCSize() {
   return cache_size;
 }
 
-/*!
+/**
  * @brief Tile the CSR matrix to roughly make sure that the column tiles and
  *        corresponding neighbor features fit into LLC and the row tiles
  *        are assigned to OMP threads.
@@ -165,7 +165,7 @@ inline void SpMMCreateBlocks(
   }
 }
 
-/*!
+/**
  * @brief Create libxsmm kernel.
  * @param has_idx For the edge features, are there indices available.
  * @param N Feature size.
@@ -269,7 +269,7 @@ inline libxsmm_meltwfunction_opreduce_vecs_idx SpMMCreateLibxsmmKernel(
   return kernel;
 }
 
-/*!
+/**
  * @brief Use libxsmm to perform SpMM-Sum on all blocks.
  * @param block_csr_array The array containing csr matrices of all blocks.
  * @param B The feature on source nodes.
@@ -326,7 +326,7 @@ inline void SpMMBlockwiseOpSum(
   }
 }
 
-/*!
+/**
  * @brief Use libxsmm to perform SpMM-Max/Min on all blocks.
  * @param block_csr_array The array containing csr matrices of all blocks.
  * @param B The feature on source nodes.
@@ -390,7 +390,7 @@ inline void SpMMBlockwiseOpCmp(
   }
 }
 
-/*!
+/**
  * @brief Free the tiled CSR matrix data.
  * @param block_csr_array The array containing csr matrices of all blocks.
  * @param num_M_blocks Number of blocks to create along the rows of adjacency
@@ -412,7 +412,7 @@ inline void SpMMFreeBlocks(
   free(block_csr_array);
 }
 
-/*!
+/**
  * @brief Optimized CPU kernel of SpMM-Sum/Max/Min on Csr format.
  * @param bcast Broadcast information.
  * @param csr The Csr matrix.
@@ -550,7 +550,7 @@ void SpMMRedopCsrOpt(
 #endif  // DEBUG
 }
 
-/*!
+/**
  * @brief Optimized CPU kernel of SpMM-Sum on Csr format.
  * @param bcast Broadcast information.
  * @param csr The Csr matrix.
@@ -568,7 +568,7 @@ void SpMMSumCsrLibxsmm(
       bcast, csr, ufeat, efeat, out, dummy, dummy);
 }
 
-/*!
+/**
  * @brief Optimized CPU kernel of SpMM-Min/Max on Csr format.
  * @param bcast Broadcast information.
  * @param csr The Csr matrix.

src/array/cpu/traversal.cc

-/*!
+/**
  *  Copyright (c) 2020 by Contributors
  * @file array/cpu/traversal.cc
  * @brief Graph traversal implementation

src/array/cpu/traversal.h

-/*!
+/**
  *  Copyright (c) 2020 by Contributors
  * @file array/cpu/traversal.h
  * @brief Graph traversal routines.
@@ -21,7 +21,7 @@ namespace dgl {
 namespace aten {
 namespace impl {
 
-/*!
+/**
  * @brief Traverse the graph in a breadth-first-search (BFS) order.
  *
  * The queue object must suffice following interface:
@@ -85,7 +85,7 @@ void BFSTraverseNodes(
   }
 }
 
-/*!
+/**
  * @brief Traverse the graph in a breadth-first-search (BFS) order, returning
  *        the edges of the BFS tree.
  *
@@ -153,7 +153,7 @@ void BFSTraverseEdges(
   }
 }
 
-/*!
+/**
  * @brief Traverse the graph in topological order.
  *
  * The queue object must suffice following interface:
@@ -226,13 +226,13 @@ void TopologicalNodes(
   }
 }
 
-/*!\brief Tags for ``DFSEdges``. */
+/** @brief Tags for ``DFSEdges``. */
 enum DFSEdgeTag {
   kForward = 0,
   kReverse,
   kNonTree,
 };
-/*!
+/**
  * @brief Traverse the graph in a depth-first-search (DFS) order.
  *
  * The traversal visit edges in its DFS order. Edges have three tags:

src/array/cuda/array_cumsum.cu

-/*!
+/**
  *  Copyright (c) 2020 by Contributors
  * @file array/cpu/array_cumsum.cu
  * @brief Array cumsum GPU implementation

src/array/cuda/array_index_select.cu

-/*!
+/**
  *  Copyright (c) 2019 by Contributors
  * @file array/cpu/array_index_select.cu
  * @brief Array index select GPU implementation

src/array/cuda/array_index_select.cuh

-/*!
+/**
  *  Copyright (c) 2021-2022 by Contributors
  * @file array/cuda/array_index_select.cuh
  * @brief Array index select GPU kernel implementation