Fix doc style and unused parameters.

include/ck/tensor_operation/gpu/device/device_grouped_gemm_splitk.hpp

@@ -82,24 +82,26 @@ struct DeviceGroupedGemmSplitK : public DeviceGroupedGemm<ALayout,
                                                           BElementwiseOperation,
                                                           CElementwiseOperation>
 {
-    ///------------------------------------------------------------------------//
+    //----------------------------------------------------------------------------------------------
     /// @brief      Sets the k batch size.
     ///
     /// @param      p_arg   Pointer to the Argument we're going to change.
     /// @param[in]  kbatch  The kbatch value.
     ///
-    virtual void SetKBatchSize(BaseArgument* /*p_arg*/, index_t /*kbatch*/) const {}
+    virtual void SetKBatchSize([[maybe_unused]] BaseArgument* p_arg,
+                               [[maybe_unused]] index_t kbatch) const
+    {
+    }
 
-    ///------------------------------------------------------------------------//
-    ///
+    //----------------------------------------------------------------------------------------------
     /// @brief      Sets the device kernel arguments pointer.
     ///
     /// @param      p_arg              The pointer to the Argument we're going to update.
     /// @param[in]  p_dev_kernel_args  The pointer to the device memory which contains kernel
     ///                                arguments.
     ///
-    virtual void SetDeviceKernelArgs(BaseArgument* /*p_arg*/,
-                                     const void* /*p_dev_kernel_args*/) const
+    virtual void SetDeviceKernelArgs([[maybe_unused]] BaseArgument* p_arg,
+                                     [[maybe_unused]] const void* p_dev_kernel_args) const
     {
     }
 };

include/ck/tensor_operation/gpu/device/impl/device_grouped_gemm_xdl_splitk_cshuffle_tile_loop.hpp

@@ -28,7 +28,7 @@ namespace device {
 ///
 /// @brief      Entry point kernel for device-wide Grouped GEMM operation.
 ///
-/// @param[in]  gemm_desc_const             The pointer to the array of GEMM descriptor structures.
+/// @param[in]  gemm_descs_const            The pointer to the array of GEMM descriptor structures.
 /// @param[in]  tile_count                  The overall number of output tiles we divided all groups
 ///                                         into.
 /// @param[in]  k_batch                     The number of batches we split the K dimension into.

include/ck/tensor_operation/gpu/grid/block_to_ctile_map.hpp

@@ -643,12 +643,12 @@ struct BlockToCTileMap_3DGrid_KSplit
     ///
     /// @brief      Constructs a new instance.
     ///
-    /// @param      <unnamed>  Swallow blockIdx.
+    /// @param[in]  top_idx  Swallow blockIdx.
     ///
     /// @tparam     TopIdx   The type of block index.
     ///
     template <typename TopIdx>
-    __host__ __device__ BlockToCTileMap_3DGrid_KSplit(TopIdx)
+    __host__ __device__ BlockToCTileMap_3DGrid_KSplit([[maybe_unused]] TopIdx top_idx)
     {
     }