fix doxygen warnings

cub/block/block_adjacent_difference.cuh

@@ -550,7 +550,7 @@ public:
      * @param[in] difference_op
      *   Binary difference operator
      *
-     * @param[in]
+     * @param[in] valid_items
      *   Number of valid items in thread block
      */
     template <int ITEMS_PER_THREAD,

cub/block/block_shuffle.cuh

@@ -183,7 +183,7 @@ public:
      */
     __device__ __forceinline__ void Rotate(
         T   input,                  ///< [in] The calling thread's input item
-        T&  output,                 ///< [out] The \p input item from thread <em>thread</em><sub>(<em>i</em>+<tt>distance></tt>)%<tt><BLOCK_THREADS></tt></sub> (may be aliased to \p input).  This value is not updated for <em>thread</em><sub>BLOCK_THREADS-1</sub>
+        T&  output,                 ///< [out] The \p input item from thread <em>thread</em><sub>(<em>i</em>+<tt>distance></tt>)%<tt>BLOCK_THREADS</tt></sub> (may be aliased to \p input).  This value is not updated for <em>thread</em><sub>BLOCK_THREADS-1</sub>
         unsigned int distance = 1)  ///< [in] Offset distance (0 < \p distance < <tt>BLOCK_THREADS</tt>)
     {
         temp_storage[linear_tid] = input;

cub/device/device_partition.cuh

@@ -548,6 +548,24 @@ struct DevicePartition
      *   `d_num_selected_out[0]` and total number of items selected by
      *   @p select_second_part_op is stored as `d_num_selected_out[1]`,
      *   respectively
+     *
+     * @param[in] num_items
+     *   Total number of items to select from
+     *
+     * @param[in] select_first_part_op
+     *   Unary selection operator to select @p d_first_part_out
+     *
+     * @param[in] select_second_part_op
+     *   Unary selection operator to select @p d_second_part_out
+     *
+     * @param[in] stream
+     *   **[optional]** CUDA stream to launch kernels within.
+     *   Default is stream<sub>0</sub>.
+     *
+     * @param[in] debug_synchronous
+     *   **[optional]** Whether or not to synchronize the stream after every
+     *   kernel launch to check for errors. May cause significant slowdown.
+     *   Default is @p false.
      */
     template <typename InputIteratorT,
               typename FirstOutputIteratorT,

cub/device/device_scan.cuh

@@ -509,7 +509,7 @@ struct DeviceScan
      * \tparam KeysInputIteratorT      <b>[inferred]</b> Random-access input iterator type for reading scan keys inputs \iterator
      * \tparam ValuesInputIteratorT    <b>[inferred]</b> Random-access input iterator type for reading scan values inputs \iterator
      * \tparam ValuesOutputIteratorT   <b>[inferred]</b> Random-access output iterator type for writing scan values outputs \iterator
-     * \tparam EqualityOpT             <b>[inferred][/b] Functor type having member <tt>T operator()(const T &a, const T &b)</tt> for binary operations that defines the equality of keys
+     * \tparam EqualityOpT             <b>[inferred]</b> Functor type having member <tt>T operator()(const T &a, const T &b)</tt> for binary operations that defines the equality of keys
      *
      * [decoupled look-back]: https://research.nvidia.com/publication/single-pass-parallel-prefix-scan-decoupled-look-back
      */
@@ -625,7 +625,7 @@ struct DeviceScan
      * \tparam ValuesOutputIteratorT   <b>[inferred]</b> Random-access output iterator type for writing scan values outputs \iterator
      * \tparam ScanOp                  <b>[inferred]</b> Binary scan functor type having member <tt>T operator()(const T &a, const T &b)</tt>
      * \tparam InitValueT              <b>[inferred]</b> Type of the \p init_value value used in Binary scan functor type having member <tt>T operator()(const T &a, const T &b)</tt>
-     * \tparam EqualityOpT             <b>[inferred][/b] Functor type having member <tt>T operator()(const T &a, const T &b)</tt> for binary operations that defines the equality of keys
+     * \tparam EqualityOpT             <b>[inferred]</b> Functor type having member <tt>T operator()(const T &a, const T &b)</tt> for binary operations that defines the equality of keys
      *
      * [decoupled look-back]: https://research.nvidia.com/publication/single-pass-parallel-prefix-scan-decoupled-look-back
      */
@@ -711,7 +711,7 @@ struct DeviceScan
      * \tparam KeysInputIteratorT      <b>[inferred]</b> Random-access input iterator type for reading scan keys inputs \iterator
      * \tparam ValuesInputIteratorT    <b>[inferred]</b> Random-access input iterator type for reading scan values inputs \iterator
      * \tparam ValuesOutputIteratorT   <b>[inferred]</b> Random-access output iterator type for writing scan values outputs \iterator
-     * \tparam EqualityOpT             <b>[inferred][/b] Functor type having member <tt>T operator()(const T &a, const T &b)</tt> for binary operations that defines the equality of keys
+     * \tparam EqualityOpT             <b>[inferred]</b> Functor type having member <tt>T operator()(const T &a, const T &b)</tt> for binary operations that defines the equality of keys
      *
      * [decoupled look-back]: https://research.nvidia.com/publication/single-pass-parallel-prefix-scan-decoupled-look-back
      */
@@ -818,7 +818,7 @@ struct DeviceScan
      * \tparam ValuesInputIteratorT    <b>[inferred]</b> Random-access input iterator type for reading scan values inputs \iterator
      * \tparam ValuesOutputIteratorT   <b>[inferred]</b> Random-access output iterator type for writing scan values outputs \iterator
      * \tparam ScanOp                  <b>[inferred]</b> Binary scan functor type having member <tt>T operator()(const T &a, const T &b)</tt>
-     * \tparam EqualityOpT             <b>[inferred][/b] Functor type having member <tt>T operator()(const T &a, const T &b)</tt> for binary operations that defines the equality of keys
+     * \tparam EqualityOpT             <b>[inferred]</b> Functor type having member <tt>T operator()(const T &a, const T &b)</tt> for binary operations that defines the equality of keys
      *
      * [decoupled look-back]: https://research.nvidia.com/publication/single-pass-parallel-prefix-scan-decoupled-look-back
      */

cub/util_device.cuh

@@ -613,7 +613,7 @@ cudaError_t MaxSmOccupancy(
     int&                max_sm_occupancy,          ///< [out] maximum number of thread blocks that can reside on a single SM
     KernelPtr           kernel_ptr,                 ///< [in] Kernel pointer for which to compute SM occupancy
     int                 block_threads,              ///< [in] Number of threads per thread block
-    int                 dynamic_smem_bytes = 0)
+    int                 dynamic_smem_bytes = 0)	    ///< [in] Dynamically allocated shared memory in bytes. Default is 0.
 {
 #ifndef CUB_RUNTIME_ENABLED
 

cub/version.cuh

@@ -25,7 +25,7 @@
  *
  ******************************************************************************/
 
-/*! \file version.h
+/*! \file version.cuh
  *  \brief Compile-time macros encoding CUB release version
  *
  *         <cub/version.h> is the only CUB header that is guaranteed to

cub/warp/warp_load.cuh

@@ -658,7 +658,7 @@ public:
    *                                           thread_data,
    *                                           valid_items,
    *                                           -1);
-   * @endcod
+   * @endcode
    * @par
    * Suppose the input @p d_data is <tt>0, 1, 2, 3, 4, 5, ...</tt>, @p valid_items
    * is @p 5, and the out-of-bounds default is @p -1.