Enable doxygen XML and fix issues

.gitignore

@@ -91,6 +91,7 @@ thirdparty/googletest/
 
 ## Doxygen
 doxygen/html
+doxygen/xml
 
 #Java
 target

doxygen/Doxyfile

@@ -2012,7 +2012,7 @@ MAN_LINKS              = NO
 # captures the structure of the code including all documentation.
 # The default value is: NO.
 
-GENERATE_XML           = NO
+GENERATE_XML           = YES
 
 # The XML_OUTPUT tag is used to specify where the XML pages will be put. If a
 # relative path is entered the value of OUTPUT_DIRECTORY will be put in front of

include/rmm/cuda_stream_view.hpp

@@ -166,7 +166,7 @@ inline bool operator!=(cuda_stream_view lhs, cuda_stream_view rhs) { return not(
  * @brief Output stream operator for printing / logging streams
  *
  * @param os The output ostream
- * @param sv The cuda_stream_view to output
+ * @param stream The cuda_stream_view to output
  * @return std::ostream& The output ostream
  */
 inline std::ostream& operator<<(std::ostream& os, cuda_stream_view stream)

include/rmm/mr/device/aligned_resource_adaptor.hpp

@@ -140,7 +140,7 @@ class aligned_resource_adaptor final : public device_memory_resource {
    *
    * @throws Nothing.
    *
-   * @param p Pointer to be deallocated
+   * @param ptr Pointer to be deallocated
    * @param bytes Size of the allocation
    * @param stream Stream on which to perform the deallocation
    */

include/rmm/mr/device/arena_memory_resource.hpp

@@ -293,7 +293,8 @@ class arena_memory_resource final : public device_memory_resource {
    * @param stream to execute on.
    * @return std::pair containing free_size and total_size of memory.
    */
-  std::pair<std::size_t, std::size_t> do_get_mem_info(cuda_stream_view) const override
+  std::pair<std::size_t, std::size_t> do_get_mem_info(
+    [[maybe_unused]] cuda_stream_view stream) const override
   {
     return std::make_pair(0, 0);
   }

include/rmm/mr/device/binning_memory_resource.hpp

@@ -182,7 +182,7 @@ class binning_memory_resource final : public device_memory_resource {
    *
    * @throws nothing
    *
-   * @param p Pointer to be deallocated
+   * @param ptr Pointer to be deallocated
    * @param bytes The size in bytes of the allocation. This must be equal to the
    * value of `bytes` that was passed to the `allocate` call that returned `p`.
    * @param stream Stream on which to perform deallocation
@@ -201,7 +201,8 @@ class binning_memory_resource final : public device_memory_resource {
    * @param stream the stream being executed on
    * @return std::pair with available and free memory for resource
    */
-  [[nodiscard]] std::pair<std::size_t, std::size_t> do_get_mem_info(cuda_stream_view) const override
+  [[nodiscard]] std::pair<std::size_t, std::size_t> do_get_mem_info(
+    [[maybe_unused]] cuda_stream_view stream) const override
   {
     return std::make_pair(0, 0);
   }

include/rmm/mr/device/callback_memory_resource.hpp

@@ -100,11 +100,34 @@ class callback_memory_resource final : public device_memory_resource {
     default;  ///< @default_move_assignment{callback_memory_resource}
 
  private:
+  /**
+   * @brief Allocates memory of size at least \p bytes.
+   *
+   * The returned pointer will have at minimum 256 byte alignment.
+   *
+   * If supported by the callback, this operation may optionally be executed on
+   * a stream.  Otherwise, the stream is ignored and the null stream is used.
+   *
+   * @param bytes The size of the allocation
+   * @param stream Stream on which to perform allocation
+   * @return void* Pointer to the newly allocated memory
+   */
   void* do_allocate(std::size_t bytes, cuda_stream_view stream) override
   {
     return allocate_callback_(bytes, stream, allocate_callback_arg_);
   }
 
+  /**
+   * @brief Deallocate memory pointed to by \p p.
+   *
+   * If supported by the callback, this operation may optionally be executed on
+   * a stream.  Otherwise, the stream is ignored and the null stream is used.
+   *
+   * @param ptr Pointer to be deallocated
+   * @param bytes The size in bytes of the allocation. This must be equal to the
+   * value of `bytes` that was passed to the `allocate` call that returned `p`.
+   * @param stream Stream on which to perform deallocation
+   */
   void do_deallocate(void* ptr, std::size_t bytes, cuda_stream_view stream) override
   {
     deallocate_callback_(ptr, bytes, stream, deallocate_callback_arg_);

include/rmm/mr/device/cuda_async_memory_resource.hpp

@@ -173,13 +173,12 @@ class cuda_async_memory_resource final : public device_memory_resource {
 #endif
 
   /**
-   * @brief Allocates memory of size at least `bytes` using cudaMalloc.
+   * @brief Allocates memory of size at least \p bytes.
    *
-   * The returned pointer has at least 256B alignment.
+   * The returned pointer will have at minimum 256 byte alignment.
    *
-   * @throws `rmm::bad_alloc` if the requested allocation could not be fulfilled
-   *
-   * @param bytes The size, in bytes, of the allocation
+   * @param bytes The size of the allocation
+   * @param stream Stream on which to perform allocation
    * @return void* Pointer to the newly allocated memory
    */
   void* do_allocate(std::size_t bytes, rmm::cuda_stream_view stream) override
@@ -197,17 +196,18 @@ class cuda_async_memory_resource final : public device_memory_resource {
   /**
    * @brief Deallocate memory pointed to by \p p.
    *
-   * @throws Nothing.
-   *
-   * @param p Pointer to be deallocated
+   * @param ptr Pointer to be deallocated
+   * @param bytes The size in bytes of the allocation. This must be equal to the
+   * value of `bytes` that was passed to the `allocate` call that returned `p`.
+   * @param stream Stream on which to perform deallocation
    */
-  void do_deallocate(void* ptr, std::size_t size, rmm::cuda_stream_view stream) override
+  void do_deallocate(void* ptr, std::size_t bytes, rmm::cuda_stream_view stream) override
   {
 #ifdef RMM_CUDA_MALLOC_ASYNC_SUPPORT
-    pool_.deallocate(ptr, size, stream);
+    pool_.deallocate(ptr, bytes, stream);
 #else
     (void)ptr;
-    (void)size;
+    (void)bytes;
     (void)stream;
 #endif
   }

include/rmm/mr/device/cuda_async_view_memory_resource.hpp

@@ -108,13 +108,12 @@ class cuda_async_view_memory_resource final : public device_memory_resource {
 #endif
 
   /**
-   * @brief Allocates memory of size at least `bytes` using cudaMalloc.
+   * @brief Allocates memory of size at least \p bytes.
    *
-   * The returned pointer has at least 256B alignment.
+   * The returned pointer will have at minimum 256 byte alignment.
    *
-   * @throws `rmm::bad_alloc` if the requested allocation could not be fulfilled
-   *
-   * @param bytes The size, in bytes, of the allocation
+   * @param bytes The size of the allocation
+   * @param stream Stream on which to perform allocation
    * @return void* Pointer to the newly allocated memory
    */
   void* do_allocate(std::size_t bytes, rmm::cuda_stream_view stream) override
@@ -135,18 +134,22 @@ class cuda_async_view_memory_resource final : public device_memory_resource {
   /**
    * @brief Deallocate memory pointed to by \p p.
    *
-   * @throws Nothing.
-   *
-   * @param p Pointer to be deallocated
+   * @param ptr Pointer to be deallocated
+   * @param bytes The size in bytes of the allocation. This must be equal to the
+   * value of `bytes` that was passed to the `allocate` call that returned `p`.
+   * @param stream Stream on which to perform deallocation
    */
-  void do_deallocate(void* ptr, std::size_t, rmm::cuda_stream_view stream) override
+  void do_deallocate(void* ptr,
+                     [[maybe_unused]] std::size_t bytes,
+                     rmm::cuda_stream_view stream) override
   {
 #ifdef RMM_CUDA_MALLOC_ASYNC_SUPPORT
     if (ptr != nullptr) {
       RMM_ASSERT_CUDA_SUCCESS(rmm::detail::async_alloc::cudaFreeAsync(ptr, stream.value()));
     }
 #else
     (void)ptr;
+    (void)bytes;
     (void)stream;
 #endif
   }

include/rmm/mr/device/cuda_memory_resource.hpp

@@ -55,18 +55,17 @@ class cuda_memory_resource final : public device_memory_resource {
 
  private:
   /**
-   * @brief Allocates memory of size at least `bytes` using cudaMalloc.
+   * @brief Allocates memory of size at least \p bytes.
    *
-   * The returned pointer has at least 256B alignment.
+   * The returned pointer will have at minimum 256 byte alignment.
    *
-   * @note Stream argument is ignored
+   * The stream argument is ignored.
    *
-   * @throws `rmm::bad_alloc` if the requested allocation could not be fulfilled
-   *
-   * @param bytes The size, in bytes, of the allocation
+   * @param bytes The size of the allocation
+   * @param stream This argument is ignored
    * @return void* Pointer to the newly allocated memory
    */
-  void* do_allocate(std::size_t bytes, cuda_stream_view) override
+  void* do_allocate(std::size_t bytes, [[maybe_unused]] cuda_stream_view stream) override
   {
     void* ptr{nullptr};
     RMM_CUDA_TRY_ALLOC(cudaMalloc(&ptr, bytes));
@@ -76,13 +75,16 @@ class cuda_memory_resource final : public device_memory_resource {
   /**
    * @brief Deallocate memory pointed to by \p p.
    *
-   * @note Stream argument is ignored.
-   *
-   * @throws Nothing.
+   * The stream argument is ignored.
    *
-   * @param p Pointer to be deallocated
+   * @param ptr Pointer to be deallocated
+   * @param bytes The size in bytes of the allocation. This must be equal to the
+   * value of `bytes` that was passed to the `allocate` call that returned `p`.
+   * @param stream This argument is ignored.
    */
-  void do_deallocate(void* ptr, std::size_t, cuda_stream_view) override
+  void do_deallocate(void* ptr,
+                     [[maybe_unused]] std::size_t bytes,
+                     [[maybe_unused]] cuda_stream_view stream) override
   {
     RMM_ASSERT_CUDA_SUCCESS(cudaFree(ptr));
   }

include/rmm/mr/device/device_memory_resource.hpp

@@ -201,7 +201,7 @@ class device_memory_resource {
    * If supported, this operation may optionally be executed on a stream.
    * Otherwise, the stream is ignored and the null stream is used.
    *
-   * @param p Pointer to be deallocated
+   * @param ptr Pointer to be deallocated
    * @param bytes The size in bytes of the allocation. This must be equal to the
    * value of `bytes` that was passed to the `allocate` call that returned `p`.
    * @param stream Stream on which to perform deallocation

include/rmm/mr/device/limiting_resource_adaptor.hpp

@@ -185,7 +185,8 @@ class limiting_resource_adaptor final : public device_memory_resource {
    * @param stream Stream on which to get the mem info.
    * @return std::pair contaiing free_size and total_size of memory
    */
-  [[nodiscard]] std::pair<std::size_t, std::size_t> do_get_mem_info(cuda_stream_view) const override
+  [[nodiscard]] std::pair<std::size_t, std::size_t> do_get_mem_info(
+    [[maybe_unused]] cuda_stream_view stream) const override
   {
     return {allocation_limit_ - allocated_bytes_, allocation_limit_};
   }
@@ -209,7 +210,7 @@ class limiting_resource_adaptor final : public device_memory_resource {
  *
  * @tparam Upstream Type of the upstream `device_memory_resource`.
  * @param upstream Pointer to the upstream resource
- * @param limit Maximum amount of memory to allocate
+ * @param allocation_limit Maximum amount of memory to allocate
  */
 template <typename Upstream>
 limiting_resource_adaptor<Upstream> make_limiting_adaptor(Upstream* upstream,

include/rmm/mr/device/logging_resource_adaptor.hpp

@@ -329,6 +329,8 @@ class logging_resource_adaptor final : public device_memory_resource {
  * @param upstream Pointer to the upstream resource
  * @param filename Name of the file to write log info. If not specified,
  * retrieves the log file name from the environment variable "RMM_LOG_FILE".
+ * @param auto_flush If true, flushes the log for every (de)allocation. Warning, this will degrade
+ * performance.
  */
 template <typename Upstream>
 logging_resource_adaptor<Upstream> make_logging_adaptor(
@@ -346,6 +348,8 @@ logging_resource_adaptor<Upstream> make_logging_adaptor(
  * @tparam Upstream Type of the upstream `device_memory_resource`.
  * @param upstream Pointer to the upstream resource
  * @param stream The ostream to write log info.
+ * @param auto_flush If true, flushes the log for every (de)allocation. Warning, this will degrade
+ * performance.
  */
 template <typename Upstream>
 logging_resource_adaptor<Upstream> make_logging_adaptor(Upstream* upstream,

include/rmm/mr/device/managed_memory_resource.hpp

@@ -55,18 +55,17 @@ class managed_memory_resource final : public device_memory_resource {
 
  private:
   /**
-   * @brief Allocates memory of size at least `bytes` using cudaMallocManaged.
+   * @brief Allocates memory of size at least \p bytes.
    *
-   * The returned pointer has at least 256B alignment.
+   * The returned pointer will have at minimum 256 byte alignment.
    *
-   * @note Stream argument is ignored
+   * The stream is ignored.
    *
-   * @throws `rmm::bad_alloc` if the requested allocation could not be fulfilled
-   *
-   * @param bytes The size, in bytes, of the allocation
+   * @param bytes The size of the allocation
+   * @param stream This argument is ignored
    * @return void* Pointer to the newly allocated memory
    */
-  void* do_allocate(std::size_t bytes, cuda_stream_view) override
+  void* do_allocate(std::size_t bytes, [[maybe_unused]] cuda_stream_view stream) override
   {
     // FIXME: Unlike cudaMalloc, cudaMallocManaged will throw an error for 0
     // size allocations.
@@ -78,15 +77,18 @@ class managed_memory_resource final : public device_memory_resource {
   }
 
   /**
-   * @brief Deallocate memory pointed to by `ptr`.
-   *
-   * @note Stream argument is ignored.
+   * @brief Deallocate memory pointed to by \p p.
    *
-   * @throws Nothing.
+   * The stream is ignored.
    *
    * @param ptr Pointer to be deallocated
+   * @param bytes The size in bytes of the allocation. This must be equal to the
+   * value of `bytes` that was passed to the `allocate` call that returned `p`.
+   * @param stream This argument is ignored
    */
-  void do_deallocate(void* ptr, std::size_t, cuda_stream_view) override
+  void do_deallocate(void* ptr,
+                     [[maybe_unused]] std::size_t bytes,
+                     [[maybe_unused]] cuda_stream_view stream) override
   {
     RMM_ASSERT_CUDA_SUCCESS(cudaFree(ptr));
   }
@@ -116,7 +118,8 @@ class managed_memory_resource final : public device_memory_resource {
    * @param stream to execute on
    * @return std::pair contaiing free_size and total_size of memory
    */
-  [[nodiscard]] std::pair<std::size_t, std::size_t> do_get_mem_info(cuda_stream_view) const override
+  [[nodiscard]] std::pair<std::size_t, std::size_t> do_get_mem_info(
+    [[maybe_unused]] cuda_stream_view stream) const override
   {
     std::size_t free_size{};
     std::size_t total_size{};

include/rmm/thrust_rmm_allocator.h

@@ -33,9 +33,9 @@ using exec_policy_t = std::unique_ptr<par_t, deleter_t>;
  * @brief Returns a unique_ptr to a Thrust CUDA execution policy that uses RMM
  * for temporary memory allocation.
  *
- * @Param stream The stream that the allocator will use
+ * @param stream The stream that the allocator will use
  *
- * @Returns A Thrust execution policy that will use RMM for temporary memory
+ * @return A Thrust execution policy that will use RMM for temporary memory
  * allocation.
  */
 [[deprecated("Use new exec_policy in rmm/exec_policy.hpp")]] inline exec_policy_t exec_policy(