From c8d60c1a2c81f4f5b312bf0d0649f9b5a4f7a9c1 Mon Sep 17 00:00:00 2001
From: Vyas Ramasubramani <vyasr@nvidia.com>
Date: Mon, 7 Aug 2023 10:31:04 -0700
Subject: [PATCH 01/17] Add docs to envs

---
 dependencies.yaml | 1 +
 1 file changed, 1 insertion(+)

diff --git a/dependencies.yaml b/dependencies.yaml
index 765a812cb..68020fd0a 100644
--- a/dependencies.yaml
+++ b/dependencies.yaml
@@ -12,6 +12,7 @@ files:
       - develop
       - docs
       - run
+      - docs
       - test_python
   test_python:
     output: none

From ca7c3875895d332bc22a58868133a501fa2a193f Mon Sep 17 00:00:00 2001
From: Vyas Ramasubramani <vyasr@nvidia.com>
Date: Mon, 7 Aug 2023 10:32:02 -0700
Subject: [PATCH 02/17] Add breathe and alphabetize

---
 conda/environments/all_cuda-118_arch-x86_64.yaml | 1 +
 conda/environments/all_cuda-120_arch-x86_64.yaml | 1 +
 dependencies.yaml                                | 2 +-
 3 files changed, 3 insertions(+), 1 deletion(-)

diff --git a/conda/environments/all_cuda-118_arch-x86_64.yaml b/conda/environments/all_cuda-118_arch-x86_64.yaml
index 5c5ecb453..71609ca35 100644
--- a/conda/environments/all_cuda-118_arch-x86_64.yaml
+++ b/conda/environments/all_cuda-118_arch-x86_64.yaml
@@ -4,6 +4,7 @@ channels:
 - rapidsai
 - conda-forge
 dependencies:
+- breathe
 - cmake>=3.26.4
 - cuda-python>=11.7.1,<12.0a0
 - cuda-version=11.8
diff --git a/conda/environments/all_cuda-120_arch-x86_64.yaml b/conda/environments/all_cuda-120_arch-x86_64.yaml
index e01de1ca2..ce89249a9 100644
--- a/conda/environments/all_cuda-120_arch-x86_64.yaml
+++ b/conda/environments/all_cuda-120_arch-x86_64.yaml
@@ -4,6 +4,7 @@ channels:
 - rapidsai
 - conda-forge
 dependencies:
+- breathe
 - cmake>=3.26.4
 - cuda-nvcc
 - cuda-python>=12.0,<13.0a0
diff --git a/dependencies.yaml b/dependencies.yaml
index 68020fd0a..1a1be7e4f 100644
--- a/dependencies.yaml
+++ b/dependencies.yaml
@@ -12,7 +12,6 @@ files:
       - develop
       - docs
       - run
-      - docs
       - test_python
   test_python:
     output: none
@@ -158,6 +157,7 @@ dependencies:
       - output_types: [conda]
         packages:
           - *doxygen
+          - breathe
           - graphviz
           - ipython
           - make

From d58bef25fee6041387b319d85b9f625317042af0 Mon Sep 17 00:00:00 2001
From: Vyas Ramasubramani <vyasr@nvidia.com>
Date: Tue, 8 Aug 2023 17:36:48 -0700
Subject: [PATCH 03/17] Add C++ docs

---
 doxygen/Doxyfile           |  2 +-
 python/docs/conf.py        | 12 ++++++++++++
 python/docs/index.rst      |  1 +
 python/docs/librmm_api.rst |  4 ++++
 4 files changed, 18 insertions(+), 1 deletion(-)
 create mode 100644 python/docs/librmm_api.rst

diff --git a/doxygen/Doxyfile b/doxygen/Doxyfile
index 2ace4abba..28a789a9f 100644
--- a/doxygen/Doxyfile
+++ b/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
diff --git a/python/docs/conf.py b/python/docs/conf.py
index d8c7460dc..e224f59dd 100644
--- a/python/docs/conf.py
+++ b/python/docs/conf.py
@@ -46,8 +46,12 @@
     "IPython.sphinxext.ipython_directive",
     "nbsphinx",
     "recommonmark",
+    "breathe",
 ]
 
+# Breathe Configuration
+breathe_projects = {"librmm": "../../doxygen/xml"}
+breathe_default_project = "librmm"
 
 copybutton_prompt_text = ">>> "
 
@@ -197,9 +201,17 @@
 ]
 
 
+def on_missing_reference(app, env, node, contnode):
+    return contnode
+    if node["refdomain"] == "cpp":
+        return contnode
+    return None
+
+
 def setup(app):
     app.add_js_file("copybutton_pydocs.js")
     app.add_css_file("https://docs.rapids.ai/assets/css/custom.css")
     app.add_js_file(
         "https://docs.rapids.ai/assets/js/custom.js", loading_method="defer"
     )
+    app.connect("missing-reference", on_missing_reference)
diff --git a/python/docs/index.rst b/python/docs/index.rst
index 8dffdeb02..4a675d297 100644
--- a/python/docs/index.rst
+++ b/python/docs/index.rst
@@ -12,6 +12,7 @@ Welcome to rmm's documentation!
 
    basics.md
    api.rst
+   librmm_api.rst
 
 
 Indices and tables
diff --git a/python/docs/librmm_api.rst b/python/docs/librmm_api.rst
new file mode 100644
index 000000000..087eda247
--- /dev/null
+++ b/python/docs/librmm_api.rst
@@ -0,0 +1,4 @@
+API Reference
+==============
+
+.. doxygennamespace:: rmm

From 2a479d5dcc3470768105c03507748d306716f70d Mon Sep 17 00:00:00 2001
From: Vyas Ramasubramani <vyasr@nvidia.com>
Date: Wed, 9 Aug 2023 08:13:31 -0700
Subject: [PATCH 04/17] Fix doxygen code blocks so that they also display
 nicely in Sphinx

---
 include/rmm/device_uvector.hpp                              | 2 +-
 include/rmm/mr/device/device_memory_resource.hpp            | 2 +-
 include/rmm/mr/device/failure_callback_resource_adaptor.hpp | 2 +-
 include/rmm/mr/device/per_device_resource.hpp               | 2 +-
 include/rmm/mr/device/polymorphic_allocator.hpp             | 2 +-
 5 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/include/rmm/device_uvector.hpp b/include/rmm/device_uvector.hpp
index 47ed1adff..74103e9d4 100644
--- a/include/rmm/device_uvector.hpp
+++ b/include/rmm/device_uvector.hpp
@@ -39,7 +39,7 @@ namespace rmm {
  * `thrust::uninitialized_fill`.
  *
  * Example:
- * @code{c++}
+ * @code{.cpp}
  * rmm::mr::device_memory_resource * mr = new my_custom_resource();
  * rmm::cuda_stream_view s{};
  *
diff --git a/include/rmm/mr/device/device_memory_resource.hpp b/include/rmm/mr/device/device_memory_resource.hpp
index 5f511e393..5149411d5 100644
--- a/include/rmm/mr/device/device_memory_resource.hpp
+++ b/include/rmm/mr/device/device_memory_resource.hpp
@@ -70,7 +70,7 @@ namespace rmm::mr {
  * pool_memory_resource objects for each device and sets them as the per-device resource for that
  * device.
  *
- * @code{c++}
+ * @code{.cpp}
  * std::vector<unique_ptr<pool_memory_resource>> per_device_pools;
  * for(int i = 0; i < N; ++i) {
  *   cudaSetDevice(i);
diff --git a/include/rmm/mr/device/failure_callback_resource_adaptor.hpp b/include/rmm/mr/device/failure_callback_resource_adaptor.hpp
index 48f0513d8..a6778d1b0 100644
--- a/include/rmm/mr/device/failure_callback_resource_adaptor.hpp
+++ b/include/rmm/mr/device/failure_callback_resource_adaptor.hpp
@@ -55,7 +55,7 @@ using failure_callback_t = std::function<bool(std::size_t, void*)>;
  * When implementing a callback function for allocation retry, care must be taken to avoid an
  * infinite loop. The following example makes sure to only retry the allocation once:
  *
- * @code{c++}
+ * @code{.cpp}
  * using failure_callback_adaptor =
  *   rmm::mr::failure_callback_resource_adaptor<rmm::mr::device_memory_resource>;
  *
diff --git a/include/rmm/mr/device/per_device_resource.hpp b/include/rmm/mr/device/per_device_resource.hpp
index aa7217758..594b48749 100644
--- a/include/rmm/mr/device/per_device_resource.hpp
+++ b/include/rmm/mr/device/per_device_resource.hpp
@@ -61,7 +61,7 @@
  * pool_memory_resource objects for each device and sets them as the per-device resource for that
  * device.
  *
- * @code{c++}
+ * @code{.cpp}
  * std::vector<unique_ptr<pool_memory_resource>> per_device_pools;
  * for(int i = 0; i < N; ++i) {
  *   cudaSetDevice(i);
diff --git a/include/rmm/mr/device/polymorphic_allocator.hpp b/include/rmm/mr/device/polymorphic_allocator.hpp
index a52ec14d1..c3b8ac6ea 100644
--- a/include/rmm/mr/device/polymorphic_allocator.hpp
+++ b/include/rmm/mr/device/polymorphic_allocator.hpp
@@ -133,7 +133,7 @@ bool operator!=(polymorphic_allocator<T> const& lhs, polymorphic_allocator<U> co
  *`deallocate` functions.
  *
  * Example:
- *\code{c++}
+ *\code{.cpp}
  * my_stream_ordered_allocator<int> a{...};
  * cuda_stream_view s = // create stream;
  *

From 565d6317ea158dd0ffe74c97d8d4abb73383ff53 Mon Sep 17 00:00:00 2001
From: Vyas Ramasubramani <vyasr@nvidia.com>
Date: Wed, 9 Aug 2023 17:53:34 -0700
Subject: [PATCH 05/17] Temporarily allow building with warnings so that CI can
 complete and generate previews

---
 python/docs/Makefile | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/python/docs/Makefile b/python/docs/Makefile
index 72d1dcd85..68e58b474 100644
--- a/python/docs/Makefile
+++ b/python/docs/Makefile
@@ -3,7 +3,7 @@
 
 # You can set these variables from the command line, and also
 # from the environment for the first two.
-SPHINXOPTS    = -n -v -W --keep-going
+SPHINXOPTS    = -n -v
 SPHINXBUILD   ?= sphinx-build
 SOURCEDIR     = .
 BUILDDIR      = _build

From 0bb577029c13fef6d51181a93b0eab0b2b892a02 Mon Sep 17 00:00:00 2001
From: Vyas Ramasubramani <vyasr@nvidia.com>
Date: Mon, 11 Sep 2023 13:42:00 -0700
Subject: [PATCH 06/17] Also ignore xml

---
 .gitignore | 1 +
 1 file changed, 1 insertion(+)

diff --git a/.gitignore b/.gitignore
index ad6c8ebf7..8d5823ab2 100644
--- a/.gitignore
+++ b/.gitignore
@@ -91,6 +91,7 @@ thirdparty/googletest/
 
 ## Doxygen
 doxygen/html
+doxygen/xml
 
 #Java
 target

From ddbdc56b5edc8d68b227accba99195d1a456c8f6 Mon Sep 17 00:00:00 2001
From: Vyas Ramasubramani <vyasr@nvidia.com>
Date: Wed, 20 Sep 2023 15:40:33 -0700
Subject: [PATCH 07/17] Fix doxygen issues revealed by building XML

---
 include/rmm/cuda_stream_view.hpp              |  2 +-
 .../mr/device/aligned_resource_adaptor.hpp    |  2 +-
 .../rmm/mr/device/arena_memory_resource.hpp   |  3 +-
 .../rmm/mr/device/binning_memory_resource.hpp |  5 ++--
 .../mr/device/callback_memory_resource.hpp    | 23 ++++++++++++++
 .../mr/device/cuda_async_memory_resource.hpp  | 24 +++++++++------
 .../cuda_async_view_memory_resource.hpp       | 22 +++++++++-----
 .../rmm/mr/device/cuda_memory_resource.hpp    | 27 +++++++++--------
 .../rmm/mr/device/device_memory_resource.hpp  |  2 +-
 .../mr/device/limiting_resource_adaptor.hpp   |  5 ++--
 .../mr/device/logging_resource_adaptor.hpp    |  2 ++
 .../rmm/mr/device/managed_memory_resource.hpp | 30 +++++++++++--------
 include/rmm/thrust_rmm_allocator.h            |  4 +--
 13 files changed, 99 insertions(+), 52 deletions(-)

diff --git a/include/rmm/cuda_stream_view.hpp b/include/rmm/cuda_stream_view.hpp
index 150fe3957..b6e3cc022 100644
--- a/include/rmm/cuda_stream_view.hpp
+++ b/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)
diff --git a/include/rmm/mr/device/aligned_resource_adaptor.hpp b/include/rmm/mr/device/aligned_resource_adaptor.hpp
index 0b933fb56..f79f0c15a 100644
--- a/include/rmm/mr/device/aligned_resource_adaptor.hpp
+++ b/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
    */
diff --git a/include/rmm/mr/device/arena_memory_resource.hpp b/include/rmm/mr/device/arena_memory_resource.hpp
index e16fbaf7c..882cbf694 100644
--- a/include/rmm/mr/device/arena_memory_resource.hpp
+++ b/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);
   }
diff --git a/include/rmm/mr/device/binning_memory_resource.hpp b/include/rmm/mr/device/binning_memory_resource.hpp
index 705df63c6..54cccca28 100644
--- a/include/rmm/mr/device/binning_memory_resource.hpp
+++ b/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);
   }
diff --git a/include/rmm/mr/device/callback_memory_resource.hpp b/include/rmm/mr/device/callback_memory_resource.hpp
index 2d9695be2..8a42df6e2 100644
--- a/include/rmm/mr/device/callback_memory_resource.hpp
+++ b/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, 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, 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_);
diff --git a/include/rmm/mr/device/cuda_async_memory_resource.hpp b/include/rmm/mr/device/cuda_async_memory_resource.hpp
index 329d8f29a..cd75f5add 100644
--- a/include/rmm/mr/device/cuda_async_memory_resource.hpp
+++ b/include/rmm/mr/device/cuda_async_memory_resource.hpp
@@ -173,13 +173,15 @@ 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
+   * If supported, this operation may optionally be executed on a stream.
+   * Otherwise, the stream is ignored and the null stream is used.
    *
-   * @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 +199,21 @@ class cuda_async_memory_resource final : public device_memory_resource {
   /**
    * @brief Deallocate memory pointed to by \p p.
    *
-   * @throws Nothing.
+   * 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
    */
-  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
   }
diff --git a/include/rmm/mr/device/cuda_async_view_memory_resource.hpp b/include/rmm/mr/device/cuda_async_view_memory_resource.hpp
index c685cd75f..190897b82 100644
--- a/include/rmm/mr/device/cuda_async_view_memory_resource.hpp
+++ b/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
@@ -135,11 +134,17 @@ class cuda_async_view_memory_resource final : public device_memory_resource {
   /**
    * @brief Deallocate memory pointed to by \p p.
    *
-   * @throws Nothing.
+   * 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
    */
-  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) {
@@ -147,6 +152,7 @@ class cuda_async_view_memory_resource final : public device_memory_resource {
     }
 #else
     (void)ptr;
+    (void)bytes;
     (void)stream;
 #endif
   }
diff --git a/include/rmm/mr/device/cuda_memory_resource.hpp b/include/rmm/mr/device/cuda_memory_resource.hpp
index b0bf9ae09..6bd15bd27 100644
--- a/include/rmm/mr/device/cuda_memory_resource.hpp
+++ b/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 +75,17 @@ class cuda_memory_resource final : public device_memory_resource {
   /**
    * @brief Deallocate memory pointed to by \p p.
    *
-   * @note Stream argument is ignored.
-   *
-   * @throws Nothing.
+   * 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
    */
-  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));
   }
diff --git a/include/rmm/mr/device/device_memory_resource.hpp b/include/rmm/mr/device/device_memory_resource.hpp
index 5149411d5..247774145 100644
--- a/include/rmm/mr/device/device_memory_resource.hpp
+++ b/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
diff --git a/include/rmm/mr/device/limiting_resource_adaptor.hpp b/include/rmm/mr/device/limiting_resource_adaptor.hpp
index 895c404b0..ebc67fed1 100644
--- a/include/rmm/mr/device/limiting_resource_adaptor.hpp
+++ b/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,
diff --git a/include/rmm/mr/device/logging_resource_adaptor.hpp b/include/rmm/mr/device/logging_resource_adaptor.hpp
index 0ff9e950b..263387b44 100644
--- a/include/rmm/mr/device/logging_resource_adaptor.hpp
+++ b/include/rmm/mr/device/logging_resource_adaptor.hpp
@@ -329,6 +329,7 @@ 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
  */
 template <typename Upstream>
 logging_resource_adaptor<Upstream> make_logging_adaptor(
@@ -346,6 +347,7 @@ 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
  */
 template <typename Upstream>
 logging_resource_adaptor<Upstream> make_logging_adaptor(Upstream* upstream,
diff --git a/include/rmm/mr/device/managed_memory_resource.hpp b/include/rmm/mr/device/managed_memory_resource.hpp
index 4a0f7701a..d53da4e61 100644
--- a/include/rmm/mr/device/managed_memory_resource.hpp
+++ b/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 +77,19 @@ 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.
+   * If supported, 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, 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 +119,8 @@ 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{};
diff --git a/include/rmm/thrust_rmm_allocator.h b/include/rmm/thrust_rmm_allocator.h
index 3bbcab0a5..0b78d5497 100644
--- a/include/rmm/thrust_rmm_allocator.h
+++ b/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(

From 5c7f8c7cb4cf6abe9679872918482168a7027228 Mon Sep 17 00:00:00 2001
From: Vyas Ramasubramani <vyasr@nvidia.com>
Date: Wed, 20 Sep 2023 15:46:48 -0700
Subject: [PATCH 08/17] Fix some typos in the docs

---
 include/rmm/mr/device/callback_memory_resource.hpp        | 8 ++++----
 include/rmm/mr/device/cuda_async_memory_resource.hpp      | 6 ------
 include/rmm/mr/device/cuda_async_view_memory_resource.hpp | 3 ---
 include/rmm/mr/device/cuda_memory_resource.hpp            | 5 ++---
 include/rmm/mr/device/managed_memory_resource.hpp         | 5 ++---
 5 files changed, 8 insertions(+), 19 deletions(-)

diff --git a/include/rmm/mr/device/callback_memory_resource.hpp b/include/rmm/mr/device/callback_memory_resource.hpp
index 8a42df6e2..289ceb9d6 100644
--- a/include/rmm/mr/device/callback_memory_resource.hpp
+++ b/include/rmm/mr/device/callback_memory_resource.hpp
@@ -105,8 +105,8 @@ class callback_memory_resource final : public device_memory_resource {
    *
    * The returned pointer will have at minimum 256 byte alignment.
    *
-   * If supported, this operation may optionally be executed on a stream.
-   * Otherwise, the stream is ignored and the null stream is used.
+   * 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
@@ -120,8 +120,8 @@ class callback_memory_resource final : public device_memory_resource {
   /**
    * @brief Deallocate memory pointed to by \p p.
    *
-   * If supported, this operation may optionally be executed on a stream.
-   * Otherwise, the stream is ignored and the null stream is used.
+   * 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
diff --git a/include/rmm/mr/device/cuda_async_memory_resource.hpp b/include/rmm/mr/device/cuda_async_memory_resource.hpp
index cd75f5add..9bbd9cc09 100644
--- a/include/rmm/mr/device/cuda_async_memory_resource.hpp
+++ b/include/rmm/mr/device/cuda_async_memory_resource.hpp
@@ -177,9 +177,6 @@ class cuda_async_memory_resource final : public device_memory_resource {
    *
    * The returned pointer will have at minimum 256 byte alignment.
    *
-   * If supported, 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
@@ -199,9 +196,6 @@ class cuda_async_memory_resource final : public device_memory_resource {
   /**
    * @brief Deallocate memory pointed to by \p p.
    *
-   * If supported, 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`.
diff --git a/include/rmm/mr/device/cuda_async_view_memory_resource.hpp b/include/rmm/mr/device/cuda_async_view_memory_resource.hpp
index 190897b82..569f65df7 100644
--- a/include/rmm/mr/device/cuda_async_view_memory_resource.hpp
+++ b/include/rmm/mr/device/cuda_async_view_memory_resource.hpp
@@ -134,9 +134,6 @@ class cuda_async_view_memory_resource final : public device_memory_resource {
   /**
    * @brief Deallocate memory pointed to by \p p.
    *
-   * If supported, 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`.
diff --git a/include/rmm/mr/device/cuda_memory_resource.hpp b/include/rmm/mr/device/cuda_memory_resource.hpp
index 6bd15bd27..5164bc5c5 100644
--- a/include/rmm/mr/device/cuda_memory_resource.hpp
+++ b/include/rmm/mr/device/cuda_memory_resource.hpp
@@ -75,13 +75,12 @@ class cuda_memory_resource final : public device_memory_resource {
   /**
    * @brief Deallocate memory pointed to by \p p.
    *
-   * If supported, this operation may optionally be executed on a stream.
-   * Otherwise, the stream is ignored and the null stream is used.
+   * The stream argument 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 Stream on which to perform deallocation
+   * @param stream This argument is ignored.
    */
   void do_deallocate(void* ptr,
                      [[maybe_unused]] std::size_t bytes,
diff --git a/include/rmm/mr/device/managed_memory_resource.hpp b/include/rmm/mr/device/managed_memory_resource.hpp
index d53da4e61..e987de15d 100644
--- a/include/rmm/mr/device/managed_memory_resource.hpp
+++ b/include/rmm/mr/device/managed_memory_resource.hpp
@@ -79,13 +79,12 @@ class managed_memory_resource final : public device_memory_resource {
   /**
    * @brief Deallocate memory pointed to by \p p.
    *
-   * If supported, this operation may optionally be executed on a stream.
-   * Otherwise, the stream is ignored and the null stream is used.
+   * 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 Stream on which to perform deallocation
+   * @param stream This argument is ignored
    */
   void do_deallocate(void* ptr,
                      [[maybe_unused]] std::size_t bytes,

From d8a16ad6e8a6598eb709f8938b0cda801ef532c0 Mon Sep 17 00:00:00 2001
From: Vyas Ramasubramani <vyasr@nvidia.com>
Date: Thu, 21 Sep 2023 10:37:08 -0700
Subject: [PATCH 09/17] Alphabetize deps

---
 dependencies.yaml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/dependencies.yaml b/dependencies.yaml
index 1a1be7e4f..dab7daab8 100644
--- a/dependencies.yaml
+++ b/dependencies.yaml
@@ -156,8 +156,8 @@ dependencies:
     common:
       - output_types: [conda]
         packages:
-          - *doxygen
           - breathe
+          - *doxygen
           - graphviz
           - ipython
           - make

From 6d941e30b0029f67369e68e24e9554a670950be3 Mon Sep 17 00:00:00 2001
From: Vyas Ramasubramani <vyasr@nvidia.com>
Date: Thu, 21 Sep 2023 10:41:28 -0700
Subject: [PATCH 10/17] Fix typos

---
 include/rmm/mr/device/logging_resource_adaptor.hpp | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/include/rmm/mr/device/logging_resource_adaptor.hpp b/include/rmm/mr/device/logging_resource_adaptor.hpp
index 263387b44..c59fdf61a 100644
--- a/include/rmm/mr/device/logging_resource_adaptor.hpp
+++ b/include/rmm/mr/device/logging_resource_adaptor.hpp
@@ -330,6 +330,7 @@ class logging_resource_adaptor final : public device_memory_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(
@@ -348,6 +349,7 @@ logging_resource_adaptor<Upstream> make_logging_adaptor(
  * @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,

From 576f87e19c2b9a9bc20b92f443f8b01056fd8181 Mon Sep 17 00:00:00 2001
From: Vyas Ramasubramani <vyasr@nvidia.com>
Date: Thu, 21 Sep 2023 10:41:36 -0700
Subject: [PATCH 11/17] Remove patch for now

---
 python/docs/conf.py | 11 +++++------
 1 file changed, 5 insertions(+), 6 deletions(-)

diff --git a/python/docs/conf.py b/python/docs/conf.py
index e224f59dd..021a077c1 100644
--- a/python/docs/conf.py
+++ b/python/docs/conf.py
@@ -201,11 +201,10 @@
 ]
 
 
-def on_missing_reference(app, env, node, contnode):
-    return contnode
-    if node["refdomain"] == "cpp":
-        return contnode
-    return None
+# def on_missing_reference(app, env, node, contnode):
+#     if node["refdomain"] == "cpp":
+#         return contnode
+#     return None
 
 
 def setup(app):
@@ -214,4 +213,4 @@ def setup(app):
     app.add_js_file(
         "https://docs.rapids.ai/assets/js/custom.js", loading_method="defer"
     )
-    app.connect("missing-reference", on_missing_reference)
+    # app.connect("missing-reference", on_missing_reference)

From bc79db2dc620f1b4756b18ff1488917f2d715075 Mon Sep 17 00:00:00 2001
From: Vyas Ramasubramani <vyasr@nvidia.com>
Date: Thu, 21 Sep 2023 10:54:17 -0700
Subject: [PATCH 12/17] First pass of reorg

---
 python/docs/cpp.rst                         | 8 ++++++++
 python/docs/{librmm_api.rst => cpp_api.rst} | 2 +-
 python/docs/{basics.md => guide.md}         | 2 +-
 python/docs/index.rst                       | 5 ++---
 python/docs/python.rst                      | 9 +++++++++
 python/docs/{api.rst => python_api.rst}     | 0
 6 files changed, 21 insertions(+), 5 deletions(-)
 create mode 100644 python/docs/cpp.rst
 rename python/docs/{librmm_api.rst => cpp_api.rst} (73%)
 rename python/docs/{basics.md => guide.md} (99%)
 create mode 100644 python/docs/python.rst
 rename python/docs/{api.rst => python_api.rst} (100%)

diff --git a/python/docs/cpp.rst b/python/docs/cpp.rst
new file mode 100644
index 000000000..e60f15129
--- /dev/null
+++ b/python/docs/cpp.rst
@@ -0,0 +1,8 @@
+Welcome to the rmm C++ documentation!
+========================================
+
+.. toctree::
+   :maxdepth: 2
+   :caption: Contents:
+
+   cpp_api.rst
diff --git a/python/docs/librmm_api.rst b/python/docs/cpp_api.rst
similarity index 73%
rename from python/docs/librmm_api.rst
rename to python/docs/cpp_api.rst
index 087eda247..ca0afbb0d 100644
--- a/python/docs/librmm_api.rst
+++ b/python/docs/cpp_api.rst
@@ -1,4 +1,4 @@
 API Reference
-==============
+=============
 
 .. doxygennamespace:: rmm
diff --git a/python/docs/basics.md b/python/docs/guide.md
similarity index 99%
rename from python/docs/basics.md
rename to python/docs/guide.md
index 997745f00..c06135ca8 100644
--- a/python/docs/basics.md
+++ b/python/docs/guide.md
@@ -1,4 +1,4 @@
-# RMM - the RAPIDS Memory Manager
+# User Guide
 
 Achieving optimal performance in GPU-centric workflows frequently requires
 customizing how GPU ("device") memory is allocated.
diff --git a/python/docs/index.rst b/python/docs/index.rst
index 4a675d297..524351253 100644
--- a/python/docs/index.rst
+++ b/python/docs/index.rst
@@ -10,9 +10,8 @@ Welcome to rmm's documentation!
    :maxdepth: 2
    :caption: Contents:
 
-   basics.md
-   api.rst
-   librmm_api.rst
+   Python <python.rst>
+   C++ <cpp.rst>
 
 
 Indices and tables
diff --git a/python/docs/python.rst b/python/docs/python.rst
new file mode 100644
index 000000000..bff919627
--- /dev/null
+++ b/python/docs/python.rst
@@ -0,0 +1,9 @@
+Welcome to the rmm Python documentation!
+========================================
+
+.. toctree::
+   :maxdepth: 2
+   :caption: Contents:
+
+   guide.md
+   python_api.rst
diff --git a/python/docs/api.rst b/python/docs/python_api.rst
similarity index 100%
rename from python/docs/api.rst
rename to python/docs/python_api.rst

From 1272bc8c750e681318aa6dd6569cb902ffa2fd58 Mon Sep 17 00:00:00 2001
From: Vyas Ramasubramani <vyasr@nvidia.com>
Date: Mon, 25 Sep 2023 09:47:07 -0700
Subject: [PATCH 13/17] Remove breathe

---
 conda/environments/all_cuda-118_arch-x86_64.yaml | 1 -
 conda/environments/all_cuda-120_arch-x86_64.yaml | 1 -
 dependencies.yaml                                | 1 -
 3 files changed, 3 deletions(-)

diff --git a/conda/environments/all_cuda-118_arch-x86_64.yaml b/conda/environments/all_cuda-118_arch-x86_64.yaml
index 71609ca35..5c5ecb453 100644
--- a/conda/environments/all_cuda-118_arch-x86_64.yaml
+++ b/conda/environments/all_cuda-118_arch-x86_64.yaml
@@ -4,7 +4,6 @@ channels:
 - rapidsai
 - conda-forge
 dependencies:
-- breathe
 - cmake>=3.26.4
 - cuda-python>=11.7.1,<12.0a0
 - cuda-version=11.8
diff --git a/conda/environments/all_cuda-120_arch-x86_64.yaml b/conda/environments/all_cuda-120_arch-x86_64.yaml
index ce89249a9..e01de1ca2 100644
--- a/conda/environments/all_cuda-120_arch-x86_64.yaml
+++ b/conda/environments/all_cuda-120_arch-x86_64.yaml
@@ -4,7 +4,6 @@ channels:
 - rapidsai
 - conda-forge
 dependencies:
-- breathe
 - cmake>=3.26.4
 - cuda-nvcc
 - cuda-python>=12.0,<13.0a0
diff --git a/dependencies.yaml b/dependencies.yaml
index dab7daab8..765a812cb 100644
--- a/dependencies.yaml
+++ b/dependencies.yaml
@@ -156,7 +156,6 @@ dependencies:
     common:
       - output_types: [conda]
         packages:
-          - breathe
           - *doxygen
           - graphviz
           - ipython

From 51bbbd62cdb64d20b5a28aeeb5f75f76c3c23814 Mon Sep 17 00:00:00 2001
From: Vyas Ramasubramani <vyasr@nvidia.com>
Date: Mon, 25 Sep 2023 09:48:11 -0700
Subject: [PATCH 14/17] Revert code block changes

---
 include/rmm/device_uvector.hpp                              | 2 +-
 include/rmm/mr/device/device_memory_resource.hpp            | 2 +-
 include/rmm/mr/device/failure_callback_resource_adaptor.hpp | 2 +-
 include/rmm/mr/device/per_device_resource.hpp               | 2 +-
 4 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/include/rmm/device_uvector.hpp b/include/rmm/device_uvector.hpp
index 74103e9d4..47ed1adff 100644
--- a/include/rmm/device_uvector.hpp
+++ b/include/rmm/device_uvector.hpp
@@ -39,7 +39,7 @@ namespace rmm {
  * `thrust::uninitialized_fill`.
  *
  * Example:
- * @code{.cpp}
+ * @code{c++}
  * rmm::mr::device_memory_resource * mr = new my_custom_resource();
  * rmm::cuda_stream_view s{};
  *
diff --git a/include/rmm/mr/device/device_memory_resource.hpp b/include/rmm/mr/device/device_memory_resource.hpp
index 247774145..1813fb726 100644
--- a/include/rmm/mr/device/device_memory_resource.hpp
+++ b/include/rmm/mr/device/device_memory_resource.hpp
@@ -70,7 +70,7 @@ namespace rmm::mr {
  * pool_memory_resource objects for each device and sets them as the per-device resource for that
  * device.
  *
- * @code{.cpp}
+ * @code{c++}
  * std::vector<unique_ptr<pool_memory_resource>> per_device_pools;
  * for(int i = 0; i < N; ++i) {
  *   cudaSetDevice(i);
diff --git a/include/rmm/mr/device/failure_callback_resource_adaptor.hpp b/include/rmm/mr/device/failure_callback_resource_adaptor.hpp
index a6778d1b0..48f0513d8 100644
--- a/include/rmm/mr/device/failure_callback_resource_adaptor.hpp
+++ b/include/rmm/mr/device/failure_callback_resource_adaptor.hpp
@@ -55,7 +55,7 @@ using failure_callback_t = std::function<bool(std::size_t, void*)>;
  * When implementing a callback function for allocation retry, care must be taken to avoid an
  * infinite loop. The following example makes sure to only retry the allocation once:
  *
- * @code{.cpp}
+ * @code{c++}
  * using failure_callback_adaptor =
  *   rmm::mr::failure_callback_resource_adaptor<rmm::mr::device_memory_resource>;
  *
diff --git a/include/rmm/mr/device/per_device_resource.hpp b/include/rmm/mr/device/per_device_resource.hpp
index 594b48749..aa7217758 100644
--- a/include/rmm/mr/device/per_device_resource.hpp
+++ b/include/rmm/mr/device/per_device_resource.hpp
@@ -61,7 +61,7 @@
  * pool_memory_resource objects for each device and sets them as the per-device resource for that
  * device.
  *
- * @code{.cpp}
+ * @code{c++}
  * std::vector<unique_ptr<pool_memory_resource>> per_device_pools;
  * for(int i = 0; i < N; ++i) {
  *   cudaSetDevice(i);

From 657b09847697aa33b25d78c3e33c6bd2731d3eaf Mon Sep 17 00:00:00 2001
From: Vyas Ramasubramani <vyasr@nvidia.com>
Date: Mon, 25 Sep 2023 09:49:17 -0700
Subject: [PATCH 15/17] Revert Sphinx configuration changes

---
 python/docs/Makefile |  2 +-
 python/docs/conf.py  | 11 -----------
 2 files changed, 1 insertion(+), 12 deletions(-)

diff --git a/python/docs/Makefile b/python/docs/Makefile
index 68e58b474..72d1dcd85 100644
--- a/python/docs/Makefile
+++ b/python/docs/Makefile
@@ -3,7 +3,7 @@
 
 # You can set these variables from the command line, and also
 # from the environment for the first two.
-SPHINXOPTS    = -n -v
+SPHINXOPTS    = -n -v -W --keep-going
 SPHINXBUILD   ?= sphinx-build
 SOURCEDIR     = .
 BUILDDIR      = _build
diff --git a/python/docs/conf.py b/python/docs/conf.py
index 021a077c1..d8c7460dc 100644
--- a/python/docs/conf.py
+++ b/python/docs/conf.py
@@ -46,12 +46,8 @@
     "IPython.sphinxext.ipython_directive",
     "nbsphinx",
     "recommonmark",
-    "breathe",
 ]
 
-# Breathe Configuration
-breathe_projects = {"librmm": "../../doxygen/xml"}
-breathe_default_project = "librmm"
 
 copybutton_prompt_text = ">>> "
 
@@ -201,16 +197,9 @@
 ]
 
 
-# def on_missing_reference(app, env, node, contnode):
-#     if node["refdomain"] == "cpp":
-#         return contnode
-#     return None
-
-
 def setup(app):
     app.add_js_file("copybutton_pydocs.js")
     app.add_css_file("https://docs.rapids.ai/assets/css/custom.css")
     app.add_js_file(
         "https://docs.rapids.ai/assets/js/custom.js", loading_method="defer"
     )
-    # app.connect("missing-reference", on_missing_reference)

From 665c7eb4aa35cc18656029ad31f1b9c25a678bfe Mon Sep 17 00:00:00 2001
From: Vyas Ramasubramani <vyasr@nvidia.com>
Date: Mon, 25 Sep 2023 09:51:09 -0700
Subject: [PATCH 16/17] Revert Sphinx content changes

---
 python/docs/{python_api.rst => api.rst} | 0
 python/docs/{guide.md => basics.md}     | 2 +-
 python/docs/cpp.rst                     | 8 --------
 python/docs/cpp_api.rst                 | 4 ----
 python/docs/index.rst                   | 4 ++--
 python/docs/python.rst                  | 9 ---------
 6 files changed, 3 insertions(+), 24 deletions(-)
 rename python/docs/{python_api.rst => api.rst} (100%)
 rename python/docs/{guide.md => basics.md} (99%)
 delete mode 100644 python/docs/cpp.rst
 delete mode 100644 python/docs/cpp_api.rst
 delete mode 100644 python/docs/python.rst

diff --git a/python/docs/python_api.rst b/python/docs/api.rst
similarity index 100%
rename from python/docs/python_api.rst
rename to python/docs/api.rst
diff --git a/python/docs/guide.md b/python/docs/basics.md
similarity index 99%
rename from python/docs/guide.md
rename to python/docs/basics.md
index c06135ca8..997745f00 100644
--- a/python/docs/guide.md
+++ b/python/docs/basics.md
@@ -1,4 +1,4 @@
-# User Guide
+# RMM - the RAPIDS Memory Manager
 
 Achieving optimal performance in GPU-centric workflows frequently requires
 customizing how GPU ("device") memory is allocated.
diff --git a/python/docs/cpp.rst b/python/docs/cpp.rst
deleted file mode 100644
index e60f15129..000000000
--- a/python/docs/cpp.rst
+++ /dev/null
@@ -1,8 +0,0 @@
-Welcome to the rmm C++ documentation!
-========================================
-
-.. toctree::
-   :maxdepth: 2
-   :caption: Contents:
-
-   cpp_api.rst
diff --git a/python/docs/cpp_api.rst b/python/docs/cpp_api.rst
deleted file mode 100644
index ca0afbb0d..000000000
--- a/python/docs/cpp_api.rst
+++ /dev/null
@@ -1,4 +0,0 @@
-API Reference
-=============
-
-.. doxygennamespace:: rmm
diff --git a/python/docs/index.rst b/python/docs/index.rst
index 524351253..8dffdeb02 100644
--- a/python/docs/index.rst
+++ b/python/docs/index.rst
@@ -10,8 +10,8 @@ Welcome to rmm's documentation!
    :maxdepth: 2
    :caption: Contents:
 
-   Python <python.rst>
-   C++ <cpp.rst>
+   basics.md
+   api.rst
 
 
 Indices and tables
diff --git a/python/docs/python.rst b/python/docs/python.rst
deleted file mode 100644
index bff919627..000000000
--- a/python/docs/python.rst
+++ /dev/null
@@ -1,9 +0,0 @@
-Welcome to the rmm Python documentation!
-========================================
-
-.. toctree::
-   :maxdepth: 2
-   :caption: Contents:
-
-   guide.md
-   python_api.rst

From 14cb408f50150de952b42ea323645b3deb44fff5 Mon Sep 17 00:00:00 2001
From: Vyas Ramasubramani <vyasr@nvidia.com>
Date: Mon, 25 Sep 2023 09:52:05 -0700
Subject: [PATCH 17/17] One missed reversion

---
 include/rmm/mr/device/polymorphic_allocator.hpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/include/rmm/mr/device/polymorphic_allocator.hpp b/include/rmm/mr/device/polymorphic_allocator.hpp
index c3b8ac6ea..a52ec14d1 100644
--- a/include/rmm/mr/device/polymorphic_allocator.hpp
+++ b/include/rmm/mr/device/polymorphic_allocator.hpp
@@ -133,7 +133,7 @@ bool operator!=(polymorphic_allocator<T> const& lhs, polymorphic_allocator<U> co
  *`deallocate` functions.
  *
  * Example:
- *\code{.cpp}
+ *\code{c++}
  * my_stream_ordered_allocator<int> a{...};
  * cuda_stream_view s = // create stream;
  *