Update _toc.yml.in

.readthedocs.yaml

@@ -6,13 +6,13 @@ version: 2
 sphinx:
    configuration: docs/conf.py
 
-formats: [htmlzip, pdf, epub]
+formats: [htmlzip, pdf]
 
 python:
    install:
    - requirements: docs/sphinx/requirements.txt
 
 build:
-   os: ubuntu-22.04
+   os: ubuntu-20.04
    tools:
       python: "3.8"

docs/apireference.rst → docs/reference/apireference.rst

@@ -1,3 +1,4 @@
+.. _apireference:
 
 API Reference
 =============

docs/reference/index.rst

@@ -0,0 +1,9 @@
+Indices and tables
+========
+.. toctree::
+   :maxdepth: 2
+   :caption: Contents:
+
+    :ref:`what-is-MIOpen`
+    :ref:`install`
+    :ref:`apireference`

docs/sphinx/_toc.yml.in

@@ -1,24 +1,57 @@
+# Anywhere {branch} is used, the branch name will be substituted.
+# These comments will also be removed.
 defaults:
   numbered: False
-  maxdepth: 6
+  maxdepth: 7
 root: index
-subtrees:
-- entries:
-  - file: releasenotes
-  - file: citation
-  - file: install
-  - file: embed
-  - file: driver
-  - file: DebugAndLogging
-  - file: cache
-  - file: perfdatabase
-  - file: finddb
-  - file: find_and_immediate
-  - file: Getting_Started_FusionAPI
-  - file: MI200AlternateImplementation
-  - file: MIOpen_Porting_Guide
-  - file: apireference
-- caption: About
-  entries:
-    - file: license
+- caption: What is MIOpen?
+  subtrees:
+    - entries:
+      - file: tutorials/what-is-MIOpen.rst
+        title: What is MIOpen?
+      - file: tutorials/find_and_immediate.rst
+        title: Find and Immediate
+      - file: tutorials/finddb.rst
+        title: Find database
+      - file: tutorials/cache.rst
+        title: Cache
+      - file: tutorials/perfdatabase.rst
+        title: Performance database
+      - file: tutorials/Getting_Started_FusionAPI.rst
+        title: Getting started with Fusion API
+      - file: tutorials/DebugAndLogging.rst
+        title: Debugging and logging
+      - file: tutorials/MI200AlternateImplementation.rst
+        title: Alternate implementation for MI200
+      - file: tutorials/MIOpen_Porting_Guide.rst
+        title: MIOpen Porting Guide
+
+   - file: tutorials/index.rst
+- caption: Quick-start
+  subtrees:
+  - entries:
+    - file: tutorials/quick-start/install.rst
+      title: MIOpen installation
+    - file: tutorials/quick-start/embed.rst
+      title: Embedding MIOpen
+    - file: tutorials/quick-start/driver.rst
+      title: About MIOpen driver 
+
+  - file: reference/index.rst
+- caption: API reference
+  subtrees:
+    - entries
+      - file: reference/apireference.rst
+        title: API library
+
+  - file: tutorials/index.rst
+- caption: Tutorials
+  subtrees:
+    - entries:
+      - file: tutorials/install.rst
+        title: MIOpen installation
+      - file: tutorials/embed.rst
+        title: Embedding MIOpen
+      - file: tutorials/driver.rst
+        title: Installing and building MIOpen
 

docs/MI200AlternateImplementation.md → docs/tutorials/MI200alternateimplementation.md

@@ -10,4 +10,4 @@ An alternate implementation for the half precision data-type is available in MIO
 
 * The default MIOpen behaviors described above may be overridden using the `miopenSetConvolutionAttribute` API call and passing the convolution descriptor for the appropriate convolution operation and the `MIOPEN_CONVOLUTION_ATTRIB_FP16_ALT_IMPL` convolution attribute with a non-zero value to engage the alternate implementation. 
 
-* The behavior might also be overridden using the `MIOPEN_DEBUG_CONVOLUTION_ATTRIB_FP16_ALT_IMPL` environment variable. The above variable when set to a value of `1` engages the alternate implementation while a value of `0` disables it. Keep in mind the environment variable impacts the convolution operation in all directions. 
+* The behavior might also be overridden using the `MIOPEN_DEBUG_CONVOLUTION_ATTRIB_FP16_ALT_IMPL` environment variable. The above variable when set to a value of `1` engages the alternate implementation while a value of `0` disables it. Keep in mind the environment variable impacts the convolution operation in all directions. 

docs/MIOpen_Porting_Guide.md → docs/tutorials/MIOpenportingguide.md

@@ -1,4 +1,3 @@
-
 # MIOpen Porting Guide
 
 

docs/tutorials/index.rst

@@ -0,0 +1 @@
+

docs/embed.md → docs/tutorials/quick-start/embed.md

@@ -1,4 +1,3 @@
-
 Building MIOpen for Embedded Systems
 ====================================
 

docs/tutorials/quick-start/install.rst

@@ -0,0 +1,62 @@
+.. _install:
+
+Building MIOpen from source
+~~~~~~~~~~
+
+Configuring with cmake
+----------
+
+First create a build directory:
+
+.. code-block:: bash
+
+    mkdir build; cd build;
+
+
+Next configure cmake. The preferred backend for MIOpen can be set using the `-DMIOPEN_BACKEND` cmake variable.
+
+For the HIP backend (ROCm 3.5 and later), run
+----------
+
+Set the C++ compiler to `clang++`.
+
+.. code-block:: bash
+
+    export CXX=<location-of-clang++-compiler>
+    cmake -DMIOPEN_BACKEND=HIP -DCMAKE_PREFIX_PATH="<hip-installed-path>;<rocm-installed-path>;<miopen-dependency-path>" ..
+
+An example cmake step can be:
+
+.. code-block:: bash
+
+    export CXX=/opt/rocm/llvm/bin/clang++ && \
+    cmake -DMIOPEN_BACKEND=HIP -DCMAKE_PREFIX_PATH="/opt/rocm/;/opt/rocm/hip;/root/MIOpen/install_dir" ..
+
+Note: When specifying the path for the `CMAKE_PREFIX_PATH` variable, **do not** use the `~` shorthand for the user home directory.
+
+
+Setting Up Locations
+----------
+
+By default the install location is set to '/opt/rocm', this can be set by using `CMAKE_INSTALL_PREFIX`:
+
+.. code-block:: bash
+
+    cmake -DMIOPEN_BACKEND=OpenCL -DCMAKE_INSTALL_PREFIX=<miopen-installed-path> ..
+
+Building MIOpen using docker
+~~~~~~~~~~
+
+The easiest way is to use docker. You can build the top-level docker file:
+
+.. code-block:: bash
+
+    docker build -t miopen-image .
+
+Then to enter the development environment use `docker run`, for example:
+
+.. code-block:: bash
+
+    docker run -it -v $HOME:/data --privileged --rm --device=/dev/kfd --device /dev/dri:/dev/dri:rw  --volume /dev/dri:/dev/dri:rw -v /var/lib/docker/:/var/lib/docker --group-add video --cap-add=SYS_PTRACE --security-opt seccomp=unconfined miopen-image
+
+Prebuilt docker images can be found on [ROCm's public docker hub here](https://hub.docker.com/r/rocm/miopen/tags).

docs/tutorials/what-is-MIOpen.rst

@@ -0,0 +1,13 @@
+.. meta::
+   :description: MIOpen is AMD’s deep learning primitives library, which provides highly optimized, and hand-tuned implementations of
+   different operators such as convolution, batch normalization, pooling, softmax, activation and layers for Recurrent Neural
+      Networks (RNNs), used in both training and inference [9].
+   :keywords: MIOpen, ROCm, library, API
+
+.. _what-is-MIOpen:
+
+*********************
+What is MIOpen?
+*********************
+
+MIOpen is AMD’s deep learning primitives library, which provides highly optimized and hand-tuned implementations of different operators such as convolution, batch normalization, pooling, softmax, activation and layers for Recurrent Neural Networks (RNNs), used in both training and inference. Moreover, MIOpen is fully open source, including all its GPU kernels, complementing AMD’s open source ROCm stack. MIOpen is the first to extend the open source advantage into GPU vendor libraries thereby, continuing to embark on the same ethos as the deep learning community.