Docs: Update fusion API doc and add link to samples (#3870)

* Update fusion API doc and add link to samples

* Add new rows to table

* Update graphs

Author: amd-jnovotny

docs/data/how-to/bn_activ_fused.png

@@ -269,22 +269,110 @@ need to worry about additional cleanup.
 Supported fusions
 =================================================
 
-The following tables outline the supported fusions for ``FP32`` and ``FP16``, including any applicable
+The following tables outline the supported fusions for ``FP32``, ``FP16``, and ``BFP16``, including any applicable
 constraints.
 
 .. note::
 
-  Fusion Plans with grouped convolutions are not supported.
+   Fusion Plans with grouped convolutions are supported in the inference direction for
+   convolution, bias, and activation.
+
+The following abbreviations apply to the combination column in the following tables:
+
+*  **C**: Convolution
+*  **B**: Bias
+*  **N**: Batch Normalization
+*  **A**: Activation
+ 
+For example, CBA refers to convolution plus bias plus activation.
+
+Convolution-based FP32 fusion for inference
+-------------------------------------------
+
+The following table applies to single-precision floating point.
+
+.. csv-table::
+   :header: "Combination","Conv algo","Stride","Filter dims","N mode","Activations","Other constraints"
+   :widths: 15, 15, 15, 20, 12, 20, 20
+
+   "CBNA","Direct","1 and 2","3x3, 5x5, 7x7, 9x9, 11x11","All","All","stride and padding must be either 1 or 2"
+   "CBA","Direct","--","1x1","--","All","stride and padding not supported"
+   "CBA","Winograd","1","1x1, 2x2","N/A","Relu, Leaky Relu","c >= 18"
+   "CBA","Winograd","1","3x3","--","Relu, Leaky Relu","c >= 18 and c is even"
+   "CBA","Winograd","1","4x4, 5x5, 6x6","--","Relu, Leaky Relu","4 x c >= 18"
+   "CBA","Winograd","1","7x7, 8x8, 9x9","--","Relu, Leaky Relu","12 x c >= 18"
+   "CBA","Winograd","1","10x10, 11x11, 12x12","--","Relu, Leaky Relu","16 x c >= 18"
+   "CBA","Winograd","1","larger filter sizes","--","Relu, Leaky Relu","none"
+   "CBA","Winograd","2","1x1","--","Relu, Leaky Relu","2 x c >= 18"
+   "CBA","Winograd","2","2x2, 3x3, 4x4, 5x5, 6x6","--","Relu, Leaky Relu","4 x c >= 18"
+   "CBA","Winograd","2","7x7","--","Relu, Leaky Relu","12 x c >= 18"
+   "CBA","Winograd","2","8x8, 9x9, 10x10, 11x11, 12x12","--","Relu, Leaky Relu","16 x c >= 18"
+   "CBA","Winograd","2","larger filter sizes","--","Relu, Leaky Relu","none"
+   "CBA","CK","--","--","--","Relu, Clipped Relu, CLAMP","none"
+   "NA","--","--","--","All","All","padding not supported"
+   "CA","Direct","--","1x1","--","All","stride and padding not supported"
+   "CA","CK","--","--","--","Relu, Clipped Relu, CLAMP","none"
 
-**C = convolution, B = bias, N = batch normalization, A = activation**
+.. note::
 
-.. image:: ../data/how-to/fp32fusions.png
-  :width: 800
-  :alt: Convolution based fp32 fusion
+   N mode is either spatial or per activation. For CBA, other asymmetric kernels are supported but for brevity are not enumerated here.
 
-.. image:: ../data/how-to/fp16fusions.png
-  :width: 800
-  :alt: Convolution based fp16 fusion
+
+Convolution-based FP16 fusion for inference
+-------------------------------------------
+
+The following table applies to half-precision floating point.
+
+.. csv-table::
+   :header: "Combination","Conv algo","Stride","Filter dims","N mode","Activations","Other constraints"
+   :widths: 15, 15, 15, 20, 12, 20, 20
+
+   "CBNA","Direct","1 and 2","3x3, 5x5, 7x7, 9x9, 11x11","All","All","stride and padding must be either 1 or 2"
+   "CBA","Direct","--","1x1","--","All","stride and padding not supported"
+   "CBA","CK","--","--","--","Relu, Clipped Relu, CLAMP","none"
+   "CA","Direct","--","1x1","--","All","stride and padding not supported"
+   "CA","CK","--","--","--","Relu, Clipped Relu, CLAMP","none"
+
+.. note::
+
+   N mode is either spatial or per activation.
+
+
+Convolution-based BFP16 fusion for inference
+--------------------------------------------
+
+The following table applies to half-precision block floating point.
+
+.. csv-table::
+   :header: "Combination","Conv algo","Stride","Filter dims","N mode","Activations","Other constraints"
+   :widths: 15, 15, 15, 20, 12, 20, 20
+
+   "CBNA","Direct","1 and 2","3x3, 5x5, 7x7, 9x9, 11x11","All","All","stride and padding must be either 1 or 2"
+   "CBA","Direct","--","1x1","--","All","stride and padding not supported"
+   "CBA","CK","--","--","--","Relu, Clipped Relu, CLAMP","none"
+   "CA","Direct","--","1x1","--","All","stride and padding not supported"
+   "CA","CK","--","--","--","Relu, Clipped Relu, CLAMP","none"
+
+.. note::
+
+   N mode is either spatial or per activation.
+
+Batch Normalization-based fusion for FP32, BFP16, and FP16 for inference and training
+-------------------------------------------------------------------------------------
+
+The following table applies to both full-precision and half-precision floating point.
+
+.. csv-table::
+   :header: "Combination","N mode","Activations","Constraints"
+   :widths: 30, 15, 15, 15
+
+   "NA for inference","All","All","None"
+   "NA forward training","All","All","None"
+   "NA backward training","All","All","None"
+
+.. note::
+
+   N mode is either spatial or per activation.
 
 Comparing performance with non-fused kernels
 =================================================
@@ -298,6 +386,6 @@ non-fused version. All configurations have a batch size of 64:
 
 The following graph depicts the speedup obtained by fusing BatchNorm (in spatial mode) with activation:
 
-.. image:: ../data/how-to/na.png
+.. image:: ../data/how-to/bn_activ_fused.png
   :width: 800
   :alt: BatchNorm activation fusion

docs/data/how-to/cba.png

@@ -46,7 +46,7 @@ and is performed for each training batch.
 For more information on Batch Normalization, see `Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift <https://arxiv.org/abs/1502.03167>`_.
 
 Enabling or disabling NHWC Batch Normalization for MIOpen using PyTorch
-=============================================================
+=======================================================================
 
 The PyTorch open-source tensor library provides support for using NHWC Batch Normalization with MIOpen.
 In addition to Batch Normalization, NHWC support is also available for convolution and other MIOpen features.

docs/data/how-to/na.png

@@ -27,6 +27,9 @@ The MIOpen public repository is located at `<https://github.com/ROCm/MIOpen>`_.
     * :doc:`Build MIOpen for embedded systems <./install/embed>`
     * :doc:`Build MIOpen using Docker <./install/docker-build>`
 
+.. grid:: 2
+  :gutter: 3  
+
   .. grid-item-card:: Conceptual
 
     * :doc:`Find database <./conceptual/finddb>`
@@ -43,6 +46,10 @@ The MIOpen public repository is located at `<https://github.com/ROCm/MIOpen>`_.
     * :doc:`Use the find APIs and immediate mode <./how-to/find-and-immediate>`
     * :doc:`Use NHWC Batch Normalization with PyTorch <./how-to/use-nhwc-batchnorm-in-pytorch>`
 
+  .. grid-item-card:: Samples
+
+    * `MIOpen samples <https://github.com/ROCm/MIOpen/tree/develop/samples>`_
+
   ..  grid-item-card:: Reference
 
     * :doc:`API library <reference/index>`

docs/how-to/use-fusion-api.rst

@@ -43,6 +43,11 @@ subtrees:
   - file: how-to/use-nhwc-batchnorm-in-pytorch.rst
     title: Use NHWC Batch Normalization with PyTorch
 
+- caption: Samples
+  entries:
+  - url: https://github.com/ROCm/MIOpen/tree/develop/samples
+    title: MIOpen samples
+
 - caption: Reference
   entries:
   - file: reference/index