improve styling and link to benchmark script in Performance

docs/conf.py

@@ -52,6 +52,8 @@ def run(self):
 rst_prolog = f'''
 .. _repo: {repo}
 .. |repo| replace:: **Source Code**
+.. _script: {repo_main + '/bench/gather.py'}
+.. |script| replace:: *script*
 .. _custom: {repo_src + '/custom'}
 .. |custom| replace:: *custom*
 .. _custom_cmake: {repo_main + '/cmake/custom.cmake'}
@@ -62,6 +64,11 @@ def run(self):
 .. |doc| replace:: **Documentation**
 .. |name| replace:: *{ai3.__name__}*
 .. |pkg_name| replace:: *{pkg_name}*
+
+.. _cuDNN: https://developer.nvidia.com/cudnn
+.. |cuDNN| replace:: *cuDNN*
+.. _SYCL: https://www.khronos.org/sycl
+.. |SYCL| replace:: *SYCL*
 '''
 
 project = ai3.__name__

docs/performance.rst

@@ -20,9 +20,9 @@ Performance
 
 .. _details:
 
-The `cuDNN <https://developer.nvidia.com/cudnn>`_ and `SYCL
-<https://www.khronos.org/sycl/>`_ benchmarks for both *ai3* and *PyTorch* were
+The |cudnn|_ and |sycl|_ benchmarks for both *ai3* and *PyTorch* were
 gathered using an *NVIDIA GeForce L40S GPU* with *16* gigabytes of memory. The
 final latencies used are the average over *10* runs after *10* warm up runs.
 The implementations for the algorithms include select ones provided by *cuDNN*
-and implementations from *ai3* which leverage *SYCL*.
+and implementations from *ai3* which leverage *SYCL*. Benchmarks are
+gathered using this |script|_.