Added GPU install procedure

Author: JeanLucPons

docs/p/howto/multiprocessing.rst

@@ -111,3 +111,55 @@ to a `MPI <https://www.mpi-forum.org/docs/>`_ one or vice-versa, remove the
     rm -rf build
     pip install --config-settings mpi=1 ".[mpi]"
 
+`GPU Tracking`_
+---------------
+PyAT can be installed with GPU support, either `OpenCL <https://github.com/KhronosGroup/OpenCL-Guide/tree/main>`_ or
+`CUDA <https://developer.nvidia.com/cuda-toolkit>`_, compatibility. GPU are especially interesting for tracking large
+number of particle. The performance of the tracking is mainly related to the GPU double precision arithmetic performance.
+
+OpenCL Installation
+...................
+`OpenCL`_ and installable client driver (ICD) must be preliminary installed on the system either using linux packages or by
+building the `OpenCL SDK <https://github.com/KhronosGroup/OpenCL-SDK>`::
+
+    sudo apt install opencl-headers ocl-icd-opencl-dev -y
+
+Then you need to set the environment variable ``OCL_PATH`` to the SDK install path if you don't use a standard install::
+
+    export OCL_PATH=<sdk_intall_path>
+or (on Windows)::
+
+    set OCL_PATH=C:\clpeak\build\sdk_instal
+
+Note: `clpeak <https://github.com/krrishnarraj/clpeak>` is a great OpenCL benchmarking tool that can be used to check system
+performance (especially double precision floating point arithmetic) and to build the OpenCL SDK.
+
+Install PyAT using the ``opencl`` flag::
+
+    cd <at>
+    rm -rf build
+    pip install --config-settings opencl=1 .
+
+You can check the install as bellow, when no GPU support is enable, the method ``at.tracking.gpu_info()`` will return an empty list::
+
+    Z:\at>python
+    Python 3.11.5 (tags/v3.11.5:cce6ba9, Aug 24 2023, 14:38:34) [MSC v.1936 64 bit (AMD64)] on win32
+    Type "help", "copyright", "credits" or "license" for more information.
+    >>> import at
+    >>> at.tracking.gpu_info()
+    [['NVIDIA GeForce GTX 1050 Ti', '6.1', 6, 'NVIDIA CUDA OpenCL 3.0 CUDA 12.3.68'], ['Intel(R) UHD Graphics 630', '0.0', 24, 'Intel(R) OpenCL HD Graphics OpenCL 3.0 ']]
+
+CUDA Installation
+.................
+
+`NVidia CUDA`_ toolkit must be preliminary installed on the system from `NVidia <https://developer.nvidia.com/cuda-downloads>`.
+Set the environment variable ``CUDA_PATH``::
+
+    export CUDA_PATH=/cvmfs/hpc.esrf.fr/software/packages/ubuntu20.04/x86_64/cuda/12.3.1
+or on Windows::
+
+    set CUDA_PATH=CUDA_PATH=C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.3
+Install PyAT using the ``cuda`` flag::
+
+    pip install --config-settings cuda=1 .
+You can check the install using the method ``at.tracking.gpu_info()`` as described above.