Update docs

docs/installation.rst

@@ -99,6 +99,19 @@ Installing PARDISO
   # or conda install -c conda-forge pypardiso
   # or pip install tdgl[pardiso]
 
+
+GPU acceleration
+----------------
+
+For users with an NVIDIA or AMD GPU, ``tdgl`` can be accelerated using the `CuPy <https://cupy.dev/>`_ library.
+First install the appropriate version of ``cupy`` for your GPU hardware and driver version
+(see installation instructions `here <https://docs.cupy.dev/en/stable/install.html>`_).
+Then set the ``sparse_solver`` attribute of :class:`tdgl.SolverOptions` to ``"cupy"``.
+Due to overheads related to transferring data between the CPU and GPU, it is expected that ``cupy`` will provide
+a speedup only for models with large meshes and/or models that include screening.
+Please open a `GitHub issue <https://github.com/loganbvh/py-tdgl/issues>`_ if you have any problems using ``tdgl`` with ``cupy``.
+Note that ``cupy`` support for AMD GPUs is `currently experimental <https://docs.cupy.dev/en/stable/install.html#using-cupy-on-amd-gpu-experimental>`_.
+
 Verify the installation
 -----------------------
 

tdgl/solver/options.py

@@ -34,20 +34,20 @@ class SolverOptions:
             given solve iteration before giving up.
         adaptive_time_step_multiplier: The factor by which to multiple the time
             step ``dt`` for each adaptive solve retry.
-        sparse_solver: One of "superlu", "umfpack", "pardiso", or "cupy".
-            "umfpack" requires suitesparse, which can be installed via conda,
-            and scikit-umfpack, which can be installed via pip. "pardiso"
+        sparse_solver: One of ``"superlu"``, ``"umfpack"``, ``"pardiso"``, or ``"cupy"``.
+            ``"umfpack"`` requires suitesparse, which can be installed via conda,
+            and scikit-umfpack, which can be installed via pip. ``"pardiso"``
             requires an Intel CPU and the pypardiso package, which can be
-            installed via pip or conda. "cupy" requires and  NVIDIA GPU
-            and the CuPy Python package.
+            installed via pip or conda. ``"cupy"`` requires a GPU and the
+            CuPy Python package, which can be installed via pip.
         terminal_psi: Fixed value for the order parameter in current terminals.
         field_units: The units for magnetic fields.
         current_units: The units for currents.
         output_file: Path to an HDF5 file in which to save the data.
             If the file name already exists, a unique name will be generated.
             If ``output_file`` is ``None``, the solver results will not be saved
             to disk.
-        pause_on_interrupt: Pause the simulation in the event of a KeyboardInterrupt.
+        pause_on_interrupt: Pause the simulation in the event of a ``KeyboardInterrupt``.
         save_every: Save interval in units of solve steps.
         progress_interval: Minimum number of solve steps between progress bar updates.
         include_screening: Whether to include screening in the simulation.