Documentation of the libraries

conda.yaml

@@ -1,12 +1,9 @@
 # Definition of the Conda Environment.
 #
-# WARNING: The GPU libraries (CUDA, CuPy) should be installed separately.
-#          These libraries are not included in the Conda Environment.
-#          The GPU support is extremely platform/version dependent.
-#
-# WARNING: Jupyter is not included in the Conda Environment.
-#
-# WARNING: FFTW and UMFPACK are not included in the Conda Environment.
+# WARNING: Only the required packages are in the environment.
+#          The following optional libraries are not installed:
+#              - Jupyter and Trame
+#              - CuPy, FFTW, UMFPACK, and MKL/PARDISO
 #
 # (c) Thomas Guillod - Dartmouth College
 

docs/technical.md

@@ -71,16 +71,20 @@ Afterwards, a custom configuration (JSON or YAML) file can be set:
 > **Warning**: The plotting code is likely sensitive to the environment (platform and the version of the libraries).
 > More particularly, the interactions between **Qt/PyVista/Matplotlib** and **Jupyter/PyVista/Matplotlib**.
 > Therefore, the Qt/Jupyter dependencies are minimized and insulated from the rest of the code.
+> The plotting code (viewer and plotter) is separated from the simulation code (mesher and solver).
 
-> **Warning**: The optional **GPU libraries** (CUDA, CuPy) should be installed separately.
-> These libraries are not included in the package dependencies and environments. 
-> The GPU support is extremely platform/version dependent (GPU type and CUDA version).
+> **Warning**: **Jupyter** is not included in the package dependencies.
+> For Jupyter, PyVista has to be installed with the optional Trame support.
+> Jupyter support is optional, PyPEEC is fully functional without Jupyter.
 
-> **Warning**: **Jupyter** is not included in the package dependencies and environments.
+> **Warning**: The **GPU** libraries (CUDA, CuPy) are not included in the package dependencies.
+> The GPU support is extremely hardware/platform/version dependent).
+> GPU support is optional, PyPEEC is fully functional without GPU support.
 
-> **Warning**: The optional **UMFPACK** library is known to be difficult to install on MS Windows.
-
-> **Warning**: The optional **FFTW** library should be compiled with multithreading support.
+> **Warning**: **FFTW**, **UMFPACK**, and **MKL/PARDISO** are not included in the package dependencies.
+> These libraries can be tricky to install, especially on MS Windows.
+> Make sure that these libraries are compiled with multithreading support.
+> FFTW, UMFPACK, and MKL/PARDISO are optional, PyPEEC is fully functional without them.
 
 # General Warnings
 
@@ -92,5 +96,5 @@ Afterwards, a custom configuration (JSON or YAML) file can be set:
 > Only load Pickle files that you trust.
 > Do not commit the Pickle files in the Git repository.
 
-> **Warning**: Some dependencies are under copyleft licences.
-> Make sure to respect these licenses if you package these libraries.
+> **Warning**: Some dependencies are under various licences (including copyleft and non-free).
+> Make sure to respect these licenses if you package and/or distribute these libraries.

examples/run_config.py

@@ -297,8 +297,8 @@ def get_data_tolerance():
     factorization_options = {
         # matrix factorization options
         #   - SuperLU is typically slower but is always available (integrated with SciPy)
-        #   - UMFPACK is typically faster but has to be installed separately
-        #   - PARDISO is typically faster but has to be installed separately
+        #   - UMFPACK is typically faster than SuperLU (with SciKits)
+        #   - PARDISO is typically faster than UMFPACK (with Pydiso)
         #   - GCROT is quite unstable but has low memory requirements (integrated with SciPy)
         #   - BICG is quite unstable but has low memory requirements (integrated with SciPy)
         #   - GMRES is quite unstable but has low memory requirements (integrated with SciPy)

pypeec/lib_matrix/fourier_transform.py

@@ -1,10 +1,10 @@
 """
+Module for computer FFT/iFFT of tensors.
+
 This module is used as a common interface for different FFT libraries:
-    - NumPy FFT library
     - SciPy FFT library
     - FFTW FFT library (available through pyFFTW)
-
-WARNING: Not all versions of FFTW are compiled with multithreading support.
+    - CuPy FFT library (computation with GPUs)
 """
 
 __author__ = "Thomas Guillod"

pypeec/lib_matrix/matrix_factorization.py

@@ -1,22 +1,13 @@
 """
 Module for factorizing sparse matrix.
-This module is used as a common interface for different solvers:
-    - SuperLU: built-in library (with SciPy)
-    - UMFPACK: extra library (with SciKits)
-    - PARDISO: extra library (with Pydiso)
-    - GCROT: built-in library (with SciPy)
-    - BICG: built-in library (with SciPy)
-    - GMRES: built-in library (with SciPy)
-
-SuperLU is available on all platforms and easy to install.
-UMFPACK and PARDISO need to be installed separately.
-
-GCROT, BICG, and GMRES are iterative solvers and are used without preconditioner.
-Therefore, these solvers are typically unstable and should only be used for particular problems.
 
-WARNING: UMFPACK is difficult to install on MS Windows.
-
-WARNING: PARDISO is difficult to install on MS Windows.
+This module is used as a common interface for different solvers:
+    - SuperLU is typically slower but is always available (integrated with SciPy)
+    - UMFPACK is typically faster than SuperLU (with SciKits)
+    - PARDISO is typically faster than UMFPACK (with Pydiso)
+    - GCROT is quite unstable but has low memory requirements (integrated with SciPy)
+    - BICG is quite unstable but has low memory requirements (integrated with SciPy)
+    - GMRES is quite unstable but has low memory requirements (integrated with SciPy)
 """
 
 __author__ = "Thomas Guillod"

requirements.txt

@@ -1,12 +1,9 @@
 # Definition of the Python Virtual Environment.
 #
-# WARNING: The GPU libraries (CUDA, CuPy) should be installed separately.
-#          These libraries are not included in the Python Virtual Environment.
-#          The GPU support is extremely platform/version dependent.
-#
-# WARNING: Jupyter is not included in the Python Virtual Environment.
-#
-# WARNING: FFTW and UMFPACK are not included in the Python Virtual Environment.
+# WARNING: Only the required packages are in the environment.
+#          The following optional libraries are not installed:
+#              - Jupyter and Trame
+#              - CuPy, FFTW, UMFPACK, and MKL/PARDISO
 #
 # (c) Thomas Guillod - Dartmouth College
 

setup.cfg

@@ -1,18 +1,15 @@
 # Configuration file for defining the package data.
 #
-# WARNING: The GPU libraries (CUDA, CuPy) should be installed separately.
-#          These libraries are not included in the package dependencies.
-#          The GPU support is extremely platform/version dependent.
-#
-# WARNING: Jupyter is not included in the Python Virtual Environment.
-#
-# WARNING: FFTW and UMFPACK are not included in the Python Virtual Environment.
+# WARNING: Only the required packages are listed as the package dependencies.
+#          The following optional libraries are not installed:
+#              - Jupyter and Trame
+#              - CuPy, FFTW, UMFPACK, and MKL/PARDISO
 #
 # (c) Thomas Guillod - Dartmouth College
 
 [metadata]
 name = pypeec
-description = PyPEEC - 3D PEEC Solver
+description = PyPEEC - 3D PEEC Solve
 license = All rights reserved
 author = Thomas Guillod
 author_email = [email protected]