Overview | Installation | Dependencies | Parallelism | Examples | Acknowledgments | License | Authors

ParaMonte: Parallel Monte Carlo and Machine Learning

ParaMonte is a serial/parallel library of Monte Carlo routines for sampling mathematical density functions of arbitrary dimensions and Machine Learning (ML) algorithms for scientific inference, with the design goal of unifying automation (of simulations and tasks), user-friendliness (of algorithms), accessibility (from any platform or programming environment), high-performance (at runtime), and scalability (across many parallel processors).

For more information on the installation, usage, and examples, visit https://www.cdslab.org/paramonte

ParaMonte Design Goals

ParaMonte has been developed while bearing the following design goals in mind:

• Full automation of Monte Carlo and Machine Learning simulations as much as possible to ensure user-friendliness of the library and minimal time investment requirements for building, running, and post-processing simulation models.

• Interoperability of the core library with as many programming languages as currently possible, including C, C++, Fortran, MATLAB, and Python, with ongoing efforts to support other popular programming languages.

• High-Performance meticulously-low-level library implementation to ensure the fastest-possible inferences.

• Parallelizability of all simulations via shared-memory OpenMP threads and two-sided and one-sided MPI/Coarray distributed communications while requiring zero-parallel-coding efforts by the user.

• Zero-dependence on external libraries to ensure hassle-free ParaMonte library builds and ParaMonte simulation runs.

• Fully-deterministic reproducibility and automatically-enabled restart functionality for all simulations with arbitrary digits of precision as requested by the user.

• Comprehensive reporting and post-processing of simulations and ML tasks, as well as their automatic storage in external files to ensure the simulation results will be understandable and reproducible at any time in the foreseeable future.

Quick Start

• Follow the quick start instructions in this QUICKSTART.md file.
• For programming languages other than C, C++, Fortran, see the quick start section of the README.md file in the corresponding language's source subfolder in the src folder in the root directory of the ParaMonte repository.

Installation

The prebuilt, ready-to-use libraries are available on the release page of the ParaMonte library on GitHub.
Each prebuilt ParaMonte library automatically ships with a full-fledged set of example codes and build scripts.

Alternatively, you can build the library from the source in the project's GitHub repository.
The ParaMonte library installation/build process is fully automated for all supported programming languages.
Currently, the following compiler suites are supported for builds from source,

Compiler Suite	Windows (amd64)	Linux (amd64)	macOS (amd64)	macOS (arm64)
GNU Compiler Collection >= 10.3	✅	✅	✅	✅
Intel OneAPI >= 2021.8.0	✅	✅	✅	❌

with the following essential build systems,

Dependency	Windows (amd64)	Linux (amd64)	macOS (amd64)	macOS (arm64)
CMake >= 3.21	✅	✅	✅	✅

• Read the install.md installation instruction notes to learn about the library's optional build configurations.
• For a quickstart in the programming language of your choice, visit the ParaMonte library homepage.

Dependencies

Aside from an optional MPI runtime library for MPI-parallel simulations, the ParaMonte library core has zero dependency on external third-party libraries or packages.

Parallelism

The ParaMonte library relies on three independent parallelism paradigms for parallel applications:

1. The Coarray parallelism (only available for select routines in the Fortran programming language).
2. The Message Passing Interface (MPI) standard for inter-processor distributed parallelism.
   • On Windows and Linux operating systems, we highly recommend downloading and installing the Intel MPI runtime libraries, which is available to the public free of charge.
     The ParaMonte auxil prerelease exclusively contains the optional Intel MPI runtime libraries for users' convenience.
     The Intel MPI requires the ParaMonte library to be (or have been prebuilt) using the Intel compilers.
   • On macOS (where Intel MPI is unavailable) and Linux operating systems, users can install either Open-MPI or MPICH MPI runtime libraries depending on the prebuilt version of the ParaMonte library that you have downloaded or the configuration with which you intend to build the library.

Optional Dependency Compatibility

Dependency	Windows (amd64)	Linux (amd64)	macOS (amd64)	macOS (arm64)
Intel MPI (IMPI) >= 2021.8	✅	✅	✅	❌
MPICH MPI (MMPI) >= 3	❌	✅	✅	✅
OpenMPI (OMPI) >= 4	❌	✅	✅	✅

Read the install.md installation instruction notes to learn more about the library's optional parallel build configurations.

For more information, visit https://www.cdslab.org/paramonte/.

Example Usage Instructions

• The best way to get started with example usage is to build the language-specific library examples.
  The library example build and run scripts are automatically generated for each ParaMonte build and are automatically stored in the output installation directory of the build.

• For complete, organized, up-to-date instructions, visit: cdslab.org/pm

• For a quick look into language-specific README.md instructions, visit:

  • C: https://github.com/cdslaborg/paramonte/tree/main/src/c
  • C++: https://github.com/cdslaborg/paramonte/tree/main/src/cpp
  • Fortran: https://github.com/cdslaborg/paramonte/tree/main/src/fortran
  • MATLAB: https://github.com/cdslaborg/paramonte/tree/main/src/matlab

• The ParaMonte library for Python is currently undergoing internal testing before their impending release. For now, the previous versions (1) of the ParaMonte Python libraries can be downloaded and used from the previous library major release on GitHub Release Page.

Citing ParaMonte

The ParaMonte library is an honor-ware and its currency is acknowledgment and citations.

If you use ParaMonte or any ideas from this software, please acknowledge it by citing the ParaMonte library's main publications as listed in ACKNOWLEDGMENT.md.

Visit the ParaMonte library homepage to access the PDF version of these files free of charge.

License

The majority of the ParaMonte library routines are distributed under the permissive MIT License plus acknowledgments that are detailed in LICENSE.md.

However, there are occasionally modernized and extended routines from external projects, particularly within the ParaMonte Fortran library. These routines may have a license that does not fully overlap with the default ParaMonte library license LICENSE.md. In such cases, the documentation and source code of the specific routine explicitly list the original license information.

This is free software, so help us keep it freely available to the public by acknowledging the usage and contributing to it. If you have questions or concerns about the license, please get in touch with the project's lead (shahmoradi@utexas.edu).

Authors and Contributors

• Amir Shahmoradi

  • Astrophysicist/bioinformatician by training,
  • Ph.D. in computational physics/bioinformatics, The University of Texas at Austin,
  • Currently a faculty member of the Data Science and Physics programs at UT Arlington,
  • Contact: shahmoradi@utexas.edu

• Fatemeh Bagheri

  • PhD in Astronomy,
  • PhD in Space Physics,
  • Deep philosophical thinker,
  • Currently at NASA Goddard Space Flight Center,
  • Contact: bagheri.fateme@gmail.com

• Joshua Osborne

  • Physicist / computational data scientist by training,
  • Currently a Physics PhD candidate at UT Arlington,
  • Contact: joshuaalexanderosborne@gmail.com

Contributing to ParaMonte

Please read the generic contribution guidelines listed in CONTRIBUTING.md.

For more information, visit cdslab.org/pm or contact Amir Shahmoradi: shahmoradi@utexas.edu