ParaMonte kernel C/C++/Fortran release notes

This project follows Semantic Versioning. To access the latest release of the package, visit the ParaMonte GitHub repository release page.

See also,

  • CHANGES.md for the ParaMonte::MATLAB release notes.
  • CHANGES.md for the ParaMonte::Python release notes.

Version 1.x.x

Version 1.5.1 – January 1, 2020

Minor Enhancements

  • This is a minor enhancement release, but is a major step toward further portability of the ParaMonte library. All ParaMonte kernel library dependencies are now properly handled and recognized at runtime without such aggressive actions as permanently redefining the environmental path variables, most importantly, PATH and LD_LIBRARY_PATH on Linux/macOS.

  • The ParaMonte kernel routines are now capable of handling user-input file paths that contain white-space (blank) or other exotic characters.

  • The Bash build script for the ParaMonte C/C++/Fortran examples can now handle file paths that contain white-space (blank) or other exotic characters.

  • The shared (dynamic) library file naming convention is now changed from using “dynamic” in the file name to using “shared” in the file name.

  • Improved error-handling messages in the examples’ build and run scripts.

  • Typo-fixes in the documentation of the library.

Compiler Compatibility

Compiler Suite Windows (64bit) Linux macOS
GNU Compiler Collection > 8.4
Intel Parallel Studio > 19.1.1
Microsoft C/C++ Compiler > 16.0.0

Compiler / MPI library used for this binary release

  • Windows: Intel Parallel Studio Version 19.1.1.216 Build 20200306 / Intel(R) MPI Library 2019 Update 7 for Windows
  • Linux: Intel Parallel Studio Version 19.1.1.217 20200306 / Intel(R) MPI Library for Linux OS, Version 2019 Update 7 Build 20200312
  • Linux: GNU 10.1.0 / Open-MPI 4.0.3
  • Linux: GNU 10.1.0 / MPICH 3.2
  • macOS: Intel Parallel Studio Version 19.1.0.216 20200306
  • macOS: GNU 10.2.0 / Open-MPI 4.0.5
  • macOS: GNU 10.2.0 / MPICH 3.3.2

Version 1.5.0 – December 17, 2020

Major Enhancements

  • This version introduces numerous performance and accuracy enhancements to the ParaMonte library.
  • The entire kernel library is now fully documented and verified with over 866 tests that cover close to 100% of all lines and functions in the kernel.
  • New prebuilt libraries with GNU compilers and Open-MPI on Linux are added.
  • New flags are now added to the build scripts of the library that automate the process of code coverage generation.
  • The testing builds are now removed from the ParaMonte release page as this build is mostly useful for development purposes.
  • The issue of Windows file locking, that led to the occasional crashes of the ParaDRAM and ParaDISE simulations in multiChain parallelism mode, is now resolved.

Minor Enhancements

  • Minor enhancements to the ParaMonte C/C++/Fortran example build scripts build.sh and build.bat.
  • The default build settings are now limited to heap memory allocation with dynamic library builds for only serial and MPI parallelization for all languages. The stack memory allocation results in a ~10% gain in the efficiency of the code. The benefits of stack-memory builds are marginal and are often problematic, in particular, for usage with non-compiled languages. However, users can still build the library with stack memory allocation by specifying the appropriate build flags with the install.sh on Unix or install.bat script on Windows systems. For further information, see the installation guidelines on the ParaMonte documentation website.
  • All temporary array creations in debug mode are now resolved, except when Intel compilers are used, in which case, the debug warning messages are silenced.

Compiler Compatibility

Compiler Suite Windows (64bit) Linux macOS
GNU Compiler Collection > 8.4
Intel Parallel Studio > 19.1.1
Microsoft C/C++ Compiler > 16.0.0

Compiler / MPI library used for this binary release

  • Windows: Intel Parallel Studio Version 19.1.1.216 Build 20200306 / Intel(R) MPI Library 2019 Update 7 for Windows
  • Linux: Intel Parallel Studio Version 19.1.1.217 20200306 / Intel(R) MPI Library for Linux OS, Version 2019 Update 7 Build 20200312
  • Linux: GNU 10.1.0 / Open-MPI 4.0.3
  • Linux: GNU 10.1.0 / MPICH 3.2
  • macOS: Intel Parallel Studio Version 19.1.0.216 20200306
  • macOS: GNU 10.2.0 / Open-MPI 4.0.5
  • macOS: GNU 10.2.0 / MPICH 3.3.2

Version 1.4.1 – November 15, 2020

Enhancements

  • The ParaMonte C/C++/Fortran example build scripts build.sh and build.bat now check for the existence of both Intel and GNU compilers in the appropriate order that is automatically inferred at the compilation time. Also, the dependencies on the MPI compiler wrappers is now removed as the MPI libraries are not required to build the ParaMonte examples, even in cases of parallel ParaMonte example builds.

Compiler support

Compiler Suite Windows (64bit) Linux macOS
GNU Compiler Collection > 7.5
Intel Parallel Studio > 18.0.0
Microsoft C/C++ Compiler > 16.0.0

Compiler / MPI library used for this binary release

  • Windows: Intel Parallel Studio Version 19.0.4.245 Build 20190417 / Intel(R) MPI Library 2019 Update 4 for Windows
  • Linux: Intel Parallel Studio Version 18.0.2 20180210 / Intel(R) MPI Library for Linux OS, Version 2018 Update 2 Build 20180125
  • Linux: GNU 10.1.0 / MPICH 3.2
  • macOS: Intel Parallel Studio Version 19.1.0.216 20200306
  • macOS: GNU 10.2.0 / Open-MPI 4.0.5

Version 1.4.0 – October 29, 2020

Enhancements

  • The IO debugging info of all ParaMonte samplers have been enhanced. In cases of wrong syntax or syntax-breaking input values in the simulation output files, the error messages are now more informative and point directly to the exact location of of error in the input file.

  • The Integrated Autocorrelation (IAC) for sample refinement in ParaDRAM sampler of ParaMonte is now set to the average of all variables’ IAC values instead of the maximum IAC value. This will lead to less aggressive decorrelation of the final sample, which means significantly larger final sample sizes, without compromising the i.i.d. property of the final refined sample. This behavior can be reversed back to the original by specifying “max” or “maximum” along with the requested refinement method, SampleRefinementMethod = "batchmeans max" or SampleRefinementMethod = "BatchMeans-max" (case-insensitive).

Compiler support

Compiler Suite Windows (64bit) Linux macOS
GNU Compiler Collection > 7.5
Intel Parallel Studio > 18.0.0
Microsoft C/C++ Compiler > 16.0.0

Compiler / MPI library used for this binary release

  • Windows: Intel Parallel Studio Version 19.0.4.245 Build 20190417 / Intel(R) MPI Library 2019 Update 4 for Windows
  • Linux: Intel Parallel Studio Version 18.0.2 20180210 / Intel(R) MPI Library for Linux OS, Version 2018 Update 2 Build 20180125
  • Linux: GNU 9.1 / MPICH 3.2
  • macOS: Intel Parallel Studio Version 19.1.0.216 20200306
  • macOS: GNU 10.2.0 / Open-MPI 4.0.5

Version 1.3.0 – October 3, 2020

Enhancements

  • A new simulation specification overwriteRequested has been added to all ParaMonte samplers. If TRUE and the ParaMonte sampler detects an existing set of old simulation output files in the output path of the current simulation with the same names as the output file names of the current simulation, then, the ParaMonte sampler will overwrite the existing simulation files.

Compiler support

Compiler Suite Windows (64bit) Linux macOS
GNU Compiler Collection > 7.5
Intel Parallel Studio > 18.0.0
Microsoft C/C++ Compiler > 16.0.0

Compiler / MPI library used for this binary release

  • Windows: Intel Parallel Studio Version 19.0.4.245 Build 20190417 / Intel(R) MPI Library 2019 Update 4 for Windows
  • Linux: Intel Parallel Studio Version 18.0.2 20180210 / Intel(R) MPI Library for Linux OS, Version 2018 Update 2 Build 20180125
  • Linux: GNU 9.1 / MPICH 3.2
  • macOS: Intel Parallel Studio Version 19.1.0.216 20200306
  • macOS: GNU 10.2.0 / Open-MPI 4.0.5

Version 1.2.0 – September 22, 2020

Enhancements

  • The post-processing report in the output report file of ParaDRAM simulation has been significantly improved:
    • The parallel simulation summary now also provides the predicted strong-scaling speedup behavior of the parallel ParaDRAM simulations in “single chain” parallelism mode. This can help make wiser decisions regarding the the number of processors for similar parallel simulations in the future.
  • The ParaDRAM restart output file in ASCII mode now contains all proposal updates, including the first user-specified proposal specs.

  • All parallel simulations now avoid the unnecessary creation of temporary files by all processors for System and OS operations. This is particularly important for large-scale parallel simulations. As a side effect, this will also potentially improve the runtime performances of the simulation.

  • Major enhancements has been made to the parallel simulation performance analysis reported in the post-processing section of the ParaDRAM simulation output _report.txt files.

  • The ParaMonte C/C++ build processes are now separate from each other.

Bug fixes

  • minor typo fixes

Compiler support

Compiler Suite Windows (64bit) Linux macOS
GNU Compiler Collection > 7.5
Intel Parallel Studio > 18.0.0
Microsoft C/C++ Compiler > 16.0.0

Compiler / MPI library used for this binary release

  • Windows: Intel Parallel Studio Version 19.0.4.245 Build 20190417 / Intel(R) MPI Library 2019 Update 4 for Windows
  • Linux: Intel Parallel Studio Version 18.0.2 20180210 / Intel(R) MPI Library for Linux OS, Version 2018 Update 2 Build 20180125
  • Linux: GNU 9.1 / MPICH 3.2
  • macOS: Intel Parallel Studio Version 19.1.0.216 20200306
  • macOS: GNU 10.2.0 / Open-MPI 4.0.5

Version 1.1.0 – May 27, 2020

New features

  • The MATLAB interface to the ParaMonte library is now ready to use, in addition to the existing C/C++/Fortran/Python Programming language interfaces to ParaMonte.

Enhancements

  • The simulation summary report in the output report file has been improved.
  • The error handling in the build Batch-scripts on Windows is now greatly improved.

Bug fixes

  • The build Batch-script for the ParaMonte examples on Windows now properly builds and runs coarray applications in parallel.
  • The fully-deterministic restart functionality is now functional also when chainFileFormat=”verbose” in ParaDRAM simulations.

Compiler support

Compiler Suite Windows (64bit) Linux macOS
GNU Compiler Collection > 7.5
Intel Parallel Studio > 18.0.0
Microsoft C/C++ Compiler > 16.0.0

Compiler / MPI library used for this binary release

  • Windows: Intel Parallel Studio Version 19.0.4.245 Build 20190417 / Intel(R) MPI Library 2019 Update 4 for Windows
  • Linux: Intel Parallel Studio Version 18.0.2 20180210 / Intel(R) MPI Library for Linux OS, Version 2018 Update 2 Build 20180125
  • Linux: GNU 8.3 / MPICH 3.2
  • macOS: Intel Parallel Studio Version 19.1.0.166 20191121
  • macOS: GNU 9.3 / Open-MPI 4.0

Version 1.0.0 – January 1, 2020 – Initial release

This is the first public release of the ParaMonte library.

New features

  • ParaDRAM sampler: Parallel Delayed-Rejection Adaptive Metropolis-Hastings Markov Chain Monte Carlo Sampler.
  • ParaMonte Interface to the C/C++/Fortran/Python Programming languages.
  • ParaMonte simulation-output visualization via the ParaMonte Python interface.

Compiler support

Compiler Suite Windows (64bit) Linux macOS
GNU Compiler Collection > 7.5
Intel Parallel Studio > 18.0.0
Microsoft C/C++ Compiler > 16.0.0

Compiler / MPI library used for this binary release

  • Windows: Intel Parallel Studio Version 19.0.4.245 Build 20190417 / Intel(R) MPI Library 2019 Update 4 for Windows
  • Linux: Intel Parallel Studio Version 18.0.2 20180210 / Intel(R) MPI Library for Linux OS, Version 2018 Update 2 Build 20180125
  • Linux: GNU 8.3 / MPICH 3.2
  • macOS: Intel Parallel Studio Version 19.1.0.166 20191121
  • macOS: GNU 9.3 / Open-MPI 4.0


If you have any questions about the topics discussed on this page, feel free to ask in the comment section below, or raise an issue on the GitHub page of the library, or reach out to the ParaMonte library authors.