The ParaMonte kernel code coverage report

The code coverage report analyses of the ParaMonte kernel library can be generated for a wide range of input build configurations. For example, corresponding to each programming language (C, C++, Fortran, MATLAB, Python, R, …), a separate code coverage report for the kernel routines can be generated. Similarly, different library builds (debug, testing, release), types (static, shared), and memory allocation (stack or heap) can lead to different code coverage reports.

Among all build configurations, however, the single most important specification that yields the largest differences in code coverage reports is the parallelism paradigm (none, mpi, cafsingle, cafshared, cafdistributed). The the three Coarray (CAF) parallelism subtypes all yield the same code coverage reports. However, the code coverage reports for the serial, MPI, and Coarray modes are remarkably different since each parallelism type activates a different set of codes in the ParaMonte kernel routines.

As such, the code coverage reports are only generated separately for the three library parallelism builds. The corresponding latest code coverage reports are permanently available at,

The ParaMonte kernel code coverage report generation

The GNU GCOV instrumentation files and the subsequent LCOV code coverage analysis reports can be readily generated via the following install script command on a Linux or macOS machine, or a Microsoft Windows Subsystem for Linux,

./ --codecov

By default, the above command will statically build the serial kernel library for the Fortran programming language in debug mode with heap memory allocation and will generate the corresponding GCOV instrumentation files and store the final LCOV code coverage report in a folder codecov/kernel/serial in the root directory of the ParaMonte library (where the script exists). To generate code coverage for the MPI parallel kernel, simply add the following script flag,

./ --codecov --par mpi

or, for the Coarray parallel kernel,

./ --codecov --par cafshared

or for the serial and all possible parallelism modes,

./ --codecov --par "none mpi cafshared"

The meaning and the syntax rules of the flags follow those of the ParaMonte installation procedure on Linux or on macOS.

By default, the build script will perform all kernel tests. As of Dec 2020, there are $>883$ tests available that are tailored to test different functionalities for different parallelism builds. Broadly speaking, the tests can be categorized into two separate groups,

  1. Basic tests, which test the functionality and accuracy of various fundamental tasks in the library.
  2. Sampler tests, which test the functionality and accuracy of the ParaMonte sampler routines for different simulation settings.

To focus the code coverage on only a particular aforementioned category of tests, simply provide its name along with the flag -t or --test on the command prompt when calling the build script like,

./ --codecov --test basic


./ --codecov -t sampler

To get help, one can always try,

./ --help

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.