mpiP / News: Recent posts

Downloading

The current version of mpiP can be accessed at https://github.com/LLNL/mpiP/releases/latest.

Documentation

The mpiP User Guide can be found at https://software.llnl.gov/mpiP/.

We are pleased to announce the release of version 3.2 of the mpiP lightweight profiling library for MPI applications. Release v3.2 addresses the following issues: Support for MPI RMA functions, Support for glibc backtrace, and Defaulting to MPI_Wtime if platform-specific timers are not found.

We are pleased to announce the release of version 3.1.2 of the mpiP lightweight profiling library for MPI applications.

Release v3.1.2 addresses the following issues:
* Better MPI support for Init_thread, Testany, Testsome, Waitany, and Waitsome.
* Improved support for linux-mips64.
* Added option to configure for generating weak Fortran symbols in the case of multiple Fortran mangling schemes in the application object files (--enable-fortranweak).
* Addressed various outstanding issues (see ChangeLog for more details).

We are pleased to announce the release of version 3.1.1 of the mpiP lightweight profiling library for MPI applications.

Release v3.1.1 addresses the following issues:

* Revert to gettimeofday as default Linux timer.
* MIPS64-Linux stack walking support.
* Catamount dclock timer support.
* Greater install flexibility:
o 'install' target only installs lib and doc files.
o 'install-api', 'install-bin', 'install-all' targets provide additional install functionality.
o New 'uninstall' target.

We are pleased to announce the release of version 3.1.0 of the mpiP lightweight profiling library for MPI applications.

Release v3.1 addresses the following issues:

* Configure defaults to the optimal platform timer, but the timer can be specified with configure options.
* Support for rts_get_timebase on BG/L.
* Runtime flags can be used to specify that report generation be done by a single collector process (default) or by using a small fixed amount of memory and MPI collectives. The single collector process approach tends to be faster while the low-memory approach is useful for systems where memory availability may be a concern.

We are pleased to announce the release of version 3.0.1 of the mpiP lightweight profiling library for MPI applications.

Release v3.0 addresses the following issues:

* New reporting formats with summary statistics ( See runtime options 'c' and 'v' ).
* New reporting algorithms to reduce memory requirements ( See configure flag --use-low-mem-report ).
* MPI_Pcontrol calls can reset statistics, generate multiple reports per run, and generate reports in various formats.... read more