Intel® MPI Library Developer Guide for Windows* OS
Intel® MPI Library provides a tuning utility mpitune, which allows you to automatically adjust Intel® MPI Library parameters, such as collective operation algorithms, to your cluster configuration or application. The tuner iteratively launches a benchmarking application with different configurations to measure performance and stores the results of each launch. Based on these results, the tuner generates optimal values for the parameters that are being tuned.
For example, you can use the tuner to identify the best MPI_Bcast algorithms (specified with I_MPI_ADJUST_BCAST) for specific message sizes. The result may look as follows:
I_MPI_ADJUST_BCAST=2:0-0;1:1-64;2:65-509;1:510-8832;3:8833-0
Use the resulting variable value with the application launch to achieve performance gain.
All tuner parameters should be specified in a configuration file, including the library parameter to be tuned, the application launch command and other parameters. You can modify the sample configuration file etc/tune_cfg/example according to your environment (for example, to specify other collective operations to be tuned) and use it for tuning as follows:
> mpitune --config <installdir>/etc/tune_cfg/config
The mpitune usage model has changed since the 2018 release. Tuning parameters should now be specified in a configuration file rather than as command-line options.
For details on the mpitune configuration options, refer to the Developer Reference, section Command Reference > mpitune.