The IBM Exascale Extrapolator (part of the IBM ExaBounds distribution) scales software profiles to a target scale, e.g., exascale. It takes as input a set of software profiles representative of a parameter space, the input parameters of the profiled software. The parameter sets for those profiles must be generated according to a design of experiments.
The IBM Appresentor in this repository automates this design of experiments and kicks off profiling runs with IBM Platform-Independent Software Analysis, generating the complete set of profiles required as an input to the IBM Exascale Extrapolator.
To get started, please refer to the documentation.
More information on the context of this tool can be found at the Algorithms & Machines project page at at IBM Research – Zurich.