Main innovation: provide an interactive website which can be used and developed for first performance analysis of PinT algorithms.
We have developed different educational web applications (WebApps), which each consists of three columns (see also figure below):
- Documentation: displays some quick insights on what the webapp does, eventually with some illustrating formulas
- Parameters: control the parameter values that have to be set to run the webapp
- Results: shows some plots and values computed for the given parameters
Below is an example of the webapp that can be used to estimate the parallel speedup and efficiency of the Parareal algorithms, knowing the number of time sub-intervals, the number of iterations, and the computation cost of the fine and the coarse solver. Using the webapp, several parallelization strategies can be investigated, assisting HPC software developers to find the one that fits a given application best.
WebApp access
- Source code: can be found at https://github.com/Parallel-in-Time/time4apint
- Website: can be found at https://jupyterhub.mat.tu-harburg.de/blockops/