Open ivotron opened 5 years ago
@js1019 does the above sound OK?
@js1019 does the above sound OK?
Many thanks, yes. You may choose a relatively large model (a Moon model with 100k elements as the input) and change https://github.com/js1019/NormalModes/blob/master/demos/global_conf, basename = Mtopo_6L_100k.1 inputdir = models/input/Mtopo100k/ outputdir = models/output/Mtopo100k/ I believe that the strong scaling would be clear.
@js1019 thanks! we'll work on this then.
Create a new workflow that automates a parameter sweep to measure/visualize the strong scalability properties of the code. We can leverage sweepj2 (as shown in this example) to implement the sweep. The space will have two parameters, num of nodes and num of processes. Given an input dataset, it will run the parameter sweep. As part of this we will also implement post-processing and visualization of performance results.