Hello everyone,
I've implemented a FEM code in matlab, and now I'd like to make it faster. Since I have to perform nonlinear dynamic analysis, I have a lot of iterations, each one requiring to asseble the full stiffness matrix K.
What I would like to do thus is to parallelize the assembling procedure of K. Till now I tried parpool and spmd, the latter with poor results, the first one performing nicely (speedup factor x2... despite using 10 cores...) but only under a certain number of elements. Overcome a certain treshold, parallel computation (14 cores) would take as much as 10 times the single core version.
I understand this may be related to overheads in the comunication between "master" and workers and/or slicing procedures, but it seems I cannot get the hang of it...
Does anyone have suggestions and/or can point me to some useful material on this specific matter?
Thank you all in advance,
Jacopo
Normally it takes a lot of hard work and hassel to implement FEM modelling in Matlab and it is expected to get poor result. However, you may put your problem in Matlab forum and seek for a solution. Alternatively try to find few journals which are related to ykur work involving the similar ttpe of modelling and coding and send a mail to that respective author asking for their advice or suggestion. For doing that introduce yourself clearly in rhe mail giving your academic and research position and if possible from a dedicated email from your organisation. Hope the suggestion is useful.
I don't know much about your specific problem but I think you're actually populating a matrix using concurrent parfor iterations. This doesn't work so well as it causes a bottleneck in memory access, your sensitivity analisys about the number of cores confirms this hypothesis. What you actually need is vectorizing (vectorization) of your code. Try posting your code here or sending me an email (my name.surname at unirc.it, no middle name) I cannot guarantee any result but I'll try giving you some hints.
- Badges
- Science method