yade-dev team mailing list archive

[Václav Šmilauer <eu@xxxxxxxx>]

> > Sorry, I live in relations of my 2 cores laptop :-) but still overall speedup for 8 cores is 5.5x (page 66 of Vaclav's thesis), so this, depending on engine, is not "of no use" yet, at least from my point of view.. but yes, parallel 4+4 at the same time would be better than 8 and then 8..
> The reported speedup is what was measured on a cluster with a simulation
> of 10k particles. It might vary quite a lot depending on the actual
> simulation, and I suppose also depending on what constitutive law one
> uses: those that access scene->bodies are very likely to suffer when run
> in parallel.
> 
> Perhaps it is because I have very little experience, but estimating
> speedup of parallel code is crystal gazing for me.

I was just testing examples/concrete/uniax.py (for unrelated reasons)
and found the following iter/sec (approx) for 1, 2, 3, 4 threads
respectively: 220, 445, 600, 590. For sure what plays dominant role is
how much time is spent solving one interactions, therefore complicated
contact laws scale better than CundallStrack.