Giulio Fraulini wrote:
My question to Charles outside the main argument: you've said "LDA very costly in terms of time (and doesn't use all the potential CPU)". It seems to me that my workstation utilizes all the cpu during LDA.
Are you sure (~100% CPU Usage)? Or is it just sharing one thread (13% or 25%) across all the cores?
A good measure in Mechanica is to compare Total CPU Time with Total Elapsed Time. On the run I've just tried, CPU Time is ~ 2× Elapsed Time - so it was able to use (all) 4 cores for some of the run.
"Why did you include 'snap through' in two of your analysis?"
Is a forgetfulness of a previous analysis. Anyway I don't think it has any effect on an analysis where there are only enforced displacements.
The main influence is to slow down the convergence - on my system it reduced from ~7000 sec CPU time, to ~1000 after disabling snap-through detection!