Open robinzyb opened 1 year ago
Hi, just skimming through the code, there seem to be two parts that can be accelerated with parallel computing. 1) _selection_serial
2) _getMeanOnePoint
.
For the second part, there's a tutorial about how to convert a per-frame analysis function into a parallel analysis class in (https://userguide.mdanalysis.org/dev/examples/analysis/custom_parallel_analysis.html). You could start from here.
thank you !
Overall the waterdynamics code needs to get refactored to use AnalysisBase. There is a lot that could be improved that we can't do right now because it doesn't adhere to our standard API.
I'd say until someone rewrites waterdynamics to follow the standard API, we will not have the capacity to parallelize the code.
hi, just to confirm. Is this the standard API you mentioned? https://docs.mdanalysis.org/stable/documentation_pages/analysis/base.html#writing-new-analysis-tools I recently got time to rewrite this functionality
Yes, exactly — derive from AnalysisBase.
Hello, the autocorrelation function is required in water orientation analysis. however, I only found a special version of autocorrelation function for survival probability. https://docs.mdanalysis.org/stable/documentation_pages/lib/correlations.html#autocorrelation-function
Is there a general autocorrelation function implemented elsewhere or not?
If not, I can write one for 3-dim vector properties and put it in the MDAnalysis.lib.correlations
Is your feature request related to a problem?
Dear developers, I would like to suggest adding parallel computing for water dynamics analysis. In https://docs.mdanalysis.org/2.4.1/_modules/MDAnalysis/analysis/waterdynamics.html#WaterOrientationalRelaxation. Especially, the water orientational relaxation.
Describe the solution you'd like
I would like to implement by myself if there are some coding conventions/instructions to follow for parallel computing in MDA
Describe alternatives you've considered
Additional context