Profiling bioconductor packages in a straightforward fashion

@33ed5c35

Last seen 2 hours ago

Denmark

Hi there,

I’m trying to run a rather involved analysis with several compiled Bioconductor packages in an HPC environment, using OpenMPI and R’s doMPI package on my university’s ~3000 node cluster. Mostly my workload parallelises trivially, but opls from the ropls package is slow.

Please forgive me for being naïve, but what’s the best way to profile it? Running it locally I just get “(sources are not available)” in RStudio (with profvis) and although I can wget the tarball of sources and directly look at them, the nice/helpful trace function in RStudio, its own debugging tools, or the profiler don’t appear to work.

Am I missing something? Ropls itself has the “heavy numerics” at a lower level – I should be able to follow the flames down?

Thanks for your tips!