Dear Daniel,

a good starting point is: http://cran.at.r-project.org/web/views/HighPerformanceComputing.html

There are also some interesting blogs, eg.

http://blog.revolutionanalytics.com/2012/08/ryan-rosario-on-parallel-programming-in-r.html

http://www.r-bloggers.com/a-brief-foray-into-parallel-processing-with-r/

http://www.r-bloggers.com/parallel-computing-in-r/

http://www.r-bloggers.com/how-to-go-parallel-in-r-basics-tips/

http://www.r-statistics.com/2010/04/parallel-multicore-processing-with-r-on-windows/

For more search for R bloggers parallel

Best

Matthias

Here a small example. The model is from Scheffer, M. et al. (1993)  Alternative equilibria in shallow lakes. Trends in Ecology & Evolution, 8, 275-279.

library(rootSolve)

library(parallel) # requires a recent version of R

equilibrium

Thanks much! This is a great start. My model is currently killing my somewhat prehistoric machine, so looking into expansion options.

Of which type is your model and which packages are used? What we often do is writing parts or the full model in C/C++ or Fortran to make it faster. Depending on the type of model you can gain factor 2 (if the model is well formulated in matrix and vector form) upto 50..100 if good vectorization is not possible.

For our main model type (initial value problems of differential equations) we often use code generation techniques, i.e. generate R or Fortran code from plain tables.

Thomas

The model mostly propagates matrices... although to be honest, my matrix operations are probably very inefficient, as I'm coming from MATLAB and not so familiar with optimizing R. I find myself using a lot of loops, which I know is frowned upon in R modeling, and slows everything way down.

But, thanks for your suggestions, looking forward to parallelizing on my machine and hopefully expanding to a larger one.

Maybe this could be also of interest here: http://r-pbd.org/

I did not do a formal benchmark, however for some reason its much much slower in linux then windows ( I guesstimate 5x). Weird right? would've expected windows to be slower but its the reverse and I've tried this on multiple system as well, same results.

This is really weird, but we cannot guess why without a minimum reproducible example. After my experience, R's speed is more or less the same on all OSes, maybe a bit faster on Linux. The only exception that I remember was a simulation on a virtual Linux where I forgot to reserve enough RAM for the VM.