[Rd] Performance of .C and .Call functions vs. native R code

[Alireza Mahani]

Prof. Bates,

It looks like you read my mind! I am working on writing an R package for
high-performance MCMC estimation of a class of Hierarchical Bayesian models
most often used in the field of quantitative marketing. This would
essentially be a parallelized version of Peter Rossi's bayesm package. While
I've made great progress in parallelizing the most mathematically difficult
part of the algorithm, namely slice sampling of low-level coefficients, yet
I've realized that putting the entire code together while minimizing bugs is
a big challenge in C/C++/CUDA environments. I have therefore decided to
follow a more logical path of first developing the code logic in R, and then
exporting it function by function to compiled code. 

The tools that you mentioned seem to be exactly the kind of stuff I need in
order to be able to do go through this incremental, test-oriented
development process with relatively little pain.

I'm not sure if this is what you had in mind while suggesting the tools to
me, so please let me know if I'm misinterpreting your comments, or if I need
to be aware of other tools beyond what you mentioned.

Many thanks,
Alireza


--
View this message in context: http://r.789695.n4.nabble.com/Performance-of-C-and-Call-functions-vs-native-R-code-tp3665017p3679056.html
Sent from the R devel mailing list archive at Nabble.com.