I've seen 2-10 times figure mentioned as a reasonable expectation for this. (I hit around 5-6x on my last few projects.)
Avoid calling back into R (i.e., Rcpp::Function) too frequently, because crossing that interpreter boundary carries a real price. Also, if your R code is mostly staying inside its C library (e.g., matrix ops), rewriting it in Armadillo won't make much difference since it only speeds up a very thin outside layer. Davor On 2010-06-15, at 11:24 AM, Chuck Anderson wrote: > I translated some R code I have for training neural networks into C++ using > Armadillo and call it using RcppArmadillo. It was great fun to see how > Armadillo allows a pretty direct translation from R to C++. However, I only > see a speed up of 2 over the R execution time. I expected much more. > What speed up do others see? Any suggestions for finding execution > bottlenecks in my Armadillo code? > Thanks, > Chuck > _______________________________________________ > Rcpp-devel mailing list > [email protected] > https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/rcpp-devel _______________________________________________ Rcpp-devel mailing list [email protected] https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/rcpp-devel
