Dear Dirk Eddelbuettel, Romain Francois and contributors, Dear Rcpp Experts!
It is time to praise you and express my thanks. I had a very annoying problem generating up to a 300 square matrices each of sizes up to 500*500 cells. Computing their numeric cell entries took for ages, even when done using the parallel library with mclapply and up to 40 cores. In 2 to 3 % of the cases memory usage skyrocketed to 200GB and zombie processes were hanging around. I could never find the cause for that problem, because it only occurred when my program ran in background (invoked with Rscript), never when the same data and code was executed in an interactive R shell (invoked via R). Anyway, I decided to tackle that particular monster in Rcpp, even though I haven't used C/C++ in more than a decade and even though time has been and still is extremely pressing. - I have to hand in my PhD thesis in a couple of weeks… It took me only 3 days and a couple of stupid newby emails to this list, and now, even though I could not get the Rcpp code to be executed in parallel with OpenMP, I have a running version. And lo and behold: It is 830 times faster than the original R code. And still 150 times faster than the pure R implementation when executed on 10 cores. A framework like Rcpp is truly excellent, when you can get important work done in a couple of days without prior knowledge of even the programming language. Well, mine was very dusty. And Rcpp is so incredibly efficient, too. Guys: Thank you very much for providing Rcpp! I love it and you for having created it. In the end it simply saved my PhD-ing butt. Cheers and a nice rest of the weekend! _______________________________________________ Rcpp-devel mailing list [email protected] https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/rcpp-devel
