Hi all,
surfing the web I wasn't able to find a simple and effective answer to my question. In other words, I need to speed up a code I wrote using several packages such as raster, rgdal, biomod2 and ncd4 on GIS data: do I need to wait somebody else (or me too) to develop a NEW ratser etc. package which forces R to use GCPU or is it already possible siply loading additional packages before my code run such as gpuR, parallel or what else?
All my bests
Dear Maurizio,
Although I find your question really interesting and important and hope that somebody will give a more positive answer to it, I'm afraid it's not possible to let all packages use multiple CPU cores and GPU as well with a simple package loading. A workaround instead of solution: Microsoft R Open provides built-in multithreaded matrix algebra which is used by a lot of packages. You may find that this software is a useful alternative to R.
https://mran.microsoft.com/rro#intelmkl1
HTH,
Ákos
Parallel processing does not make always sense and there are many cases where it would be silly if not detrimental. It would thus not make sense to have ALL packages using parallel processing. Packages that can make efficient use of parallel computing do implement it.
And then there are problems that could be parallelized in principle, but ther is rarely a solution (of how to do this) that would generally improve the net-performance. It depends on too many things, not all being known to the package developer.
At the moment, it seems that on Ubuntu it is possible to open R environment using the graphic processor
https://starbeamrainbowlabs.com/blog/article.php?article=posts%2F254-run-program-on-amd-dedicated-graphics-card.html
Does anyone have any update on that?
- Badges
- Science topic
- Similar topics
- Computer Science