Comments on: Compile R and OpenBLAS from Source Guide http://www.lindonslog.com/linux-unix/compile-r-openblas-source-guide/ Sun, 27 Nov 2016 15:23:00 +0000 hourly 1 https://wordpress.org/?v=4.6.1 By: admin http://www.lindonslog.com/linux-unix/compile-r-openblas-source-guide/#comment-682 Wed, 22 Oct 2014 16:29:13 +0000 http://www.lindonslog.com/?p=995#comment-682 OpenBLAS is an optimized BLAS library based on GotoBLAS2, they are not the same. Go with Openblas, I’m not sure if gotoblas is being maintained anymore. Parallel linear algebra libraries are doing different things than multicore, snow and parallel <- I think these packages are just used to divide and share a large number of identical jobs between processors. There's no real benefit if all you are using are lm, its really for people who are trying to handle big data and do mcmc in R.

]]> By: isomorphismes http://www.lindonslog.com/linux-unix/compile-r-openblas-source-guide/#comment-681 Tue, 21 Oct 2014 22:55:45 +0000 http://www.lindonslog.com/?p=995#comment-681 Thanks for posting this. Could you talk more to the tradeoffs in choosing ATLAS / openBLAS (which I think goto is the same as open) versus require( multicore | snow | parallel ) ?

In particular I’m not sure how much linear algebra core R routines like plot use. Clearly lm and straight-up matrix decompositions use linear algebra but in fact I often use MASS::rlm instead of lm.

]]> By: Joe Liebig (@JochenLiebig) http://www.lindonslog.com/linux-unix/compile-r-openblas-source-guide/#comment-367 Thu, 17 Jul 2014 13:46:41 +0000 http://www.lindonslog.com/?p=995#comment-367 This is really interesting. I’m lacking the time to reproduce and dig deeper, but at the same time I am wondering whether this could help speed up R on Parallella.

]]>