dear experts I'd like to ask you, probably on behalf all of us who try to use AMD's ACML for performance optimisation a question. There seem to be problem when these two, R and ACML are coupled, described here: http://devgurus.amd.com/message/1255852#1255852 I'd like to ask, while not being a programmer/developer, how incorrect and dangerous this might be? maybe such a setup must be avoided all together? many thanks & regards
acml blas+lapac in R
3 messages · lejeczek, Brian Ripley
Note that this is in the 'R Installation and Administration Manual'. Please follow the posting guide and check the manuals before posting.
On 25/02/2014 08:42, lejeczek wrote:
dear experts I'd like to ask you, probably on behalf all of us who try to use AMD's ACML for performance optimisation a question. There seem to be problem when these two, R and ACML are coupled, described here: http://devgurus.amd.com/message/1255852#1255852 I'd like to ask, while not being a programmer/developer, how incorrect and dangerous this might be? maybe such a setup must be avoided all together?
As an R developer is reputed to have said several years ago 'Do you want it fast or do you want it right?' Handing NAs correctly is pretty fundamental to R's operation. There are several other optimized BLAS which are not known to work incorrectly, and I always use one of those.
many thanks & regards
______________________________________________ R-devel at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
Brian D. Ripley, ripley at stats.ox.ac.uk Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/ University of Oxford, Tel: +44 1865 272861 (self) 1 South Parks Road, +44 1865 272866 (PA) Oxford OX1 3TG, UK Fax: +44 1865 272595
thank you Brian if ACML is the problem I wonder if it ever was pointed out to AMD shame because ACML results in really nice performance on AMD hardware
On 25/02/14 10:39, Prof Brian Ripley wrote:
Note that this is in the 'R Installation and Administration Manual'. Please follow the posting guide and check the manuals before posting. On 25/02/2014 08:42, lejeczek wrote:
dear experts I'd like to ask you, probably on behalf all of us who try to use AMD's ACML for performance optimisation a question. There seem to be problem when these two, R and ACML are coupled, described here: http://devgurus.amd.com/message/1255852#1255852 I'd like to ask, while not being a programmer/developer, how incorrect and dangerous this might be? maybe such a setup must be avoided all together?
As an R developer is reputed to have said several years ago 'Do you want it fast or do you want it right?' Handing NAs correctly is pretty fundamental to R's operation. There are several other optimized BLAS which are not known to work incorrectly, and I always use one of those.
many thanks & regards
______________________________________________ R-devel at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel