[R-SIG-Mac] single-threaded R, 100% CPU with BLAS, vecLib and ATLAS

[Simon Urbanek]

On Oct 25, 2012, at 7:42 PM, Ray DiGiacomo, Jr. wrote:

> Hello Melanie,
> 
> How much RAM is installed on your MacBook Pro compared to your colleague's
> Linux machine?
> 
> How big is your dataset in terms of rows and columns?
> 
> I believe R can handle about 10M datapoints per GB of RAM.
> 

What exactly is that an estimate of? In R, 1GB of RAM will store ~134Mio datapoints when using numeric matrices/vectors and twice as many as integers or logicals. In practice, you will still need some room for computation on the data, though. 

Cheers,
Simon


> Note that datapoints = rows x columns
> 
> Best Regards,
> 
> Ray DiGiacomo, Jr.
> Master R Trainer
> President, Lion Data Systems LLC
> President, The Orange County R User Group
> Board Member, TDWI
> rayd at liondatasystems.com
> (Mobile) 408-425-7851
> San Juan Capistrano, California
> 
> Check out my one-on-one web-based R courses at liondatasystems.com/courses
> 
> 
> 
> 
> 
> On Thu, Oct 25, 2012 at 4:16 PM, Melanie Courtot <mcourtot at gmail.com> wrote:
> 
>> Hi,
>> 
>> I am trying to run R on my MacBook Pro 2.4 GHz Intel core i5. I am trying
>> to read a csv file, which works fine on my work colleague's machine (under
>> linux) but causes my CPU to go up to 100% and makes the GUI unresponsive
>> and hangs on the command line. Activity monitor indicates there is only one
>> R thread running.
>> 
>> I did see that by default R was using the BLAS library, which is
>> single-threaded, and that there was an option to use vecLib instead. I did
>> this, and
>> ls -l /Library/Frameworks/R.framework/Resources/lib/libRblas.dylib
>> does return
>> /Library/Frameworks/R.framework/Resources/lib/libRblas.dylib ->
>> libRblas.vecLib.dylib
>> 
>> I however still see the same behavior: 100% CPU, single thread.
>> 
>> I saw that some MacBook pro (Xeon Nehalem based) had a vecLib bug, so I
>> built the ATLAS library and symlinked R to libtatlas.dylib (unfortunately
>> the pre compiled binairies pointed to in a previous email on the list [1]
>> were not available anymore. Building ATLAS was... fun ;)) I was able to get
>> the shared libraries (using --shared in my config) but still see the same
>> behavior when trying to run my code. I was unsure if I should link to
>> libsatlas.dylib or libtatlas.dylib, so tried both (I guess the latter was
>> the right one though)
>> 
>> I tried building R from the source (specifying -arch x86_64 and
>> --enable-BLAS-shlib to be able to switch libraries), but same behavior and
>> it seems it is an identical version to the prepackaged one (I tried with
>> BLAS, vecLib and ATLAS)
>> 
>> R info: R version 2.15.1 (2012-06-22) -- "Roasted Marshmallows", Platform:
>> x86_64-apple-darwin9.8.0/x86_64 (64-bit)
>> 
>> Any help would be greatly appreciated.
>> 
>> Thanks,
>> Melanie
>> 
>> 
>> [1] https://stat.ethz.ch/pipermail/r-sig-mac/2010-October/007817.html
>> 
>> ---
>> Mélanie Courtot
>> MSFHR/PCIRN Ph.D. Candidate,
>> BCCRC - Terry Fox Laboratory - 12th floor
>> 675 West 10th Avenue
>> Vancouver, BC
>> V5Z 1L3, Canada
>> 
>> _______________________________________________
>> R-SIG-Mac mailing list
>> R-SIG-Mac at r-project.org
>> https://stat.ethz.ch/mailman/listinfo/r-sig-mac
>> 
> 
> 	[[alternative HTML version deleted]]
> 
> _______________________________________________
> R-SIG-Mac mailing list
> R-SIG-Mac at r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-mac