[R-sig-hpc] Choosing Intel MKL BLAS when compiling R
jgrn at illinois.edu
Wed Nov 2 19:58:37 CET 2011
So if I'm following all this correctly, Intel MKL BLAS cannot be used
as a "swap-in" for R like GotoBLAS or Atlas, but it CAN be used during
the compilation phase as the hard-coded BLAS used?
Incidentally, I switched over to OpenBLAS from GotoBLAS last year and
have been happy with it -- it actually solved some compilation issues
that GotoBLAS had (also, it appears the configuration phase is a lot
On Wed, Nov 2, 2011 at 9:06 AM, Ei-ji Nakama <nakama at ki.rim.or.jp> wrote:
>> o configure looks for support for OpenMP and if found compiles R
>> with appropriate flags and also makes them available for use in
>> packages: see 'Writing R Extensions'.
> When people make by default option in GotoBLAS for Linux,
> it raise a thread at the time of the initialization of the library
> and set affinity.
> Therefore openmp will not cause a thread.
> The good method should control the number of the threads in
> omp_set_num_threads and goto_set_num_threads.
> !!!must perform this!!!
> performance when there is a part to use the thread for elsewhere is the worst
> MKL thinks that KMP_AFFINITY=verbose helps analysis.
> and ... people remove the number of the CPU's of the hyper thread in
> BLAS and should carry it out.
> # BLAS may count the number of the cores precisely partly.
> I wrote regret it in strange English.XD
> Best Regards,
> EI-JI Nakama <nakama (a) ki.rim.or.jp>
> "\u4e2d\u9593\u6804\u6cbb" <nakama (a) ki.rim.or.jp>
> R-sig-hpc mailing list
> R-sig-hpc at r-project.org
Jonathan A. Greenberg, PhD
Department of Geography
University of Illinois at Urbana-Champaign
607 South Mathews Avenue, MC 150
Urbana, IL 61801
AIM: jgrn307, MSN: jgrn307 at hotmail.com, Gchat: jgrn307, Skype: jgrn3007
More information about the R-sig-hpc