[R-sig-hpc] Choosing Intel MKL BLAS when compiling R

[Jonathan Greenberg]

Sorry, small update.  I had this problem before when dealing with
Intel MKL, but I need to revisit it now -- I got the linking working
via alternatives, but this does not seem to work:

(from alternatives:)
Current `best' version is /usr/local/intel-11.1/mkl/lib/em64t/libmkl_core.so

Booting up R gives me this error:
Error in dyn.load(file, DLLpath = DLLpath, ...) :
  unable to load shared object
'/usr/apps/oa/lib64/R/library/stats/libs/stats.so':
  /usr/apps/oa/lib64/R/library/stats/libs/stats.so: undefined symbol: dswap_
During startup - Warning message:
package 'stats' in options("defaultPackages") was not found

I had this problem before -- is libmkl_core.so not the right file to
be linking against?  If so, which one should I be using?  Here's a
dump of the em64t folder:

libmkl_avx.so                   libmkl_gf_lp64.a         libmkl_pgi_thread.so
libmkl_blacs_ilp64.a            libmkl_gf_lp64.so
libmkl_scalapack_ilp64.a
libmkl_blacs_intelmpi_ilp64.a   libmkl_gnu_thread.a
libmkl_scalapack_ilp64.so
libmkl_blacs_intelmpi_ilp64.so  libmkl_gnu_thread.so     libmkl_scalapack_lp64.a
libmkl_blacs_intelmpi_lp64.a    libmkl_intel_ilp64.a
libmkl_scalapack_lp64.so
libmkl_blacs_intelmpi_lp64.so   libmkl_intel_ilp64.so    libmkl_sequential.a
libmkl_blacs_lp64.a             libmkl_intel_lp64.a      libmkl_sequential.so
libmkl_blacs_openmpi_ilp64.a    libmkl_intel_lp64.so     libmkl_solver_ilp64.a
libmkl_blacs_openmpi_lp64.a     libmkl_intel_sp2dp.a
libmkl_solver_ilp64_sequential.a
libmkl_blacs_sgimpt_ilp64.a     libmkl_intel_sp2dp.so    libmkl_solver_lp64.a
libmkl_blacs_sgimpt_lp64.a      libmkl_intel_thread.a
libmkl_solver_lp64_sequential.a
libmkl_blas95_ilp64.a           libmkl_intel_thread.so   libmkl_vml_avx.so
libmkl_blas95_lp64.a            libmkl_lapack.so         libmkl_vml_def.so
libmkl_cdft_core.a              libmkl_lapack95_ilp64.a  libmkl_vml_mc.so
libmkl_core.a                   libmkl_lapack95_lp64.a   libmkl_vml_mc2.so
libmkl_core.so                  libmkl_mc.so             libmkl_vml_mc3.so
libmkl_def.so                   libmkl_mc3.so            libmkl_vml_p4n.so
libmkl_gf_ilp64.a               libmkl_p4n.so            locale
libmkl_gf_ilp64.so              libmkl_pgi_thread.a

--j


On Tue, Nov 1, 2011 at 3:40 PM, Jonathan Greenberg <jgrn at illinois.edu> wrote:
> Dirk:
>
> Thanks -- I was able to get this running on a Debian box previously
> (http://lostingeospace.blogspot.com/2010/07/multithreaded-r-using-blas-in-debian_28.html)
> but I'm having trouble figuring out specific steps I need to do to
> choose the Intel MKL BLAS, particularly now that I don't have a
> package manager to rely on -- is it an issue of using
> update-alternatives?  How can I check to see what R is currently
> using?
>
> --j
>
> On Tue, Nov 1, 2011 at 3:12 PM, Dirk Eddelbuettel <edd at debian.org> wrote:
>>
>> On 1 November 2011 at 14:36, Jonathan Greenberg wrote:
>> | r-sig-hpc'ers:
>> |
>> | I am compiling R 2.14 from scratch on a cluster computer (I can't use
>> | any package managers for the install, unfortunately), and they have
>> | the Intel MKL libraries available for BLAS/LAPACK.  I have not
>> | successfully gotten a configure/make/make install combo yet that uses
>> | multiple CPUs to do, say:
>> |
>> | a = matrix(rnorm(5000*5000), 5000, 5000)
>> | b = matrix(rnorm(5000*5000), 5000, 5000)
>> | # This following line, minimally, should be getting distributed.
>> | c = a%*%b
>> |
>> | First off, I've tried compiling it with the configure flags:
>> | ./configure --prefix=/myprefix/ --enable-BLAS-shlib
>> |
>> | How do I then select the Intel MKL libraries to be used as the BLAS?
>> | They are located in /usr/local/intel-11.1/mkl/lib/em64t
>>
>> BLAS is an interface so you do not have to configure for a given BLAS. You
>> just build for the generic BLAS, and the drop better ones in.
>>
>> This is demonstrated in a little package / vignette / paper I have at
>>
>>   https://r-forge.r-project.org/scm/viewvc.php/pkg/gcbd/?root=gcb
>>
>> and you may want to hit the pdf directly at
>>
>>   https://r-forge.r-project.org/scm/viewvc.php/pkg/gcbd/inst/doc/gcbd.pdf?root=gcb&view=log
>>
>> This uses the package manager on Debian/Ubuntu to drop the basic Blas, Atlas,
>> Goto and MKL in and out in a fully automatable (and hence scripted) manner so
>> that benchmarks can be run.   Besides the fact that is uses the dpkg package
>> manager --- and of course the fact that the respective package maintainer
>> have done the hard work of conforming to the Debian standard for pluggable
>> Blas --- nothing is really distro specific.  So you could use this as
>> 'proof-of-concept' and try to make your cluster machines conform.
>>
>> In essence, this means trying to configure R to any external Blas and then
>> plugging the MKL in.
>>
>> Some other details are in the 'R Installation and Administration' manual.
>>
>> Good luck,  Dirk
>>
>> --
>> "Outside of a dog, a book is a man's best friend. Inside of a dog, it is too
>> dark to read." -- Groucho Marx
>>
>
>
>
> --
> Jonathan A. Greenberg, PhD
> Assistant Professor
> Department of Geography
> University of Illinois at Urbana-Champaign
> 607 South Mathews Avenue, MC 150
> Urbana, IL 61801
> Phone: 415-763-5476
> AIM: jgrn307, MSN: jgrn307 at hotmail.com, Gchat: jgrn307, Skype: jgrn3007
> http://www.geog.illinois.edu/people/JonathanGreenberg.html
>



-- 
Jonathan A. Greenberg, PhD
Assistant Professor
Department of Geography
University of Illinois at Urbana-Champaign
607 South Mathews Avenue, MC 150
Urbana, IL 61801
Phone: 415-763-5476
AIM: jgrn307, MSN: jgrn307 at hotmail.com, Gchat: jgrn307, Skype: jgrn3007
http://www.geog.illinois.edu/people/JonathanGreenberg.html