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

[Dirk Eddelbuettel]

On 1 November 2011 at 16:25, Jonathan Greenberg wrote:
| Ok, so I did a new test and installed a local version of GotoBLAS
| (OpenBLAS fork), used alternatives --install, and its working
| perfectly.  According to your paper, I'm probably good to go (you got
| the best results from GotoBLAS, right?)  However, just to see if I can

Correct, though that was a) a little over a year ago and b) it used the older
MKL "we all" once got access to because Revo (then called REvolution
Computing) had it integrated into Ubuntu 9.10 (Disclaimer: I was on a
consultant on that project.)

On the other hand, with R 2.14.0 you can no longer use MKL anyway it
conflicts (via R's use of OpenMP) with the MKL build (as they both provide
conflicting MP primitives, if memory serves).  So my money is still on Goto,
and hopefully one day on the OpenBlas replacement.

| settle the issue (I have colleagues who would prefer to use Intel
| MKL's BLAS), can you advise on how to properly link against the Intel
| MKL libraries based on the previous error I reported?  What is the
| specific file I should be pointing R towards?

I suggest yoru colleagues request help from the vendor they paid dearly for
this.  I am after all just a volunteer and not really in charge of your
cluster (nor do I have the time to dive into this..., sorry!)

Dirk

| 
| ***
| 
| 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
| 
| ***
| 
| --j
| 
| 
| On Tue, Nov 1, 2011 at 4:09 PM, Dirk Eddelbuettel <edd at debian.org> wrote:
| >
| > On 1 November 2011 at 15:40, Jonathan Greenberg 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?
| >
| > By examining the Makefile(s) created after configure, as well as the
| > configure logs.  "My" builds for Debian end the configure step in
| >
| > R is now configured for i486-pc-linux-gnu
| >
| >  Source directory:          .
| >  Installation directory:    /usr
| >
| >  C compiler:                gcc -std=gnu99  -O3 -pipe  -g
| >  Fortran 77 compiler:       gfortran  -O3 -pipe  -g
| >
| >  C++ compiler:              g++  -O3 -pipe  -g
| >  Fortran 90/95 compiler:    gfortran -O3 -pipe  -g
| >  Obj-C compiler:
| >
| >  Interfaces supported:      X11, tcltk
| >  External libraries:        readline, BLAS(generic), LAPACK(generic), lzma
| >  Additional capabilities:   PNG, JPEG, TIFF, NLS, cairo
| >  Options enabled:           shared R library, R profiling, memory profiling, Java
| >
| >  Recommended packages:      no
| >
| >
| > which shows you that external Blas and Lapack are used -- that's needed for
| > the drop-in behaviour which you then deal with at the dynamoc linker stage
| > (eg with LD_PRELOAD etc).
| >
| > 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

-- 
"Outside of a dog, a book is a man's best friend. Inside of a dog, it is too
dark to read." -- Groucho Marx