Package Details: r-mkl 3.6.1-1

Git Clone URL: https://aur.archlinux.org/r-mkl.git (read-only, click to copy)
Package Base: r-mkl
Description: Language and environment for statistical computing and graphics, linked to the Intel(R) MKL.
Upstream URL: http://www.r-project.org/
Keywords: hpc mathematics modelling r statistics
Licenses: GPL
Conflicts: microsoft-r-open, r
Provides: r=3.6.1
Submitter: giniu
Maintainer: alexanderp
Last Packager: alexanderp
Votes: 19
Popularity: 0.46
First Submitted: 2010-05-06 00:10
Last Updated: 2019-07-11 20:06

Dependencies (26)

Required by (264)

Sources (4)

Latest Comments

« First ‹ Previous 1 2 3 4 5 6 7 ... Next › Last »

petronny commented on 2019-07-23 07:24

And is building with gcc the recommend way by the official R group to build with MKL support?
If so I can give up building with icc and using gcc now.

petronny commented on 2019-07-23 07:21

Hi, thanks for maintaining this package.
Could you help with building with icc?
I've un-comment the line _CC="icc" but I get errors:

icc -I../../src/extra -I/usr/include/tirpc -I. -I../../src/include -I../../src/include  -D_FORTIFY_SOURCE=2 -I../../src/nmath -DHAVE_CONFIG_H   -fopenmp -fpic  -O3 -fPIC -m64 -march=native -fp-model precise -fp-model source -I/opt/intel/compilers_and_libraries_2019.4.243/li
nux/mkl/include  -c envir.c -o envir.o
arithmetic.c(59): warning #274: declaration is not visible outside of function
  int matherr(struct exception *exc)
                     ^

arithmetic.c(61): error: pointer to incomplete class type is not allowed
      switch (exc->type) {
              ^

arithmetic.c(62): error: identifier "DOMAIN" is undefined
      case DOMAIN:
           ^

arithmetic.c(63): error: identifier "SING" is undefined
      case SING:
           ^

arithmetic.c(66): error: identifier "OVERFLOW" is undefined
      case OVERFLOW:
           ^

arithmetic.c(69): error: identifier "UNDERFLOW" is undefined
      case UNDERFLOW:
           ^

arithmetic.c(70): error: pointer to incomplete class type is not allowed
        exc->retval = 0.0;
        ^

icc -I../../src/extra -I/usr/include/tirpc -I. -I../../src/include -I../../src/include  -D_FORTIFY_SOURCE=2 -I../../src/nmath -DHAVE_CONFIG_H   -fopenmp -fpic  -O3 -fPIC -m64 -march=native -fp-model precise -fp-model source -I/opt/intel/compilers_and_libraries_2019.4.243/li
nux/mkl/include  -c errors.c -o errors.o
icc -I../../src/extra -I/usr/include/tirpc -I. -I../../src/include -I../../src/include  -D_FORTIFY_SOURCE=2 -I../../src/nmath -DHAVE_CONFIG_H   -fopenmp -fpic  -O3 -fPIC -m64 -march=native -fp-model precise -fp-model source -I/opt/intel/compilers_and_libraries_2019.4.243/li
nux/mkl/include  -c eval.c -o eval.o
icc -I../../src/extra -I/usr/include/tirpc -I. -I../../src/include -I../../src/include  -D_FORTIFY_SOURCE=2 -I../../src/nmath -DHAVE_CONFIG_H   -fopenmp -fpic  -O3 -fPIC -m64 -march=native -fp-model precise -fp-model source -I/opt/intel/compilers_and_libraries_2019.4.243/li
nux/mkl/include  -c format.c -o format.o
compilation aborted for arithmetic.c (code 2)
make[3]: *** [../../Makeconf:124: arithmetic.o] Error 2
make[3]: *** Waiting for unfinished jobs....
make[3]: Leaving directory '/build/r-mkl/src/R-3.6.1/src/main'
make[2]: *** [Makefile:135: R] Error 2
make[2]: Leaving directory '/build/r-mkl/src/R-3.6.1/src/main'
make[1]: *** [Makefile:28: R] Error 1
make[1]: Leaving directory '/build/r-mkl/src/R-3.6.1/src'
make: *** [Makefile:61: R] Error 1
==> ERROR: A failure occurred in build().

blazko commented on 2019-07-15 07:40

Is anyone else facing validity checks problem with parallel_studio_xe?

==> Validating source files with sha256sums...
    parallel_studio_xe_2019_update4_cluster_edition.tgz ... FAILED
    intel_compilers.sh ... Passed
    intel_vtune-amplifier.sh ... Passed
    intel_advisor.sh ... Passed
    intel_inspector.sh ... Passed
    intel-composer.install ... Passed
    intel-common-libs.conf ... Passed
    intel-fortran.conf ... Passed
    intel-openmp.conf ... Passed
    intel-mkl.conf ... Passed
    intel-mpi.conf ... Passed
    intel-ipp.conf ... Passed
    intel-tbb.conf ... Passed
    intel-mkl.sh ... Passed
    intel-mkl.install ... Passed
    intel-mkl-th.conf ... Passed
    intel-tbb.install ... Passed
    EULA.txt ... Passed
==> ERROR: One or more files did not pass the validity check!

alexanderp commented on 2019-05-17 11:46

Thanks for the report @blazko. To be honest, those numbers look strange to me... Did you consider reporting this to the Intel MKL forums? (https://software.intel.com/en-us/forums/intel-math-kernel-library)

By the way, have you done any benchmark comparisons between the versions?

blazko commented on 2019-05-17 11:40

@alexanderp I sorted this out. For a 4 cores 8 threads CPU I should set export MKL_NUM_THREADS=32 on my laptop (instead of 8) and export MKL_NUM_THREADS=128 on the xeon (instead of 32),

After this they all work at max speeds and resources.

alexanderp commented on 2019-05-07 10:59

@blazko, Not sure... It seems that the way GNU OpenMP utilizes cores/threads is different. Just for reference, I am getting the same behaviour when using the microsoft-r-open package.

blazko commented on 2019-05-07 08:49

@alexanderp thanks for your tests - i'm running a bigger batch of those sims right now, and I would avoid stopping them.

Your test clearly shows, that performancewise OpenMP is better. Is there a reason we can't run that on 100%?

alexanderp commented on 2019-05-06 20:15

@blazko, I modified your code as such:

library(simr)
dataf <- read.csv("Rdataframe.csv", stringsAsFactors = F)
ns = 500
ss<-1
m1 <- lmer(scc ~ naC + (naC|subj), dataf)
summary(m1)
fixef(m1)["naC"] <- -0.15
set.seed(ss)
system.time(
  p1 <- powerCurve(m1, along="subj", nsim=100, breaks = seq(20, 100, 20))
)

These were the results:

R 3.5.3 with Intel threading:

  • all cores 100% in top
  • user 1015.352, system 73.983, elapsed 206.349

R 3.6.0 with GNU OpenMP:

  • half of the cores active in top
  • user 245.528, system 3.107, elapsed 181.824

Please run this on your side as well and report back.

alexanderp commented on 2019-05-06 19:57

@mys_721tx, I've never had to source mklvars.sh manually, since it's loaded by profile.d on login.

mys_721tx commented on 2019-05-06 19:36

@alexanderpLD_LIBRARY_PATH and LIBRARY_PATH are not captured in Makeconf. Therefore /opt/intel/mkl/bin/mklvars.sh must be sourced before building any package that uses limf.