I am having a matrix error when using the computer cluster at my university that I cannot reproduce on my local machine. I think it might be due to a difference of matrix libraries (BLAS, LAPACK, ATLAS, etc.). I don't know much about these libraries other than what I've read here, but I'm thinking there should be a way to get R to tell me which matrix libraries it was installed with (i.e. which ones it's using), analogous to sessionInfo() to tell me which version of R packages it's using.
Though there may not be an explicit R function for this, perhaps you can capitalize on shell commands (e.g., file
and ldd
) to get some clues without requiring rebuilding R nor root access on the cluster:
(rpath <- Sys.which("R"))
# R
# "/usr/bin/R"
To make sure that ldd
will work, see what type of file it is:
system2("file", rpath)
# /usr/bin/R: Bourne-Again shell script, ASCII text executable
If yours shows an actual executable (such as ELF 64-bit LSB executable, x86-64, ...
), then skip this one step.
script <- readLines(rpath)
script[grepl("/bin/", script)]
# [1] "#!/bin/bash"
# [2] " if [ -x \"/usr/${libnn}/R/bin/exec/R\" ]; then"
# [3] " elif [ -x \"/usr/${libnn_fallback}/R/bin/exec/R\" ]; then"
# [4] "## some systems have a more portable sed, e.g. /usr/xpg4/bin/sed on Solaris,"
# [5] "SED=/bin/sed"
# [6] " exec sh \"${R_HOME}/bin/Rcmd\" \"${@}\" ;;"
# [7] "R_binary=\"${R_HOME}/bin/exec${R_ARCH}/R\""
This tells me that the actual executable is /usr/lib/R/bin/exec/R
(or /usr/lib64/...
). It is taking some inference, but it's a step. This is working for me with R-3.3.2 on ubuntu, so I can only assume it'll be similar on different OSes. If this is uninformative, you can also grep for "/lib"
or "/exec"
(or just examine the whole script file for other clues).
Once you know the path called by the script (or if it isn't a script to begin with), find the shared library dependencies:
system2("ldd", "/usr/lib/R/bin/exec/R")
# linux-vdso.so.1 => (0x00007ffcfadcd000)
# libR.so => /usr/lib/R/lib/libR.so (0x00007f746f615000)
# libpthread.so.0 => /lib/x86_64-linux-gnu/libpthread.so.0 (0x00007f746f3eb000)
# libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007f746f025000)
# libblas.so.3 => /usr/lib/libblas.so.3 (0x00007f746eda8000)
# libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007f746eaa2000)
# libreadline.so.6 => /lib/x86_64-linux-gnu/libreadline.so.6 (0x00007f746e85b000)
# libpcre.so.3 => /lib/x86_64-linux-gnu/libpcre.so.3 (0x00007f746e61d000)
# liblzma.so.5 => /lib/x86_64-linux-gnu/liblzma.so.5 (0x00007f746e3fb000)
# libbz2.so.1.0 => /lib/x86_64-linux-gnu/libbz2.so.1.0 (0x00007f746e1ea000)
# libz.so.1 => /lib/x86_64-linux-gnu/libz.so.1 (0x00007f746dfd1000)
# librt.so.1 => /lib/x86_64-linux-gnu/librt.so.1 (0x00007f746ddc9000)
# libdl.so.2 => /lib/x86_64-linux-gnu/libdl.so.2 (0x00007f746dbc4000)
# libgomp.so.1 => /usr/lib/x86_64-linux-gnu/libgomp.so.1 (0x00007f746d9b5000)
# /lib64/ld-linux-x86-64.so.2 (0x0000560abd5fa000)
# libtinfo.so.5 => /lib/x86_64-linux-gnu/libtinfo.so.5 (0x00007f746d78b000)
The 5th line of this output suggests that R is using the BLAS library, specifically libblas3
.
As far as I am aware there isn’t a specific function (especially not in base R) that will give you the BLAS version. R will most likely use the standard library that comes with your system (unless someone built it from source with a specific BLAS version).
You’d have to do it from the terminal, type
locate libblas.so
It’s probably going to be in /usr/lib/libblas.so
or /usr/lib64/libblas.so
, and the location is probably a symlink. So follow the links
ls -l /usr/lib/libblas.so
If you want to to change the symbolic links to point to a different BLAS library, you can do
update-alternatives --config libblas.so.3gf
and select the version you want to use. This post explains it very nicely in more detail: https://www.r-bloggers.com/for-faster-r-use-openblas-instead-better-than-atlas-trivial-to-switch-to-on-ubuntu/
Another way to definitely know what version you are using, would be to install whatever implementation of BLAS you want – for example OpenBLAS, and then replace the standard R BLAS library with a symlink to your new library. Make sure to install R with --with-shared-blas
option, as explained in here: https://cran.r-project.org/doc/manuals/r-release/R-admin.html#BLAS
Then you just need to:
# navigate to your R directory
cd …/path-to-R-installation/lib/R/lib
# backup the original
mv libRblas.so libRblas.backup
# create simlink
ln –s …/path-to-new-0blas-library/new-lib.so libRblas.so
If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!
Donate Us With