[R-SIG-Mac] Clarifying the status of OMP in CRAN binaries on Mac

[Simon Urbanek]

Philipp,

the instructions (if really insist on using Xcode 16.3 - not recommended) are here: https://mac.r-project.org/openmp/ relevant part quoted:

"Note: If you are using CRAN R binary and there is a potential conflict between its run-time and your run-time (which is generally a bad idea!) then you can force the use of your binary by replacing -lomp above with /usr/local/lib/libomp.dylib (note that lack of -l!)."

i.e., you set PKG_LIBS=/usr/local/lib/libomp.dylib when compiling data.table.

Note, however that you then have to re-compile *all* packages that may use OpenMP as your result will be incompatible with the CRAN binaries.

Cheers,
Simon


> On May 10, 2025, at 1:10 AM, Philipp Schneider via R-SIG-Mac <r-sig-mac using r-project.org> wrote:
> 
> Hey,
> 
> I'm a bit late to the party, but I was just made aware of this thread in my GitHub thread (https://github.com/Rdatatable/data.table/issues/6949), which is very related to the one mentioned here.
> 
> To make it as short as possible, I am unable to build data.table from source since Xcode 16.3, which expected reading this thread. I thought I was using the OpenMP from LLVM 19.1.0, which should work again with Xcode 16.3. While doing some research I stumbled over an old R-help thread by me that states that R is always using the libomp.dylib bundled with R (/Library/Frameworks/R.framework/Versions/4.x-arm64/Resources/lib/libomp.dylib). Here's the R-help thread just in case: https://stat.ethz.ch/pipermail/r-help/2024-April/479157.html. Turns out I never really solved this issue since other stuff came up, but now this haunts me again :). I always untar the .tar.gz archive from https://mac.r-project.org/openmp/ as mentioned there and then use a ~/.R/Makevars also as mentioned there, but how do I actually tell R to use the libomp.dylib in /usr/local/lib and not /Library/Frameworks/R.framework/Versions/4.x-arm64/Resources/lib/, when building a package from source?
> 
> Thank in advance,
> Philipp
> 
> _______________________________________________
> R-SIG-Mac mailing list
> R-SIG-Mac using r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-mac
>