[R-SIG-Mac] GCC on Lion and above

Simon Urbanek simon.urbanek at r-project.org
Sat Apr 7 19:05:26 CEST 2012


Charlie,

On Apr 7, 2012, at 12:34 PM, Charlie Sharpsteen wrote:

> 
> 
> On Monday, April 2, 2012 8:30:46 AM UTC-7, Timothy Bates wrote:
> Hi All,
> HPC seem to be maintaining the gcc toolchain up to date (they have GCC 4.7 compiled with autovectoring using OpenMP…)
> 
> http://hpc.sourceforge.net
> 
> BUT the page  http://r.research.att.com/tools/  says "do not use compilers from HPC, they won't work correctly!” Is that the case?
> 
> Also, I wondered if http://www.macports.org might be the way to go to get a version of gcc with a non-crashing OpenMP library?
> 
> best wishes and thanks for your help
> 
> tim
> 
> PS: The att.com page talks about install disks for OS X, but I think it’s all via the app store now, including X Code.
> 
> 
> As far as I am aware, the problem with HPC binaries or MacPorts builds is that they cannot compile universal binaries. In order to do this, you need to build GFortran from Apple's patched sources for GCC 4.2.x or use Simon's binaries from r.research.att.com.
> 
> If you don't care about a multi-architecture version of R, then the HPC and MacPorts builds will probably work just fine---you may need to set R_ARCHS or build a single-architecture version yourself so that R does not try to build universal binaries when compiling packages from source.
> 

You seem to be implying that you would use CRAN binary and then another compiler - that's not what we talked about and it has even more issues, because you will have potential problem with run-time libraries, ABIs etc. They differ by compiler version and R can load only one, so typically it will load the older, system library which may cause problem in the code that is compiled with the other compiler. If you want to use a different compiler, you should really compile R from sources, otherwise there is no guarantee that anything will work. It may or may not and the issues may be subtle, so beware. The discussion about compilers was really implying that you compile everything (including R) from sources, because otherwise it makes no sense as you will still have the same issues with GOMP etc.

Cheers,
Simon



More information about the R-SIG-Mac mailing list