[R-SIG-Mac] Using GCC4/GFORTRAN with R (and R.app)

[Simon Urbanek]

Paul,

On Dec 1, 2005, at 4:41 PM, Paul Roebuck wrote:

> Seem to remember Ripley bringing up some issues with gfortran bugs/ 
> optimization on rdev a while back. Were those resolved?

I didn't re-iterate on gcc4 recently (will do later this month), but  
the last status was that the stdio problem persists, some simple  
Fortran code segfaults, but R will pass make check (when linked to  
static libgfortran) - so yes, it's possible. Right now I'm slightly  
out of spare cycles, but I plan to experiment with universal R binary  
later this month and that involves producing full binaries using  
gcc4 ...

Jan, if you have a spare minute, can you run some benchmarks of your  
gcc4-optimized R (preferably involving mostly Fortran code vs just C)  
against the CRAN binary? That would be quite helpful in assessing the  
current speed benefits...

> Is there a plan for handling the binary distributions for the  
> differing ABIs?

Yes in the sense that they will be separate. We have added one more  
directory level to the Mac binaries for that reason.

However, the ABI-fix currently applies to R internal code only, so  
packages that need those specific LAPACK functions will be on their  
own. I guess it would be possible to use a more elaborate fix or  
provide additional entry points, but there was no need (yet?).

Cheers,
Simon