You may also want to take a look at the document I posted regarding
compiler flags on my .mac website. It contains a great overview of the
effects of compiler flag optimizations with both GCC and IBM compilers:
http://homepage.mac.com/wyuen/hpc/Menu11.html
Please ignore the MPICH link. ;-).
On Oct 8, 2004, at 11:12 AM, Brian Barrett wrote:
> On Oct 8, 2004, at 10:52 AM, Christian Simon wrote:
>
>> I use IBM's XLC/XLF under MacOSX Server 10.3.5. and I need fortran
>> support,
>> so I recompiled the lam-mpi libs (7.1 version, with 7.1.1, I did not
>> tried), very simply:
>>
>> ./configure CC=xlc FC=xlf FFLAGS=-qextname
>> --prefix=/usr/local/lam-7.1 --with-trillium --disable-tv-queue
>> --with-memory-manager=none
>>
>> Things work. Ok. But I would like to improve performances.
>> What would you suggest (apart adding CFLAGS=-O3) ?
>
> LAM generally doesn't see much speedup from one optimization level to
> another - it just doesn't have much that a compiler can work with. So
> the real question is what will help your application out. Apple has a
> couple of pages up on optimizing applications for the G5, some of
> which include information on the XL compilers. You might want to have
> a look at their developer pages for more information:
>
> http://developer.apple.com/
>
>> What is the configure command issued for the package (.dmg)
>> distributed ?
>
> ./configure --prefix=PREFIX --with-rsh=ssh --without-fc \
> --without-cs-fs --enable-static --enable-shared
>
>> Does it use gcc or another C compiler ?
>
> gcc/g++, from the Developer Tools.
>
>
> Hope this helps!
>
> Brian
>
> --
> Brian Barrett
> LAM/MPI developer and all around nice guy
> Have a LAM/MPI day: http://www.lam-mpi.org/
>
> _______________________________________________
> This list is archived at http://www.lam-mpi.org/MailArchives/lam/
>
|
