[MITgcm-devel] ifort/icc compiler flags and optfile

[EAPS]

-ip does function inclining within the same file. -ipo does x-file inlining at link time (essentially the code gets recompiled at link time which is why it takes longer). The latter method can find more opportunities for inlining. The few C routines MITgcm has should not benefit much if any from inlining (which reduces the overhead of calling frequently called routines that are small enough to cost too much to call compared to their runtime). I'm not sure they'd be helped by ipo either unless one of them gets called one too many times.

Constantinos

Sent from my iPad

On Oct 19, 2012, at 10:14 AM, Jean-Michel Campin <jmc at ocean.mit.edu> wrote:

> Hi Patrick,
> 
> regarding point 2, it is recommanded to use
> linux_amd64_ifort11
> for newer version of the intel compiler (line 7 from this optfile:
> #  Build options for version 11 and newer of intel compiler on Linux AMD64 platform
> )
> 
> regarding point 1, I don't care if you remove "-ip" from CFLAGS
> (may be leaving it commented out ? with few words of description ?);
> might be different if we remove it from FOPTIM where if could 
> affect performance (but I did not check; may be someone did ?).
> 
> Cheers,
> Jean-Michel
> 
> On Fri, Oct 19, 2012 at 09:42:35AM -0400, Patrick Heimbach wrote:
>> 
>> Hi there,
>> 
>> I have a couple of questions regarding our default optfile
>> linux_amd64_ifort
>> 
>> 1. how important is the flag "-ip" in CFLAGS (or, does it matter at all)?
>> Reason for asking is that some mpif wrappers use ifort for FC but gcc for CC (instead of icc).
>> Without the "-ip" flag in CFLAGS the optfile will work for both cases, but with this flag the compiling chokes.
>> Would we loose something if we just removed "-ip" in CFLAGS?
>> (the ifort man-pages claim that "-ip" improves runtime performance, but we hardly have any C code, so perhaps not relevant?).
>> 
>> 2. looks like '-w95' and '-xW' are "deprecated (in ifort 12) and will be removed in a future release". Perhaps time to remove them from this default optfile (unless they're important for older ifort version?)
>> 
>> Cheers
>> p.
>> 
>> ---
>> Patrick Heimbach | heimbach at mit.edu | http://www.mit.edu/~heimbach
>> MIT | EAPS 54-1420 | 77 Massachusetts Ave | Cambridge MA 02139 USA
>> FON +1-617-253-5259 | FAX +1-617-253-4464 | SKYPE patrick.heimbach
>> 
> 
> 
> 
>> _______________________________________________
>> MITgcm-devel mailing list
>> MITgcm-devel at mitgcm.org
>> http://mitgcm.org/mailman/listinfo/mitgcm-devel
> 
> 
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel