[Pw_forum] Re: OPTERON 64-bit and Portland pgf90 5.2-4

[Andrea Floris]

I answer myself:), hoping that it can be useful for somebody.
The ./configure put in make.sys as compilation flags:

FFLAGS         = -fast -r8 $(IFLAGS)

where -fast == -O2 -Munroll=c:1 -Mnoframe -Mlre

I changed the optimization level to -01, that is, substituting:

FFLAGS         = -fast -r8 $(IFLAGS)

with:

FFLAGS         = -O1 -Mnoframe -Mlre -r8 $(IFLAGS)

and ph.x does not give any more segmentation fault errors. If anyone has
another solution, maybe preserving the optimization level -02, please let
me know..

Thanks,

Andrea



On Tue, 26 Oct 2004 pw_forum-request at pwscf.org wrote:

>
> Message: 5
> Date: Tue, 26 Oct 2004 12:32:42 +0200 (CEST)
> From: Andrea Floris <afloris at physik.fu-berlin.de>
> To: pw_forum at pwscf.org
> Subject: [Pw_forum] OPTERON 64-bit and Portland pgf90 5.2-4
> Reply-To: pw_forum at pwscf.org
>
> Dear all, I am trying to run the PWSCF version 2.1 in a OPTERON
> cluster of 64-bit
> machines. Following the suggestion of manual, I tried the PORTLAND
> compiler, version 5.2-4. As other people reported (but with older
> versions) I was able to
> compile without problems and run pw.x also without problems. But for
> phonon calculations ph.x gives me immediately a segmentation fault error.
> What you suggest? There is a workaround to this problem?
>
> Thank yor for your answer,
>
> Andrea
>