[gentoo-science] Problems with gromacs and fft

[gentoo-science] Problems with gromacs and fft

[Mathias Weigt]

Hello!

I don't know whether to file a bug report or not. First I'd like to hear 
some suggestions.
I'm pretty desperate  now.

I cannot get gromacs to run with PME enabled (which requires fftw) on 
gentoo.
If I turn off PME the job runs.

I tried nearly all combinations of fftw (2.1.5, 3.0, 3.1) and also the 
two latest gromacs versions (3.2.1-r1, 3.3, 3.3-r1 from bugzilla).
I turned off mpi, sse, 3dnow and resetted CFLAGS to "-O2 -march=i686" 
(also on fftw) but nothing gives  me a working gromacs :-(
I also tried the (slow) built-in fftpack - same result: gromacs stops 
after some steps (sometimes with segfault)
I used gcc (GCC) 3.4.5 (Gentoo 3.4.5, ssp-3.4.5-1.0, pie-8.7.9).

Did somebody of got gromacs running with PME on gentoo?
The same job runs with the original gromacs-3.2.1-rpms installed on a 
SUSE-9.3 system.

Mathias
-- 
gentoo-science@gentoo.org mailing list

Re: [gentoo-science] Problems with gromacs and fft

[Florian Haberl]

hi,

On Wednesday 08 February 2006 12:18, Mathias Weigt wrote:
> Hello!
>
> I don't know whether to file a bug report or not. First I'd like to hear
> some suggestions.
> I'm pretty desperate  now.
>
> I cannot get gromacs to run with PME enabled (which requires fftw) on
> gentoo.
> If I turn off PME the job runs.
>
> I tried nearly all combinations of fftw (2.1.5, 3.0, 3.1) and also the
> two latest gromacs versions (3.2.1-r1, 3.3, 3.3-r1 from bugzilla).
> I turned off mpi, sse, 3dnow and resetted CFLAGS to "-O2 -march=i686"
> (also on fftw) but nothing gives  me a working gromacs :-(
> I also tried the (slow) built-in fftpack - same result: gromacs stops
> after some steps (sometimes with segfault)
> I used gcc (GCC) 3.4.5 (Gentoo 3.4.5, ssp-3.4.5-1.0, pie-8.7.9).
>
> Did somebody of got gromacs running with PME on gentoo?
> The same job runs with the original gromacs-3.2.1-rpms installed on a
> SUSE-9.3 system.

i`m working sometimes with gromacs, here it works without problems, i`ve 
tested in the past on gentoo:

3.2.1 with fftw 2.1.5 (with mpich)
3.3 with fftw 3.0.1 (without mpich)
no special optimaztion things

My own systems run without problems, also the d.lzm benchmark (included in 
gromacs benchmark suite) and pme enable works without problems.

Linux karpfen 2.6.15-ck3 #1 PREEMPT Mon Feb 6 18:01:33 GMT 2006 i686 AMD 
Athlon(tm) XP 2400+ AuthenticAMD GNU/Linux

CHOST="i686-pc-linux-gnu"
CFLAGS="-march=athlon-xp -O3 -pipe -fomit-frame-pointer"

(gcc version 3.4.5 (Gentoo 3.4.5, ssp-3.4.5-1.0, pie-8.7.9)

What does the log file say? Does it regconze your cpu, and have you started 
benchmark suite on this system or your own systems.


>
> Mathias


Greetings,

Florian

-- 
-------------------------------------------------------------------------------
 Florian Haberl                        
 Computer-Chemie-Centrum   
 Universitaet Erlangen/ Nuernberg
 Naegelsbachstr 25
 D-91052 Erlangen
 Mailto: florian.haberl AT chemie.uni-erlangen.de
-------------------------------------------------------------------------------
-- 
gentoo-science@gentoo.org mailing list

Re: [gentoo-science] Problems with gromacs and fft

[Mathias Weigt]

Servus!

Meanwhile I found the solution for the gromacs-3.3 problem and there is 
an ebuild with a patch for PME in bugzilla: 
http://bugs.gentoo.org/show_bug.cgi?id=118421

Gromacs-3.3 seems to run now. I still have no idea why gromacs-3.2.1 
didn't want to run my pme-calculation but this isn't of interest 
anymore. I think we stick to patched gromacs-3.3 now.

(The problem with pme did not occur on every system - e.g. not with the 
demo with pme enabled)

Thanks for your response

Mathias
-- 
gentoo-science@gentoo.org mailing list

Re: [gentoo-science] Problems with gromacs and fft

[Donnie Berkholz]

[-- Attachment #1: Type: text/plain, Size: 808 bytes --]

Mathias Weigt wrote:
> Servus!
> 
> Meanwhile I found the solution for the gromacs-3.3 problem and there is
> an ebuild with a patch for PME in bugzilla:
> http://bugs.gentoo.org/show_bug.cgi?id=118421
> 
> Gromacs-3.3 seems to run now. I still have no idea why gromacs-3.2.1
> didn't want to run my pme-calculation but this isn't of interest
> anymore. I think we stick to patched gromacs-3.3 now.
> 
> (The problem with pme did not occur on every system - e.g. not with the
> demo with pme enabled)

I'm looking to add that version once we get things finalized on the bug.
Until then, if somebody feels like developing it in the science overlay,
they can.

Same goes for fftw-3.1, although I haven't really touched fftw before --
that's been ribosome and phosphan.

Thanks,
Donnie


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]