[gmx-users] Gromacs slow and crashes on Leopard.

Hadas Leonov hleonov at cc.huji.ac.il
Thu Nov 29 09:52:35 CET 2007


On Wed, 2007-11-28 at 14:49 +0100, David van der Spoel wrote:
> Hadas Leonov wrote:
> > Great, thanks! 
> > 
> > It does solve the ia32 compilation problems, but not the lam-mpi
> > compilation problems - I still get undefined symbols for a few lam
> > variables. 
> > 
> > So still using open-mpi, Gromacs works a little better now, for d.villin
> > benchmark, the performance are:
> > 1 CPU: 13214 ps/day (compared to 3592 ps/day before, and 18000ps/day on
> > OSX 10.4)
> > 4 CPUs: 38000ps/day (compared to 13714 ps/day before,  41143ps/day on
> > OSX 10.4, and 48000ps/day in the gmx published benchmark). 
> >  
> > I still didn't get to make a long simulation and see if it crashes. 
> > 
> > anymore ideas? is it still something that has to be fixed in the new
> > version?
> 
> it's weird that 10.4 would be so much faster. There is a small speed 
> benefit from using the memory management routines in LAM though, but it 
> shouldn't be this big.
Will the final 3.3.3 version be fixed to compile with lam as well (on
Leopard)?

I have tried performing position restraints (2-3 hours), and it still
crashed  with the following error:

---
[tmdec2:86116] *** Process received signal ***
[tmdec2:86116] Signal: Segmentation fault (11)
[tmdec2:86116] Signal code: Address not mapped (1)
[tmdec2:86116] Failing at address: 0xcf9d849c
[ 1] [0xbfffd808, 0xcf9d849c] (-P-)
[ 2] (spread_on_grid + 0x1fa) [0xbfffd8b8, 0x0004cc53] 
[ 3] (do_pme + 0x56e) [0xbfffda18, 0x0004d1cc] 
[ 4] (force + 0x7d9) [0xbfffdc68, 0x0002e90a] 
[ 5] (do_force + 0x8a4) [0xbfffdd58, 0x0005d16e] 
[ 6] (do_md + 0x164f) [0xbfffe968, 0x000160ee] 
[ 7] (mdrunner + 0xb04) [0xbfffeae8, 0x0001453e] 
[ 8] (main + 0x463) [0xbfffeb78, 0x000186e9] 
[ 9] (start + 0x36) [0xbfffeb98, 0x00001bee] 
[10] [0x00000000, 0x0000000e] (FP-)
[tmdec2:86116] *** End of error message ***
mpirun noticed that job rank 0 with PID 86116 on node
tmdec2.ls.huji.ac.il exited on signal 11 (Segmentat
ion fault). 
1 process killed (possibly by Open MPI)
---

Help?

Thanks,
Hadas

> 
> > 
> > Thanks again,
> > Hadas.
> > 
> > 
> > On Wed, 2007-11-28 at 10:43 +0100, David van der Spoel wrote:
> >> Hadas Leonov wrote:
> >>> When is version 3.3.3 due ?
> >>>
> >> Soon.
> >>
> >> But meanwhile I put a beta version at the ftp site.
> >>
> >> ftp://ftp.gromacs.org/pub/gromacs/gromacs_3.3.3_pre1.tar.gz
> >>
> >> Please use with care, for testing purposes only.
> >>
> 
> 




More information about the gromacs.org_gmx-users mailing list