Hi there!
We have recently upgraded our OS to RedHat AS 4 (update4)
and changed a myrinet network with a Infiniband, on a
Itanium2 cluster.
We have compiled siesta (2.0 and siesta-2.0.1)
without any problem.
However, we are having problems when running parallel
jobs (using HPMPI on a 4 itanium processor Node, or
on 2 different nodes running 1 process on each node)
We have tried intel 8.1 and 9.1 compilers.
This is the error I got:
-------------------Jo0b in 4 CPUs
* Maximum dynamic memory allocated = 1 MB
siesta: ==============================
Begin CG move = 0
==============================
outcell: Unit cell vectors (Ang):
7.633698 0.000000 0.000000
0.000000 7.633698 0.000000
0.000000 0.000000 7.633698
outcell: Cell vector modules (Ang) : 7.633698 7.633698 7.633698
outcell: Cell angles (23,13,12) (deg): 90.0000 90.0000 90.0000
outcell: Cell volume (Ang**3) : 444.8411
InitMesh: MESH = 72 x 72 x 72 = 373248
InitMesh: Mesh cutoff (required, used) = 200.000 245.865 Ry
* Maximum dynamic memory allocated = 9 MB
forrtl: severe (174): SIGSEGV, segmentation fault occurred
Image PC Routine Line Source
libmpi.so.1 2000000000BA20A1 Unknown Unknown Unknown
siesta 4000000000C8A410 Unknown Unknown Unknown
siesta 4000000000C58550 Unknown Unknown Unknown
siesta 4000000000C57A30 Unknown Unknown Unknown
siesta 40000000003F08F0 Unknown Unknown Unknown
Stack trace terminated abnormally.
forrtl: severe (174): SIGSEGV, segmentation fault occurred
Image PC Routine Line Source
libmpi.so.1 2000000000BA20A1 Unknown Unknown Unknown
siesta 4000000000C8A410 Unknown Unknown Unknown
siesta 4000000000C58550 Unknown Unknown Unknown
siesta 4000000000C57A30 Unknown Unknown Unknown
siesta 40000000003F08F0 Unknown Unknown Unknown
----------------------------
When I run the same executable on a single processor the calculation finishes
correctly
an shows the following at the point where the parallel job crashes.
-----------------job on a single CPU
* Maximum dynamic memory allocated = 1 MB
siesta: ==============================
Begin CG move = 0
==============================
outcell: Unit cell vectors (Ang):
7.633698 0.000000 0.000000
0.000000 7.633698 0.000000
0.000000 0.000000 7.633698
outcell: Cell vector modules (Ang) : 7.633698 7.633698 7.633698
outcell: Cell angles (23,13,12) (deg): 90.0000 90.0000 90.0000
outcell: Cell volume (Ang**3) : 444.8411
InitMesh: MESH = 72 x 72 x 72 = 373248
InitMesh: Mesh cutoff (required, used) = 200.000 245.865 Ry
* Maximum dynamic memory allocated = 30 MB
stepf: Fermi-Dirac step function
siesta: Program's energy decomposition (eV):
siesta: Eions = 1150.308915
siesta: Ena = 50.202525
siesta: Ekin = 677.893098
siesta: Enl = -282.265243
siesta: DEna = -0.000001
siesta: DUscf = 0.000000
-----------------------------------------------
Has anyone experienced a similar problem? Or is anymone running
Siesta on a similar architecture sucesfully?
Thanks for any help!
Txema Mercero and Edu Ogando
UPV/EHU
