Dear M. Swaroop,
Did you compile siesta with a 64 bit compiler? I think you used
a 32 bit and you have overflown the number of grid points.
(64000 x829440 x984150= 7.05E17)
So you have to reduce the mesh cutoff and/or use a 64 bits compiler.
The system is really big, where you able to diagonalize
the hamiltoninan in a different CG step? If yes why it is rebuilding
the mesh, are you using variable cell CG?
For such a big system do you need so many kpoints?
Best regards,
Eduardo
On 22/05/2008, at 11:20, M.Sairam Swaroop wrote:
Dear SIESTA Users,
We are facing an annoying problem for CG runs on a HPMPI intel
machine.
The siesta job runs fine for the first CG step and then crashes for
the
next step. On analysis we found an unreasonable value for MESH =
-460324864(large & -ve ) .
Could anybody help us out.
OUT PUT FILE
................
siesta: k-point coordinates (Bohr**-1) and weights:
siesta: 1 -0.073104 -0.000015 -0.000002 0.111111
siesta: 2 0.000000 0.000000 0.000000 0.111111
siesta: 3 0.073104 0.000015 0.000002 0.111111
siesta: 4 -0.073086 0.084214 -0.000005 0.222222
siesta: 5 0.000018 0.084230 -0.000003 0.222222
siesta: 6 0.073122 0.084245 -0.000001 0.222222
siesta: k-grid: Number of k-points = 6
siesta: k-grid: Cutoff = 15.075 Ang
siesta: k-grid: Supercell and displacements
siesta: k-grid: 3 0 0 0.000
siesta: k-grid: 0 3 0 0.000
siesta: k-grid: 0 0 1 0.000
superc: Internal auxiliary supercell: 2 x 2 x 1
= 4
superc: Number of atoms, orbitals, and projectors: 736 9232
11272
outcell: Unit cell vectors (Ang):
15.160609 -0.003215 -0.000482
-0.002785 13.158104 0.000469
-0.000960 0.001078 30.149929
outcell: Cell vector modules (Ang) : 15.160610 13.158104
30.149929
outcell: Cell angles (23,13,12) (deg): 89.9959 90.0036
90.0243
outcell: Cell volume (Ang**3) : 6014.4545
InitMesh: MESH =864000 x829440 x984150 = -460324864
InitMesh: Mesh cutoff (required, used) = 400.000 414.060 Ry
thanks
swaroop
--
M.Sairam Swaroop
Theoretical Sciences Unit
JNCASR,Jakkur
Bangalore-64
INDIA
