Dear Carlo,
if I must guess, 90% that it is a problem of the gfortran
compiler. I recognize that the situation of fortran compilers
and of the floating-point behavior of CPUs is very annoying.
I'm using the Intel compiler with MKL and FFTW3 and I've run
most calculations of x86_64 CPUs. If I
ype)
La morale est une logique de l'action comme la logique est une morale de la
pens?e - Jean Piaget
-- next part --
An HTML attachment was scrubbed...
URL:
http://www.democritos.it/pipermail/pw_forum/attachments/20111025/82ed764e/attachment.htm
---
An HTML attachment was scrubbed...
URL:
http://www.democritos.it/pipermail/pw_forum/attachments/20111025/0fea3078/attachment-0001.htm
yes it's offensive.
-- next part --
An HTML attachment was scrubbed...
URL:
http://www.democritos.it/pipermail/pw_forum/attachments/20111025/4ebc7707/attachment.htm
Hello,
I tried to compile using the GOTO2 BLAS libs, but I got the same error
in gipaw.
So I tried to compile all with gfortran 4.4.5, but again I got the same
NaN message, with again the following:
Computing the magnetic susceptibility isolve=0ethr=0.1000E-13
k-point #1
You could try to compile using this option (all together, without spaces):
-ffpe-trap=invalid,zero,overflow,underflow,denormal
to force the code to crash at the first appearance of NaN, this could help
track down the source of the problem.
On Tue, 25 Oct 2011 18:18:58 +0200, Carlo Nervi
Il 25/10/2011 18.22, Lorenzo Paulatto ha scritto:
> You could try to compile using this option (all together, without spaces):
> -ffpe-trap=invalid,zero,overflow,underflow,denormal
> to force the code to crash at the first appearance of NaN, this could help
> track down the source of the problem
Recent versions of gfortran have a -traceback option that (together with
-g, I think) would give you the (almost) exact code line that causes the
problem; without that you shoudl still be able to run
addr2line -e /wherever/bin/gipaw.x 0x44e576
to get the line of cgsolve_all that triggers a NaN
rubbed...
URL:
http://www.democritos.it/pipermail/pw_forum/attachments/20111025/02144716/attachment.htm
Okay,
thanks again.
addr2line -e /home/nervi/src/QE432/espresso-4.3.2/GIPAW/src/gipaw.x 0x44e576
/home/nervi/src/QE432/espresso-4.3.2/qe-gipaw-4.3.2/src/cgsolve_all.f90:225
It should be the line with dclamda (see below) - no idea how to furter
help the developers ;-)
Carlo
#ifdef __PAR
-- next part --
An HTML attachment was scrubbed...
URL:
http://www.democritos.it/pipermail/pw_forum/attachments/20111025/e2776dca/attachment.htm
> Dear Quantum Espresso Users
>
> I think the jobs run slow when they are swapping memory to
> disk.
> Is there anything I can do to distribute the memory across
> more nodes so
> they don???t have to swap?
> I want to reduce RAM usage in practice. How it is
> possible?
It is generally a very bad
Dear PWscf forum members and developers,
in the future I would like to perform calculations on systems where
relativistic effects
are not negligible. Although I have generated USPP potentials using the code of
Prof. David Vanderbilt's group, after browsing through the forum and checking
out so
On Oct 25, 2011, at 19:05 , Carlo Nervi wrote:
> It should be the line with dclamda (see below) - no idea how to
> further
> help the developers ;-)
> dclambda = CMPLX( - a(lbnd) / c(lbnd), 0.d0,kind=DP)
print c(lbnd), check if it is zero; if it is, try to find out where
it come
On Oct 25, 2011, at 19:52 , Carlo Nervi wrote:
> in namelist &CONTROL disk_io='minimal', or even 'none'
> see here:
> http://www.quantum-espresso.org/user_guide/node18.html
note that this will actually increase RAM usage, because
wavefunctions are not stored to disk but kept in memory
P.
---
P
On Oct 25, 2011, at 19:55 , Kiss, Ioan wrote:
> If there are no such potentials widely available, can they be
> easily generated with the ld.x package
generated, yes; easily...not sure
P.
---
Paolo Giannozzi, Dept of Chemistry&Physics&Environment,
Univ. Udine, via delle Scienze 208, 33100 Udine
There are several ld1.x inputs in
http://qe-forge.org/projects/pslibrary/
They are still experimental, but you can use them to generate fully
relativistic PPs.
HTH
Andrea
On Tue, 2011-10-25 at 17:55 +, Kiss, Ioan wrote:
> Dear PWscf forum members and developers,
>
> in the future I woul
17 matches
Mail list logo