Dear Paolo,
Thanks for the suggestion! I have just run the job on 1 processor, and it
apparently have passed the initial stage of cutoff self-consistency
checking. (see attached). Since the smallest computation unit (node) has 16
processors on Stampede, I ran the job on a different cluster with QE
Thank you Ari for your reply. Yes you were right about the number of atoms. It
should be 72. I figured that out too right after I sent out the email. However,
I wanted to ask about what you said at the end of your reply. So do you think 3
2 2 or even 4 2 2 would be good enough to do the optimiza
Which problem?
On Mon, May 23, 2016 at 4:05 PM, Supriyo Roy
wrote:
> Dear Sir,
> I am trying to install QE-5.4.0.
> I am trying to install the cp package.
> But finally I am getting error message like
>
> gfortran -O3 -g -fopenmp -x f95-cpp-input -fopenmp -D__OPENMP -D__GFORTRAN
> -D__STD_F95 -D
Dear Sir,
I am trying to install QE-5.4.0.
I am trying to install the cp package.
But finally I am getting error message like
gfortran -O3 -g -fopenmp -x f95-cpp-input -fopenmp -D__OPENMP -D__GFORTRAN
-D__STD_F95 -D__FFTW -I../../iotk/src -I../../FFTXlib -I../../LAXlib
-I../../Modules -I. -c man
Anybody with system privileges can create a user-writable "/scratch"
directory if so desired. The code will anyway complain that "/scratch" if
not writable if it is not.
Paolo
On Mon, May 23, 2016 at 3:54 PM, Yue-Wen Fang wrote:
> Dr. Stefano, yes, I agree with you on the point of tmp_dir. The
Dr. Stefano, yes, I agree with you on the point of tmp_dir. There is no tag
for this.
However, I don't think the outdir "must" be right, generally in the linux
system, we don't have /scratch/ and most of the users cannot write data in
root directory "/". I think what he really need is "./scratch"
Dear A Suresh,
You could easily take an existing input file for the atomic code
('ld1.x') for the elements that you want - for example from the PSLibrary
- and just change the exchange-correlation functional to the one you want;
seen in /Modules/funct.f90:
! "hl" Hedin-
Intel compiler v.13 should work for the latest version of QE as well. It's
not that the latest version of QE needs the latest version of the compiler.
Your problem might also be related to MPI. Try first of all to run on one
processor, with EXACTLY the same job that crashes on 128 processors. Does
Dear Amir,
Did you check your geometry, for example visually (XCrysDen, VESTA,
...)? To me it looks as if there are some Si atoms missing in the
structure, and thus there are dangling bonds on oxygens. Was this your
purpose?
In principle you can later change your k point sampling in the
Dear Lorenzo, Paolo and others,
Thank you all for your very helpful suggestions! I have tried adding .d0
after the values for ecutwfc and ecutrho, but the same error persists. The
compilers for QE 5.4.0 on Stampede are ifort and icc 15.0.2. I compiled QE
5.1.2 using the same compilers, but the sam
I agree with Lorenzo. The check on cutoff consistency is done at the very
beginning. Cutoffs may be read from pseudopotential files if not present,
but this is not your case. I run your example and it passed the read phase
with no problems.
Paolo
On Mon, May 23, 2016 at 10:03 AM, Lorenzo Paulatto
On Sunday, 22 May 2016 18:00:49 CEST Jingyang Wang wrote:
> Dear QE users,
>
> Recently I have run a "relax" job on Stampede using QE v 5.4.0, and got
> this very brief error message:
>
> Error in routine set_cutoff (1):
> ecutrho <= ecutwfc?!?
>
> By checking the source code, I realiz
I do not think this is the problem: all non-conflicting combinations of
ibrav, celldm, A B C COSAB COSBC COSAC, CELL_PARAMETERS should work. In
case of ambiguous or conflicting settings, the code complains.
Paolo
On Mon, May 23, 2016 at 8:38 AM, Carlo Nervi wrote:
> I think that with ibrav=0 y
13 matches
Mail list logo