Nice catch, I implemented the patch and recompiled. Still no luck: the same
error persists.
Strangely, the example calculations that you mentioned do not fail: both
run well as expected...
The only true difference between the examples and my calculation (besides
type and number of elements) is the
Oh well. An input variable was not broadcast. Try this patch (with patch
-p1):
diff --git a/PP/src/projwfc.f90 b/PP/src/projwfc.f90
index 9210fc3a5..88e2f7cef 100644
--- a/PP/src/projwfc.f90
+++ b/PP/src/projwfc.f90
@@ -125,6 +125,7 @@ PROGRAM do_projwfc
CALL mp_bcast( lbinary_data,ionode
Hi all,
I've been looking some non-periodic systems lately, and as such have been
using the Makov-Payne correction to both correct for the dipole moment of my
molecules, but also because I occasionally am dealing with charged systems and
need to correct for the vacuum level shift. My main q
If it fits into a single GPU, run it with a single MPI process. In
parallel it is sometimes hard to figure out what fails how and why
PG
On 7/14/25 13:53, André Luiz Koch Liston wrote:
Non ricevi spesso messaggi di posta elettronica da alk2...@columbia.edu.
Scopri perché è importante
Thanks, Stefano. Yes, the results are reproducible and consistent.
Strangely, everything runs as expected when kresolveddos is set to false in
{prefix}.projwfc.in. (Though that'd also defeat the very purpose of doing
these calculations).
&PROJWFC
kresolveddos = .false.
...
/
I should also note tha
Thanks for your reply,
I was asking because if things are working properly one would expect the
sum of |psi|^2 should be less, but quite close, to the number of atomic wfs.
and the sum of the pdos for each atomic wfc projection should sum to (at
most) 1
It seams that you are missing a lot o
