Parallelization problem

[javad]

Dear yambo developers

I'm trying the new yambo, version 4.0.1, I found some problem in parallelization.

I have a system with 100 k-points and 100 q-points.
For a GW calculation I would like to parallelize on q-points for the X and quasi-particle for the GW.
I have 5 processors, so I set

X_all_q_CPU= "5 1 1 1" # [PARALLEL] CPUs for each role
X_all_q_ROLEs= "q k c v" # [PARALLEL] CPUs roles (q,k,c,v)
....
SE_CPU= "1 5 1" # [PARALLEL] CPUs for each role
SE_ROLEs= "q qp b" # [PARALLEL] CPUs roles (q,qp,b)

but I get the error message

P0005: [ERROR] STOP signal received while in :[05] Dynamic Dielectric Matrix (PPA)
P0005: [ERROR]Incomplete Parallel Index Filling

why? the number of q-points is divisible by 5 and also the number of quasi-particles.
What's wrong?
I attach pwscf input files and yambo input files.

regards
Javad

[Daniele Varsano]

Dear Javad,
The entries of the cpu for each role needs to be power of 2.
Anyway, please when posting input files, please post also the report file.

Best,
Daniele

[Zahra Taghipour]

Hello,

I am getting some error when I try to use mpirun for my GW calculation, and I was hoping you could help me resolve it.
My SCF and NSCF files, in addition to the Yambo input and r_setup files are attached.

The error I get in terminal is:
At line 84 of file PARALLEL_get_user_structure.f90
Fortran runtime error: Bad value during integer read

I have a large unit cell consisting of 64 atoms and 256 electrons. Previously, I tested k_points = 10, and nbnd= 500. However, the Bandgap was not even close. So, I decided to include more empty bands in my calculation. And then I received this error! It does the calculation in serial, though.
I have used option "-S" in my command lines, by the way.

I would be grateful if you could help me know what the problem is.

Zahra Taghipour,
PhD student,
UNM, USA

[Daniele Varsano]

Dear Zahra,
the setup looks ok.
in order to understand what is the problem please post the LOG files and report file of the crashed calculations.
Moreover it is also useful if you post your submission script.

Best,
Daniele

[Zahra Taghipour]

Dear Developers,

Recently, I am recieving the same error of " [ERROR]Incomplete Parallel Index Filling ", with any combination of parallel definition!
I was wondering if you could help me resolve this issue.
It seems Dipole and the very first q point calculation are being done like normal; however it gives this error just after it is writing ndb.em1d_fragment_1!

The report and log files are attached.

Thank You,
Best,
Zahra
PhD student, UNM, USA

[Daniele Varsano]

Dear Zahra,
Can you update your code release?
There have been bug fixes in the last releases. If you downloaded the code from the svn you need just to type:

Code: Select all

svn update

from the yambo directory and recompile by scratch. Otherwise just follow instructions from the download section of the yambo website.
If the problem persists, please post again your input/reports and your job launching script as well.

Best,
Daniele