Error ocurring in ALDA convergence with FxcGRLc and NGsBlkXd for CH4.

[Kartik.Varshney]

Dear Yambo developers,

I am trying to calculate the TDDFT optical response of a CH4 molecule using Yambo 5.3.0. The calculation stops during the setup of observables and procedures with the following error:

---------------------------------------------------------------------------------------------------------------------------------------
[ERROR] STOP signal received while in[04.01] Setup: observables and procedures
[ERROR] Impossible to define an appropriate parallel structure
---------------------------------------------------------------------------------------------------------------------------------------


The same error is reported by the MPI processes. For example:
---------------------------------------------------------------------------------------------------------------------------------------
P1-n70: [ERROR] STOP signal received while in[04.01] Setup: observables and procedures
P1-n70: [ERROR] Impossible to define an appropriate parallel structure

P10-n70: [ERROR] STOP signal received while in[04.01] Setup: observables and procedures
P10-n70: [ERROR] Impossible to define an appropriate parallel structure
---------------------------------------------------------------------------------------------------------------------------------------


I am running the calculation on an HPC cluster using SLURM with 4 nodes and 28 MPI tasks per node, i.e. 112 MPI processes.
My Yambo input is:
---------------------------------------------------------------------------------------------------------------------------------------
optics
chi
tddft

NLogCPUs=0
PAR_def_mode= "memory"

X_CPU= "1.1.1.14.8"
X_ROLEs= "g.q.k.c.v"
X_nCPU_LinAlg_INV=1
X_Threads=0

Chimod= "ALDA"
FxcGRLc= 200 eV
NGsBlkXd= 100 eV

% QpntsRXd
1 | 1 |
%

% BndsRnXd
1 | 11 |
%

% EnRngeXd
0.00000 | 20.00000 | eV
%

% DmRngeXd
0.100000 | 0.100000 | eV
%

ETStpsXd= 151

% LongDrXd
1.000000 | 0.000000 | 0.000000 |
%
---------------------------------------------------------------------------------------------------------------------------------------


The SLURM job script is:
---------------------------------------------------------------------------------------------------------------------------------------
#!/bin/bash
#SBATCH --job-name=yambo
#SBATCH --partition=test2
#SBATCH --nodes=4
#SBATCH --ntasks-per-node=28
#SBATCH --output=out_yambo.out
#SBATCH --error=err_yambo.err

module purge
module load compilers/intel/intel-OneAPI-2021.1

export PATH=$HOME/yambo-5.3.0/bin:$PATH

cd $SLURM_SUBMIT_DIR

mpirun -np $SLURM_NTASKS yambo
---------------------------------------------------------------------------------------------------------------------------------------


The Yambo executable reports:
---------------------------------------------------------------------------------------------------------------------------------------
Version 5.3.0 Revision 23959
MPI+OpenMP+SLK+HDF5_MPI_IO Build
---------------------------------------------------------------------------------------------------------------------------------------


I would like to understand why Yambo cannot construct an appropriate parallel structure for this calculation. And how does that other error that says stop signal received while in[04.01] will be resolved?
In particular, I would appreciate advice on whether my `X_CPU`/`X_ROLEs` configuration is appropriate for 112 MPI processes, and what would be the recommended parallelization settings for this TDDFT/ALDA calculation.
Should I use a different `X_CPU`, `X_ROLEs`, `PAR_def_mode`, or `X_nCPU_LinAlg_INV`?

Best regards,
Kartik Varshney
PhD, Physics
NISER, India

[Daniele Varsano]

Dear Kartik,

before going to the tddft calculation, you can try to run the setup in serial, just running yambo with an empty yambo.in file or
> yambo -i
save the yambo.in file nad run
> yambo

next you can build your tddft input file.
I suggest you to set the cutoff of gvectors in Hartree units of RL (number of reciprocal lattice vectors) instead of eV.

If the problem persists, please attach your input and also the report file, you can attach in the post by renaming them as .txt files.

Best,
Daniele