compilation error on HPC

[vitoversace]

Dear Daniele,

i managed to install yambo on HPC for openmp, but when running a test gw calculation after a while at one specific point the calculation stops with following error :

forrtl: severe (24): end-of-file during read, unit 41, file /home/vito/yambo_openmp/stable/bin/test_abinit/./Real_Axis//db.dipoles
Image PC Routine Line Source
yambo 00000000004084A2 Unknown Unknown Unknown
yambo 00000000004056F6 Unknown Unknown Unknown
libc.so.6 00002AF19D304D1D Unknown Unknown Unknown
yambo 0000000000404999 Unknown Unknown Unknown

Do you see here what the problem is ?

Bests
Vito

[Daniele Varsano]

Dear Vito,
it is hard to say, it looks you have corrupted database files.
The dipole database were built in the same run or in a previous run?
Your compilation is MPI+openmp or just openmp?

Best,
Daniele

[vitoversace]

Dear Daniele,

The dipole database were built in the same run or in a previous run?

In the same run.

Your compilation is MPI+openmp or just openmp?

Compilation was just openmp. It seems not to work properly with openmp.
Openmp is slow, too.

But any way i managed to install yambo with mpi on HPC and is very fast.
Just can not wait for the new release of yambo with all new features

Ciao
Vito

[haseebphysics1]

Dear Developers,

Is there any guide to the compilation and properly making of Yambo on the HPC cluster environment. I have a cluster to which I want to have Yambo. Each node has a FORTRAN compiler. What are the prerequisite to cluster installation? Can someone please highlight the main steps which I need to do.
e.g.
> NFS,
> Same compiler version on each node essential?
> Setup of head-node... etc.
I don't have any system admin and someone just setup the SLURM job management system in the nodes.


Regards,

[Daniele Varsano]

Dear Haseeb,
have you tried to compile the code?
You need to have the MPI fortran/C compilers and configure the code. If you do not have libraries already installed yambo will donwload and compile them.
./configure
make yambo
If you have problems post here the error message and your config.log file.

Daniele

[haseebphysics1]

Dear Daniele,

I have compiled the code using gcc version 7.2.0 (GCC) on the head node and running the code locally (without sbatch) is working fine, but when I try to run it on the cluster nodes using SLURM it gets an error of Fortran libraries or so...

I am quite sure that all the other nodes will be running the lower version of the GCC compiler! But is this an issue or should this be an issue?

The SLURM output, Yambo input, batch script, and yambo config files are attached.



Thanks,

[Daniele Varsano]

Dear Haseeb,
it seems you have some conflicts with the linear algebra libraries.
You can try the following:
> make distclean
>./configure --enable-int-linalg
> make yambo


Best,
Daniele

[haseebphysics1]

Thanks very much for the suggestion,

but then why it is working when I run it locally (not using the slurm)? I have completed the bse spectra while running the same yambo executable.


Thanks,

[Daniele Varsano]

Because most probably you are linking different libraries or some of the nodes do not have access to the requested libraries.
I think that you should discuss these issues with the people that set up the cluster, as this is not a problem related to the code.
Best,
Daniele