error in MPI_FILE_OPEN

[asubedi]

Hi,

I am trying to do a calculation with etf_mem = .false. When I do so, I get the following error:

Error in routine ephwan2blochp (1):
error in MPI_FILE_OPEN

This is an error while opening the prefix.epmatwp1 file.

Are there other strategies in lowering the memory consumption?

Thanks,
Alaska

[asubedi]

The error seems to happen when the "outdir" is not the same as "./".

I was using "outdir = ./tmp/". When I reran epw.x after copying ./tmp/prefix.epmatwp1 to ./, the code ran successfully.

Best,
Alaska

[sponce]

Dear Alaska,

Happy to hear that you could solve you problem.

Code: Select all

etf_mem

is indeed the way to reduce the RAM usage.

Best,

Samuel

[dlnguyen]

Hi,
I have the same problem while running EPW.
Beside the temporary solution by Alaska, is there any better solution for this problem?
The system administration seems not really happy when I use outdir ='./' since it will affect other works.
I ran on only 1 node with 24cores and outdir='/tmp/', after successfully write the epb files, prefix.epmatwp1 and prefix.epmatwe1 this problem occurred.
I appreciate any suggestions that you can give.

Best,
Duc-Long

[sponce]

Dear Duc-Long,

The issue of the ./ has been fixed and will be included in the next official release (september).

In the meantime, you can always create a different directory and run there with "./" without affecting other works.

The './' just mean it will write in the directory in which the input files are.

Best,

Samuel

[jir520]

Dear Dr. Samuel

I am having the same issue with error in opening epmatwp1 even though my input files are in the same way as your example files. I am doing electron linewidth calculation of Al. My wannier out put has created correctly, but epw.out stopped without any error message after 11th k point( I have 216 kpoints). But there is another file crated as epw.outerror. It has gone up to 216th k point and has written the below error at the end. This calculation has not created al.epmatwp1 file.


Writing epmatq on .epb files
Writing Hamiltonian, Dynamical matrix and EP vertex in Wann rep to file
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Error in routine diropn (115):
error opening ./al.epmatwp1
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%


Thank you very much
Jaya

[sponce]

Dear Jaya,

There was indeed some bugs related to that. I've fixed some of them.

Could you try with the new version 4.1 of EPW that has been release today to see if you still have this issue?

Thank you,

Samuel

[jir520]

Dear Dr. Sponce

Thank you very much and I will try it

Jaya

[jir520]

Dear Dr. Samuel

As you said above, I tried my calculation with new EPW version. But the problem is still the same. I am doing Al structure and used conventional cell. It creates al.wout correctly and epw.out gives the below error.

Writing epmatq on .epb files (The .epb files have been correctly written)
Writing Hamiltonian, Dynamical matrix and EP vertex in Wann rep to file
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Error in routine diropn (115):
error opening ./al.epmatwp1
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
stopping ...
--------------------------------------------------------------------------
MPI_ABORT was invoked on rank 0 in communicator MPI_COMM_WORLD
with errorcode 1.


Please let me know how I can solve this.

Thank you
Jaya

[sponce]

Dear Jaya,

Could you provide your inputs files (scf/nscf/epw) ?

Thank you,

Samuel