Page 1 of 1

Two issues with lpolar and mp_mesh_k fixed in the recent development branch of EPW

Posted: Tue Aug 25, 2020 11:58 pm
by hlee
Dear EPW users:

Two issues with lpolar and mp_mesh_k were identified and fixed in the recent development branch of EPW which can be downloaded at https://gitlab.com/QEF/q-e . The most-recent official version of EPW v5.3 partially has these issues since I identified one of them just a few hours before the release and the other after the release of EPW v5.3.

The issue with lpolar was present starting from the EPW v5.2. There was a issue in the ranges of the summation for the long-range correction to dynamical matrices; for details, please see this page at https://gitlab.com/QEF/q-e/-/merge_requests/1055 .
You can easily check whether or not this issue is present in your results obtained by using EPW v5.2 just by comparing the phonon dispersion from EPW with that from matdyn.x of the QE phonon code.

Regarding the issue with mp_mesh_k I think that it was present even in older versions of EPW. For the materials with the specific space groups, when generating the irreducible grids, previous versions of EPW used the symmetry operations which are not actually the symmetry operations of the crystal; for details, please see the last item of this page at https://gitlab.com/QEF/q-e/-/merge_requests/1038 .

I emphasize again that these two issues were not completely resolved in the most-recent official version of EPW v5.3; they were fixed only in the recent development branch of EPW.
Regarding the new features and the incompatible changes of EPW v5.3, please see the Releases page at https://docs.epw-code.org/doc/Releases.html .

Sincerely,

Hyungjun Lee

Re: Two issues with lpolar and mp_mesh_k fixed in the recent development branch of EPW

Posted: Sat Nov 14, 2020 12:53 am
by AgentZero
Dear Dr. Lee,

I download the develop version on 12th September. But I found that there is a minor bug in this developed version when solving the anisotropic Eliashberg equations. The memory required by the program gradually increases with the aggrandizement of the temperature. Finally, epw crashes due to insufficient memory. Some outputs are as follows. As we see, the memory required per pool is 0.48 GB, 1.41 GB, and 6.96 GB for 44 K, 45 K, and 51 K, respectively. Could you please tell me how to fix this bug?

Best wishes,
Miao Gao
-----------------------------------------------------------------------------------------------


temp( 1) = 44.00000 K

Solve anisotropic Eliashberg equations on imaginary-axis

Total number of frequency points nsiw( 1) = 63
Cutoff frequency wscut = 1.5128


Size of allocated memory per pool: ~= 0.4826 Gb

...
temp( 2) = 45.00000 K

Solve anisotropic Eliashberg equations on imaginary-axis

Total number of frequency points nsiw( 2) = 62
Cutoff frequency wscut = 1.5228


Size of allocated memory per pool: ~= 1.4100 Gb
...

temp( 8) = 51.00000 K

Solve anisotropic Eliashberg equations on imaginary-axis

Total number of frequency points nsiw( 8) = 54
Cutoff frequency wscut = 1.5049


Size of allocated memory per pool: ~= 6.6926 Gb

Size of allocated memory per pool: ~= 6.7094 Gb

Size of allocated memory per pool: ~= 6.9647 Gb

Re: Two issues with lpolar and mp_mesh_k fixed in the recent development branch of EPW

Posted: Sat Nov 14, 2020 1:02 pm
by hpaudya1
Hi Miao,

Yes, you are right. There was a problem in memory release while solving anisotropic Eliashberg equations in the dev. versions from 8th Sep to 30th Sep, and unfortunately you used in 12th Sep. After this, we have fixed the issue. Please consider using the latest version.

Happy EPWing,
Hari Paudyal

Re: Two issues with lpolar and mp_mesh_k fixed in the recent development branch of EPW

Posted: Sat Nov 14, 2020 4:33 pm
by hlee
Dear Miao Gao:

I would like to kindly ask you in the future to post the question separately in the corresponding board instead of posting it as replies to the announcements.

I reposted your question and its answer from Hari here (viewtopic.php?f=6&t=1344).

Sincerely,

H. Lee

Re: Two issues with lpolar and mp_mesh_k fixed in the recent development branch of EPW

Posted: Mon Nov 27, 2023 7:27 am
by Anile78
AgentZero wrote:
Sat Nov 14, 2020 12:53 am
Dear Dr. Lee,

I download the develop version on 12th September. But I found that there is a minor bug in this developed version when solving the anisotropic Eliashberg equations. The memory required by the program gradually increases with the aggrandizement of the temperature. Finally, epw crashes due to insufficient memory. Some outputs are as follows. As we see, the memory required per pool is 0.48 GB, 1.41 GB, and 6.96 GB for 44 K, 45 K, and 51 K, respectively. Could you please tell me how to fix this bug?

Encountered a minor bug in the developmental version downloaded on September 12th. While solving anisotropic Eliashberg equations, observed a memory escalation issue with temperature increase, leading to crashes.

Re: Two issues with lpolar and mp_mesh_k fixed in the recent development branch of EPW

Posted: Thu Nov 30, 2023 6:08 pm
by Anile78
hpaudya1 wrote:
Sat Nov 14, 2020 1:02 pm
Hi Miao,

Yes, you are right. There was a problem in memory release while solving anisotropic Eliashberg equations in the dev. versions from 8th Sep to 30th Sep, and unfortunately you used in 12th Sep. After this, we have fixed the issue. Please consider using the latest version.

Happy EPWing,
Hari Paudyal

Certainly!
Acknowledged. Encountered memory release issues with the anisotropic Eliashberg equations in the version dated 12th Sep. Glad to hear it has been addressed in the latest release post-30th Sep.Nrega

Re: Two issues with lpolar and mp_mesh_k fixed in the recent development branch of EPW

Posted: Tue Feb 13, 2024 5:01 pm
by nikita5
Anile78 wrote:
Thu Nov 30, 2023 6:08 pm
hpaudya1 wrote:
Sat Nov 14, 2020 1:02 pm
Hi Miao,

Yes, you are right. There was a problem in memory release while solving anisotropic Eliashberg equations in the dev. versions from 8th Sep to 30th Sep, and unfortunately you used in 12th Sep. After this, we have fixed the issue. Please consider using the latest version.

Happy EPWing,
Hari Paudyal

Certainly!
Acknowledged. Encountered memory release issues with the anisotropic Eliashberg equations in the version dated 12th Sep. Glad to hear it has been addressed in the latest release post-30th Sep.Nrega
Absolutely! It's reassuring to note that the memory release issues with the anisotropic Eliashberg equations, identified in the version dated 12th Sep, have been successfully addressed in the latest release post-30th Sep. This responsive approach ensures a smoother user experience, underscoring the commitment to continuous improvement and reliability. Kudos!