[QE-users] Some inconsistency between QE-6.7MaX and current QE-7.1
Dear QE Developers, Hi. Kindly, when I use the same scf input for QE-6.7MaX and QE-7.1, the results of forces on atoms are more or less similar but with a small difference. However, the calculated pressures are SIGNIFICANTLY_DIFFERENT. Secondly, when the number of atoms is of order say 50, the calculation of forces on atoms using QE-7.1 is VERY VERY time consuming. In the release note of QE7.1 one feature mentioned is an "automatic" optimizations for accelerating the calculations. Could it happen that in some cases instead of optimizations one encounters with deceleration? How can one disable this automatic optimization? Thank you in advance. With Best Regards, Mahmoud Payami NSTRI, AEOI, Tehran, Iran Email: mpay...@aeoi.org.ir Phone: +98 (0)21 82066504 ___ The Quantum ESPRESSO community stands by the Ukrainian people and expresses its concerns about the devastating effects that the Russian military offensive has on their country and on the free and peaceful scientific, cultural, and economic cooperation amongst peoples ___ Quantum ESPRESSO is supported by MaX (www.max-centre.eu) users mailing list users@lists.quantum-espresso.org https://lists.quantum-espresso.org/mailman/listinfo/users
Re: [QE-users] Some inconsistency between QE-6.7MaX and current QE-7.1
The optimization you are referring to is actually an automatic choice of parallelization option, but it is not used if you explicitly specify the parallelization options (e.g. with -nk N -nd M). For the original problem: one has to see under which exact coonditions it happens Paolo On 20/09/2022 05:01, Mahmoud Payami Shabestari via users wrote: Dear QE Developers, Hi. Kindly, when I use the same scf input for QE-6.7MaX and QE-7.1, the results of forces on atoms are more or less similar but with a small difference. However, the calculated pressures are SIGNIFICANTLY_DIFFERENT. Secondly, when the number of atoms is of order say 50, the calculation of forces on atoms using QE-7.1 is VERY VERY time consuming. In the release note of QE7.1 one feature mentioned is an "automatic" optimizations for accelerating the calculations. Could it happen that in some cases instead of optimizations one encounters with deceleration? How can one disable this automatic optimization? Thank you in advance. With Best Regards, Mahmoud Payami NSTRI, AEOI, Tehran, Iran Email: mpay...@aeoi.org.ir Phone: +98 (0)21 82066504 ___ The Quantum ESPRESSO community stands by the Ukrainian people and expresses its concerns about the devastating effects that the Russian military offensive has on their country and on the free and peaceful scientific, cultural, and economic cooperation amongst peoples ___ Quantum ESPRESSO is supported by MaX (www.max-centre.eu) users mailing list users@lists.quantum-espresso.org https://lists.quantum-espresso.org/mailman/listinfo/users -- Paolo Giannozzi, Dip. Scienze Matematiche Informatiche e Fisiche, Univ. Udine, via delle Scienze 208, 33100 Udine, Italy Phone +39-0432-558216, fax +39-0432-558222 ___ The Quantum ESPRESSO community stands by the Ukrainian people and expresses its concerns about the devastating effects that the Russian military offensive has on their country and on the free and peaceful scientific, cultural, and economic cooperation amongst peoples ___ Quantum ESPRESSO is supported by MaX (www.max-centre.eu) users mailing list users@lists.quantum-espresso.org https://lists.quantum-espresso.org/mailman/listinfo/users
Re: [QE-users] Some inconsistency between QE-6.7MaX and current QE-7.1
Dear Paolo, Thank you so much for the comment on automatic parallelization. On the second issue, how can one estimate the execution time for force calculation as a function of number of Hubbard levels in a DFT+U calculation with ortho-atomic U_projection? Thank you in advance. Bests, Mahmoud From: Paolo Giannozzi To: users@lists.quantum-espresso.org Date: Tue, 20 Sep 2022 07:58:42 +0200 Subject: Re: [QE-users] Some inconsistency between QE-6.7MaX and current QE-7.1 The optimization you are referring to is actually an automatic choice of parallelization option, but it is not used if you explicitly specify the parallelization options (e.g. with -nk N -nd M). For the original problem: one has to see under which exact coonditions it happens Paolo On 20/09/2022 05:01, Mahmoud Payami Shabestari via users wrote: > Dear QE Developers, > Hi. > Kindly, when I use the same scf input for QE-6.7MaX and QE-7.1, the > results of forces on atoms are more or less similar but with a small > difference. However, the calculated pressures are SIGNIFICANTLY_DIFFERENT. > Secondly, when the number of atoms is of order say 50, the calculation > of forces on atoms using QE-7.1 is VERY VERY time consuming. > In the release note of QE7.1 one feature mentioned is an "automatic" > optimizations for accelerating the calculations. > Could it happen that in some cases instead of optimizations one > encounters with deceleration? > How can one disable this automatic optimization? > Thank you in advance. > With Best Regards, > Mahmoud Payami > NSTRI, AEOI, Tehran, Iran > Email: mpay...@aeoi.org.ir > Phone: +98 (0)21 82066504 > > > ___ > The Quantum ESPRESSO community stands by the Ukrainian > people and expresses its concerns about the devastating > effects that the Russian military offensive has on their > country and on the free and peaceful scientific, cultural, > and economic cooperation amongst peoples > ___ > Quantum ESPRESSO is supported by MaX (www.max-centre.eu) > users mailing list users@lists.quantum-espresso.org > https://lists.quantum-espresso.org/mailman/listinfo/users -- Paolo Giannozzi, Dip. Scienze Matematiche Informatiche e Fisiche, Univ. Udine, via delle Scienze 208, 33100 Udine, Italy Phone +39-0432-558216, fax +39-0432-558222 ___ The Quantum ESPRESSO community stands by the Ukrainian people and expresses its concerns about the devastating effects that the Russian military offensive has on their country and on the free and peaceful scientific, cultural, and economic cooperation amongst peoples ___ Quantum ESPRESSO is supported by MaX (www.max-centre.eu) users mailing list users@lists.quantum-espresso.org https://lists.quantum-espresso.org/mailman/listinfo/users___ The Quantum ESPRESSO community stands by the Ukrainian people and expresses its concerns about the devastating effects that the Russian military offensive has on their country and on the free and peaceful scientific, cultural, and economic cooperation amongst peoples ___ Quantum ESPRESSO is supported by MaX (www.max-centre.eu) users mailing list users@lists.quantum-espresso.org https://lists.quantum-espresso.org/mailman/listinfo/users
Re: [QE-users] Some inconsistency between QE-6.7MaX and current QE-7.1
Not sure what you mean exactly. As a rule of thumb, the calculation of the Hubbard contribution to forces should take no more than 1/3 or 1/4, likely much less than that, of the time needed for a scf calculation, at least for the simple ("atomic") U projection. Paolo On 21/09/2022 19:31, Mahmoud Payami Shabestari via users wrote: Dear Paolo, Thank you so much for the comment on automatic parallelization. On the second issue, how can one estimate the execution time for force calculation as a function of number of Hubbard levels in a DFT+U calculation with ortho-atomic U_projection? Thank you in advance. Bests, Mahmoud From: Paolo Giannozzi To: users@lists.quantum-espresso.org Date: Tue, 20 Sep 2022 07:58:42 +0200 Subject: Re: [QE-users] Some inconsistency between QE-6.7MaX and current QE-7.1 The optimization you are referring to is actually an automatic choice of parallelization option, but it is not used if you explicitly specify the parallelization options (e.g. with -nk N -nd M). For the original problem: one has to see under which exact coonditions it happens Paolo On 20/09/2022 05:01, Mahmoud Payami Shabestari via users wrote: > Dear QE Developers, > Hi. > Kindly, when I use the same scf input for QE-6.7MaX and QE-7.1, the > results of forces on atoms are more or less similar but with a small > difference. However, the calculated pressures are SIGNIFICANTLY_DIFFERENT. > Secondly, when the number of atoms is of order say 50, the calculation > of forces on atoms using QE-7.1 is VERY VERY time consuming. > In the release note of QE7.1 one feature mentioned is an "automatic" > optimizations for accelerating the calculations. > Could it happen that in some cases instead of optimizations one > encounters with deceleration? > How can one disable this automatic optimization? > Thank you in advance. > With Best Regards, > Mahmoud Payami > NSTRI, AEOI, Tehran, Iran > Email: mpay...@aeoi.org.ir <mailto:mpayami%40aeoi.org.ir> > Phone: +98 (0)21 82066504 > > > ___ > The Quantum ESPRESSO community stands by the Ukrainian > people and expresses its concerns about the devastating > effects that the Russian military offensive has on their > country and on the free and peaceful scientific, cultural, > and economic cooperation amongst peoples > ___ > Quantum ESPRESSO is supported by MaX (www.max-centre.eu <https://eur01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fwww.max-centre.eu%2F&data=05%7C01%7Cpaolo.giannozzi%40uniud.it%7C2303e817de0d4bf50b0d08da9bf72673%7C6e6ade15296c4224ac581c8ec2fd53a8%7C0%7C0%7C637993783089388282%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=F4aYKbdRY0a9vRh3SpapAW4yjdSGX9Fk0TuoTxkPR08%3D&reserved=0>) > users mailing list users@lists.quantum-espresso.org <mailto:users%40lists.quantum-espresso.org> > https://lists.quantum-espresso.org/mailman/listinfo/users <https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.quantum-espresso.org%2Fmailman%2Flistinfo%2Fusers&data=05%7C01%7Cpaolo.giannozzi%40uniud.it%7C2303e817de0d4bf50b0d08da9bf72673%7C6e6ade15296c4224ac581c8ec2fd53a8%7C0%7C0%7C637993783089388282%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=KhYSLBGfZ4qsPLmoKsz86DtWUCejGdUMq389K8PioUY%3D&reserved=0> -- Paolo Giannozzi, Dip. Scienze Matematiche Informatiche e Fisiche, Univ. Udine, via delle Scienze 208, 33100 Udine, Italy Phone +39-0432-558216, fax +39-0432-558222 ___ The Quantum ESPRESSO community stands by the Ukrainian people and expresses its concerns about the devastating effects that the Russian military offensive has on their country and on the free and peaceful scientific, cultural, and economic cooperation amongst peoples ___ Quantum ESPRESSO is supported by MaX (www.max-centre.eu <https://eur01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fwww.max-centre.eu%2F&data=05%7C01%7Cpaolo.giannozzi%40uniud.it%7C2303e817de0d4bf50b0d08da9bf72673%7C6e6ade15296c4224ac581c8ec2fd53a8%7C0%7C0%7C637993783089544472%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=YKY7%2FigwjZrgcQDnGqxhc5yOf7vkJDcIR%2FV3k6cVLmA%3D&reserved=0>) users mailing list users@lists.quantum-espresso.org
Re: [QE-users] Some inconsistency between QE-6.7MaX and current QE-7.1
Dear Paolo, Thank you so much for your comments. Actually, to find some sense about the time needed to calculate force, I have performed scf calculations for just one iteration, but different numbers of atoms (4, 8, 12, 16, 32, 48, 64, 80, 96, 108) with more or less the sama number of k points in BZ. The supercells were generated from the QE-7.1/PW/example08 unitcell. The inputs and outputs are uploaded in: https://drive.google.com/drive/folders/12cWLR7PkMI1wolxgRLnOSCi4IhtnLmEO?usp=sharing The results for low number of atoms seems to take reasonable times, but for larger ones the time to calculate force becomes larger than the time for just one iteration. Do you see any un-normal behavior in force calculation routine? Thank you very much for your comments. Bests, Mahmoud From: Paolo Giannozzi To: Mahmoud Payami Shabestari , Quantum ESPRESSO users Forum Date: Thu, 22 Sep 2022 18:18:51 +0200 Subject: Re: [QE-users] Some inconsistency between QE-6.7MaX and current QE-7.1 Not sure what you mean exactly. As a rule of thumb, the calculation of the Hubbard contribution to forces should take no more than 1/3 or 1/4, likely much less than that, of the time needed for a scf calculation, at least for the simple ("atomic") U projection. Paolo On 21/09/2022 19:31, Mahmoud Payami Shabestari via users wrote: > Dear Paolo, > Thank you so much for the comment on automatic parallelization. > On the second issue, how can one estimate the execution time for force > calculation as a function of number of Hubbard levels in a DFT+U > calculation with ortho-atomic U_projection? > Thank you in advance. > Bests, > Mahmoud > > From: Paolo Giannozzi > To: users@lists.quantum-espresso.org > Date: Tue, 20 Sep 2022 07:58:42 +0200 > Subject: Re: [QE-users] Some inconsistency between QE-6.7MaX and > current QE-7.1 > The optimization you are referring to is actually an automatic choice of > parallelization option, but it is not used if you explicitly specify the > parallelization options (e.g. with -nk N -nd M). For the original > problem: one has to see under which exact coonditions it happens > > Paolo > > On 20/09/2022 05:01, Mahmoud Payami Shabestari via users wrote: > > Dear QE Developers, > > Hi. > > Kindly, when I use the same scf input for QE-6.7MaX and QE-7.1, the > > results of forces on atoms are more or less similar but with a small > > difference. However, the calculated pressures are > SIGNIFICANTLY_DIFFERENT. > > Secondly, when the number of atoms is of order say 50, the > calculation > > of forces on atoms using QE-7.1 is VERY VERY time consuming. > > In the release note of QE7.1 one feature mentioned is an "automatic" > > optimizations for accelerating the calculations. > > Could it happen that in some cases instead of optimizations one > > encounters with deceleration? > > How can one disable this automatic optimization? > > Thank you in advance. > > With Best Regards, > > Mahmoud Payami > > NSTRI, AEOI, Tehran, Iran > > Email: mpay...@aeoi.org.ir <mailto:mpayami%40aeoi.org.ir> > > Phone: +98 (0)21 82066504 > > > > > > ___ > > The Quantum ESPRESSO community stands by the Ukrainian > > people and expresses its concerns about the devastating > > effects that the Russian military offensive has on their > > country and on the free and peaceful scientific, cultural, > > and economic cooperation amongst peoples > > ___ > > Quantum ESPRESSO is supported by MaX (www.max-centre.eu > <https://eur01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fwww.max-centre.eu%2F&data=05%7C01%7Cpaolo.giannozzi%40uniud.it%7C2303e817de0d4bf50b0d08da9bf72673%7C6e6ade15296c4224ac581c8ec2fd53a8%7C0%7C0%7C637993783089388282%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=F4aYKbdRY0a9vRh3SpapAW4yjdSGX9Fk0TuoTxkPR08%3D&reserved=0>) > > users mailing list users@lists.quantum-espresso.org > <mailto:users%40lists.quantum-espresso.org> > > https://lists.quantum-espresso.org/mailman/listinfo/users > <https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.quantum-espresso.org%2Fmailman%2Flistinfo%2Fusers&data=05%7C01%7Cpaolo.giannozzi%40uniud.it%7C2303e817de0d4bf50b0d08da9bf72673%7C6e6ade15296c4224ac581c8ec2fd53a8%7C0%7C0%7C637993783089388282%7CUnknown%7CTWFpbGZsb3d8eyJWIj
Re: [QE-users] Some inconsistency between QE-6.7MaX and current QE-7.1
Hubbard forces and stresses using "ortho-atomic" Hubbard projectors are much more expensive than with "atomic". It is explained in PRB 102, 235159 (2020). Computational scaling is discussed in the appendix C. Iurii -- Dr. Iurii TIMROV Senior Research Scientist Theory and Simulation of Materials (THEOS) Swiss Federal Institute of Technology Lausanne (EPFL) CH-1015 Lausanne, Switzerland +41 21 69 34 881 http://people.epfl.ch/265334 From: users on behalf of Mahmoud Payami Shabestari via users Sent: Thursday, September 22, 2022 9:03:17 PM To: Paolo Giannozzi; Quantum ESPRESSO users Forum Subject: Re: [QE-users] Some inconsistency between QE-6.7MaX and current QE-7.1 Dear Paolo, Thank you so much for your comments. Actually, to find some sense about the time needed to calculate force, I have performed scf calculations for just one iteration, but different numbers of atoms (4, 8, 12, 16, 32, 48, 64, 80, 96, 108) with more or less the sama number of k points in BZ. The supercells were generated from the QE-7.1/PW/example08 unitcell. The inputs and outputs are uploaded in: https://drive.google.com/drive/folders/12cWLR7PkMI1wolxgRLnOSCi4IhtnLmEO?usp=sharing The results for low number of atoms seems to take reasonable times, but for larger ones the time to calculate force becomes larger than the time for just one iteration. Do you see any un-normal behavior in force calculation routine? Thank you very much for your comments. Bests, Mahmoud From: Paolo Giannozzi To: Mahmoud Payami Shabestari , Quantum ESPRESSO users Forum Date: Thu, 22 Sep 2022 18:18:51 +0200 Subject: Re: [QE-users] Some inconsistency between QE-6.7MaX and current QE-7.1 Not sure what you mean exactly. As a rule of thumb, the calculation of the Hubbard contribution to forces should take no more than 1/3 or 1/4, likely much less than that, of the time needed for a scf calculation, at least for the simple ("atomic") U projection. Paolo On 21/09/2022 19:31, Mahmoud Payami Shabestari via users wrote: > Dear Paolo, > Thank you so much for the comment on automatic parallelization. > On the second issue, how can one estimate the execution time for force > calculation as a function of number of Hubbard levels in a DFT+U > calculation with ortho-atomic U_projection? > Thank you in advance. > Bests, > Mahmoud > > From: Paolo Giannozzi > mailto:paolo.giannozzi%40uniud.it>> > To: > users@lists.quantum-espresso.org<mailto:users%40lists.quantum-espresso.org> > Date: Tue, 20 Sep 2022 07:58:42 +0200 > Subject: Re: [QE-users] Some inconsistency between QE-6.7MaX and > current QE-7.1 > The optimization you are referring to is actually an automatic choice of > parallelization option, but it is not used if you explicitly specify the > parallelization options (e.g. with -nk N -nd M). For the original > problem: one has to see under which exact coonditions it happens > > Paolo > > On 20/09/2022 05:01, Mahmoud Payami Shabestari via users wrote: > > Dear QE Developers, > > Hi. > > Kindly, when I use the same scf input for QE-6.7MaX and QE-7.1, the > > results of forces on atoms are more or less similar but with a small > > difference. However, the calculated pressures are > SIGNIFICANTLY_DIFFERENT. > > Secondly, when the number of atoms is of order say 50, the > calculation > > of forces on atoms using QE-7.1 is VERY VERY time consuming. > > In the release note of QE7.1 one feature mentioned is an "automatic" > > optimizations for accelerating the calculations. > > Could it happen that in some cases instead of optimizations one > > encounters with deceleration? > > How can one disable this automatic optimization? > > Thank you in advance. > > With Best Regards, > > Mahmoud Payami > > NSTRI, AEOI, Tehran, Iran > > Email: mpay...@aeoi.org.ir<mailto:mpayami%40aeoi.org.ir> > <mailto:mpayami%40aeoi.org.ir> > > Phone: +98 (0)21 82066504 > > > > > > ___ > > The Quantum ESPRESSO community stands by the Ukrainian > > people and expresses its concerns about the devastating > > effects that the Russian military offensive has on their > > country and on the free and peaceful scientific, cultural, > > and economic cooperation amongst peoples > > ___ > > Quantum ESPRESSO is supported by MaX > (www.max-centre.eu<http://www.max-centre.eu> > > <https://
Re: [QE-users] Some inconsistency between QE-6.7MaX and current QE-7.1
Dear Iurii, Thank you for yor reply and referencing the relevant paper. Now the important question: people prefer DFT+U, because in principle the computational cost is of order (not so larger) than that of DFT, because one just introduces a correction from "APPROXIMATE" Hubbard Model on DFT. If one would compare "DFT+U+ortho-atomic" with the "orbital-dependent HSE" method, can one be certain that the latter is still much more computationally expensive than the former? Thank you in advance. Bests, Mahmoud From: Iurii TIMROV via users To: Paolo Giannozzi , Mahmoud Payami Shabestari , Quantum ESPRESSO users Forum Date: Fri, 23 Sep 2022 07:37:57 +0000 Subject: Re: [QE-users] Some inconsistency between QE-6.7MaX and current QE-7.1 Hubbard forces and stresses using "ortho-atomic" Hubbard projectors are much more expensive than with "atomic". It is explained in PRB 102, 235159 (2020). Computational scaling is discussed in the appendix C. Iurii -- Dr. Iurii TIMROV Senior Research Scientist Theory and Simulation of Materials (THEOS) Swiss Federal Institute of Technology Lausanne (EPFL) CH-1015 Lausanne, Switzerland +41 21 69 34 881 http://people.epfl.ch/265334 From: users on behalf of Mahmoud Payami Shabestari via users Sent: Thursday, September 22, 2022 9:03:17 PM To: Paolo Giannozzi; Quantum ESPRESSO users Forum Subject: Re: [QE-users] Some inconsistency between QE-6.7MaX and current QE-7.1 Dear Paolo, Thank you so much for your comments. Actually, to find some sense about the time needed to calculate force, I have performed scf calculations for just one iteration, but different numbers of atoms (4, 8, 12, 16, 32, 48, 64, 80, 96, 108) with more or less the sama number of k points in BZ. The supercells were generated from the QE-7.1/PW/example08 unitcell. The inputs and outputs are uploaded in: https://drive.google.com/drive/folders/12cWLR7PkMI1wolxgRLnOSCi4IhtnLmEO?usp=sharing The results for low number of atoms seems to take reasonable times, but for larger ones the time to calculate force becomes larger than the time for just one iteration. Do you see any un-normal behavior in force calculation routine? Thank you very much for your comments. Bests, Mahmoud From: Paolo Giannozzi To: Mahmoud Payami Shabestari , Quantum ESPRESSO users Forum Date: Thu, 22 Sep 2022 18:18:51 +0200 Subject: Re: [QE-users] Some inconsistency between QE-6.7MaX and current QE-7.1 Not sure what you mean exactly. As a rule of thumb, the calculation of the Hubbard contribution to forces should take no more than 1/3 or 1/4, likely much less than that, of the time needed for a scf calculation, at least for the simple ("atomic") U projection. Paolo On 21/09/2022 19:31, Mahmoud Payami Shabestari via users wrote: > Dear Paolo, > Thank you so much for the comment on automatic parallelization. > On the second issue, how can one estimate the execution time for force > calculation as a function of number of Hubbard levels in a DFT+U > calculation with ortho-atomic U_projection? > Thank you in advance. > Bests, > Mahmoud > > From: Paolo Giannozzi > To: users@lists.quantum-espresso.org > Date: Tue, 20 Sep 2022 07:58:42 +0200 > Subject: Re: [QE-users] Some inconsistency between QE-6.7MaX and > current QE-7.1 > The optimization you are referring to is actually an automatic choice of > parallelization option, but it is not used if you explicitly specify the > parallelization options (e.g. with -nk N -nd M). For the original > problem: one has to see under which exact coonditions it happens > > Paolo > > On 20/09/2022 05:01, Mahmoud Payami Shabestari via users wrote: > > Dear QE Developers, > > Hi. > > Kindly, when I use the same scf input for QE-6.7MaX and QE-7.1, the > > results of forces on atoms are more or less similar but with a small > > difference. However, the calculated pressures are > SIGNIFICANTLY_DIFFERENT. > > Secondly, when the number of atoms is of order say 50, the > calculation > > of forces on atoms using QE-7.1 is VERY VERY time consuming. > > In the release note of QE7.1 one feature mentioned is an "automatic" > > optimizations for accelerating the calculations. > > Could it happen that in some cases instead of optimizations one > > encounters with deceleration? > > How can one disable this automatic optimization? > > Thank you in advance. > > With Best Regards, > > Mahmoud Payami > > NSTRI, AEOI, Tehran, Iran > > Email: mpay...@aeoi.org.ir <mailto:mpayami%40aeoi.org.ir> > > Phone: +98 (0)21 82066504 > > ---
Re: [QE-users] Some inconsistency between QE-6.7MaX and current QE-7.1
> If one would compare "DFT+U+ortho-atomic" with the "orbital-dependent HSE" > method, can one be certain that the latter is still much more computationally > expensive than the former? 1. It depends on the system size. Some comparison between DFT+U (with first-principles and self-consistent U using ortho-atomic orbitals) and HSE06 is made in the Appendix B of this paper: https://www.mdpi.com/2076-3417/11/5/2395 2. Accuracy of DFT+U vs HSE06: in DFT+U, the U parameter can be computed for each system at hand while in HSE06 the fraction of exact exchange is 0.25 for all systems. When people tune the fraction of exact exchange in HSE it is no longer first principles (though there are various methods how to compute it - some references are mentioned in the intro of the paper above). Therefore, it is not only about the speed/scaling but also about the accuracy. Iurii -- Dr. Iurii TIMROV Senior Research Scientist Theory and Simulation of Materials (THEOS) Swiss Federal Institute of Technology Lausanne (EPFL) CH-1015 Lausanne, Switzerland +41 21 69 34 881 http://people.epfl.ch/265334 From: Mahmoud Payami Shabestari Sent: Friday, September 23, 2022 11:26:33 AM To: Iurii TIMROV; Quantum ESPRESSO users Forum; Paolo Giannozzi Subject: Re: [QE-users] Some inconsistency between QE-6.7MaX and current QE-7.1 Dear Iurii, Thank you for yor reply and referencing the relevant paper. Now the important question: people prefer DFT+U, because in principle the computational cost is of order (not so larger) than that of DFT, because one just introduces a correction from "APPROXIMATE" Hubbard Model on DFT. If one would compare "DFT+U+ortho-atomic" with the "orbital-dependent HSE" method, can one be certain that the latter is still much more computationally expensive than the former? Thank you in advance. Bests, Mahmoud From: Iurii TIMROV via users To: Paolo Giannozzi , Mahmoud Payami Shabestari , Quantum ESPRESSO users Forum Date: Fri, 23 Sep 2022 07:37:57 + Subject: Re: [QE-users] Some inconsistency between QE-6.7MaX and current QE-7.1 Hubbard forces and stresses using "ortho-atomic" Hubbard projectors are much more expensive than with "atomic". It is explained in PRB 102, 235159 (2020). Computational scaling is discussed in the appendix C. Iurii -- Dr. Iurii TIMROV Senior Research Scientist Theory and Simulation of Materials (THEOS) Swiss Federal Institute of Technology Lausanne (EPFL) CH-1015 Lausanne, Switzerland +41 21 69 34 881 http://people.epfl.ch/265334 From: users on behalf of Mahmoud Payami Shabestari via users Sent: Thursday, September 22, 2022 9:03:17 PM To: Paolo Giannozzi; Quantum ESPRESSO users Forum Subject: Re: [QE-users] Some inconsistency between QE-6.7MaX and current QE-7.1 Dear Paolo, Thank you so much for your comments. Actually, to find some sense about the time needed to calculate force, I have performed scf calculations for just one iteration, but different numbers of atoms (4, 8, 12, 16, 32, 48, 64, 80, 96, 108) with more or less the sama number of k points in BZ. The supercells were generated from the QE-7.1/PW/example08 unitcell. The inputs and outputs are uploaded in: https://drive.google.com/drive/folders/12cWLR7PkMI1wolxgRLnOSCi4IhtnLmEO?usp=sharing The results for low number of atoms seems to take reasonable times, but for larger ones the time to calculate force becomes larger than the time for just one iteration. Do you see any un-normal behavior in force calculation routine? Thank you very much for your comments. Bests, Mahmoud From: Paolo Giannozzi To: Mahmoud Payami Shabestari , Quantum ESPRESSO users Forum Date: Thu, 22 Sep 2022 18:18:51 +0200 Subject: Re: [QE-users] Some inconsistency between QE-6.7MaX and current QE-7.1 Not sure what you mean exactly. As a rule of thumb, the calculation of the Hubbard contribution to forces should take no more than 1/3 or 1/4, likely much less than that, of the time needed for a scf calculation, at least for the simple ("atomic") U projection. Paolo On 21/09/2022 19:31, Mahmoud Payami Shabestari via users wrote: > Dear Paolo, > Thank you so much for the comment on automatic parallelization. > On the second issue, how can one estimate the execution time for force > calculation as a function of number of Hubbard levels in a DFT+U > calculation with ortho-atomic U_projection? > Thank you in advance. > Bests, > Mahmoud > > From: Paolo Giannozzi > mailto:paolo.giannozzi%40uniud.it>> > To: > users@lists.quantum-espresso.org<mailto:users%40lists.quantum-espresso.org> > Date: Tue, 20 Sep 2022 07:58:42 +0200 > Subject: Re: [QE-users] Some inconsistency between QE-6.7MaX and > current QE-7.1 > The optimization you are referring to is actually an automatic choice o
Re: [QE-users] Some inconsistency between QE-6.7MaX and current QE-7.1
Dear Iurii, Many thanks for your detailed explanation. I am convinced. Best regards, Mahmoud From: Iurii TIMROV To: Mahmoud Payami Shabestari , "Quantum ESPRESSO users Forum" , Paolo Giannozzi Date: Fri, 23 Sep 2022 09:35:42 + Subject: Re: [QE-users] Some inconsistency between QE-6.7MaX and current QE-7.1 > If one would compare "DFT+U+ortho-atomic" with the "orbital-dependent HSE" method, can one be certain that the latter is still much more computationally expensive than the former? It depends on the system size. Some comparison between DFT+U (with first-principles and self-consistent U using ortho-atomic orbitals) and HSE06 is made in the Appendix B of this paper: https://www.mdpi.com/2076-3417/11/5/2395 Accuracy of DFT+U vs HSE06: in DFT+U, the U parameter can be computed for each system at hand while in HSE06 the fraction of exact exchange is 0.25 for all systems. When people tune the fraction of exact exchange in HSE it is no longer first principles (though there are various methods how to compute it - some references are mentioned in the intro of the paper above). Therefore, it is not only about the speed/scaling but also about the accuracy. Iurii -- Dr. Iurii TIMROV Senior Research Scientist Theory and Simulation of Materials (THEOS) Swiss Federal Institute of Technology Lausanne (EPFL) CH-1015 Lausanne, Switzerland +41 21 69 34 881 http://people.epfl.ch/265334 From: Mahmoud Payami Shabestari Sent: Friday, September 23, 2022 11:26:33 AM To: Iurii TIMROV; Quantum ESPRESSO users Forum; Paolo Giannozzi Subject: Re: [QE-users] Some inconsistency between QE-6.7MaX and current QE-7.1 Dear Iurii, Thank you for yor reply and referencing the relevant paper. Now the important question: people prefer DFT+U, because in principle the computational cost is of order (not so larger) than that of DFT, because one just introduces a correction from "APPROXIMATE" Hubbard Model on DFT. If one would compare "DFT+U+ortho-atomic" with the "orbital-dependent HSE" method, can one be certain that the latter is still much more computationally expensive than the former? Thank you in advance. Bests, Mahmoud From: Iurii TIMROV via users To: Paolo Giannozzi , Mahmoud Payami Shabestari , Quantum ESPRESSO users Forum Date: Fri, 23 Sep 2022 07:37:57 +0000 Subject: Re: [QE-users] Some inconsistency between QE-6.7MaX and current QE-7.1 Hubbard forces and stresses using "ortho-atomic" Hubbard projectors are much more expensive than with "atomic". It is explained in PRB 102, 235159 (2020). Computational scaling is discussed in the appendix C. Iurii -- Dr. Iurii TIMROV Senior Research Scientist Theory and Simulation of Materials (THEOS) Swiss Federal Institute of Technology Lausanne (EPFL) CH-1015 Lausanne, Switzerland +41 21 69 34 881 http://people.epfl.ch/265334 From: users on behalf of Mahmoud Payami Shabestari via users Sent: Thursday, September 22, 2022 9:03:17 PM To: Paolo Giannozzi; Quantum ESPRESSO users Forum Subject: Re: [QE-users] Some inconsistency between QE-6.7MaX and current QE-7.1 Dear Paolo, Thank you so much for your comments. Actually, to find some sense about the time needed to calculate force, I have performed scf calculations for just one iteration, but different numbers of atoms (4, 8, 12, 16, 32, 48, 64, 80, 96, 108) with more or less the sama number of k points in BZ. The supercells were generated from the QE-7.1/PW/example08 unitcell. The inputs and outputs are uploaded in: https://drive.google.com/drive/folders/12cWLR7PkMI1wolxgRLnOSCi4IhtnLmEO?usp=sharing The results for low number of atoms seems to take reasonable times, but for larger ones the time to calculate force becomes larger than the time for just one iteration. Do you see any un-normal behavior in force calculation routine? Thank you very much for your comments. Bests, Mahmoud From: Paolo Giannozzi To: Mahmoud Payami Shabestari , Quantum ESPRESSO users Forum Date: Thu, 22 Sep 2022 18:18:51 +0200 Subject: Re: [QE-users] Some inconsistency between QE-6.7MaX and current QE-7.1 Not sure what you mean exactly. As a rule of thumb, the calculation of the Hubbard contribution to forces should take no more than 1/3 or 1/4, likely much less than that, of the time needed for a scf calculation, at least for the simple ("atomic") U projection. Paolo On 21/09/2022 19:31, Mahmoud Payami Shabestari via users wrote: > Dear Paolo, > Thank you so much for the comment on automatic parallelization. > On the second issue, how can one estimate the execution time for force > calculation as a function of number of Hubbard levels in a DFT+U > calculation with ortho-atomic U_projection? > Thank you in advance. > Bests, > Mahmoud > > From: Paolo Giannozzi > To: users@lists.quantum-espresso.org > Date: Tue, 20 Sep 2022 07:58:42 +0200 >