Samtale med #gpaw

(09:39:21) Emne for #gpaw er: GPAW | https://wiki.fysik.dtu.dk/gpaw | https://wiki.fysik.dtu.dk/ase | https://trac.fysik.dtu.dk/projects/gpaw | Paste code on http://paste.pocoo.org/ | Paste math on http://mathbin.net/
(11:48:22) aheld84 [~aheld@remote239-051.home.uni-freiburg.de] trådte ind i rummet.
(11:50:00) aheld84 forlod rummet.
(12:20:54) petarj [~Adium@sclwifi.ipb.ac.rs] trådte ind i rummet.
(12:30:35) arioj [82e15696@gateway/web/freenode/ip.130.225.86.150] trådte ind i rummet.
(12:30:52) chlg: arioj: Hi Ari
(14:10:47) petarj forlod rummet (quit: Quit: Leaving.).
(14:55:01) petarj [~Adium@sclwifi.ipb.ac.rs] trådte ind i rummet.
(15:52:48) peter_klue [~peterklue@vr3pc07.rhi.hi.is] trådte ind i rummet.
(15:56:28) lauri_lehtovaara [~lauri@eduroam154.univ-lyon1.fr] trådte ind i rummet.
(16:14:36) aheld84 [~aheld@remote239-051.home.uni-freiburg.de] trådte ind i rummet.
(16:15:35) aheld84 forlod rummet.
(16:16:55) alexander_held [~aheld@remote239-051.home.uni-freiburg.de] trådte ind i rummet.
(16:17:41) fabiolima [c881ca82@gateway/web/freenode/ip.200.129.202.130] trådte ind i rummet.
(16:20:19) Falco1 [~falh@planck.fysik.dtu.dk] trådte ind i rummet.
(16:28:10) naromero [~Adium@alcfwl144.alcf.anl.gov] trådte ind i rummet.
(16:28:14) rowue_ [~rowue@bone.digitalis.org] trådte ind i rummet.
(16:28:17) rowue_: re
(16:29:48) shk2 [~simon@vr3pc07.rhi.hi.is] trådte ind i rummet.
(16:31:11) jjmo1: I think it's time to start
(16:31:46) rowue_: I second this - how we should start? - Btw: Hello everybody
(16:31:46) askhl_: Who are people? I'm Ask Hjorth Larsen
(16:31:59) askhl_ kalder sig nu for askhl
(16:32:02) ***rowue_ is Rolf Würdemann
(16:32:14) naromero: Nichols A. Romero
jjmo jjmo1
(16:32:20) lauri_lehtovaara: Lauri Lehtovaara
(16:32:21) peter_klue: Peter J. Kluepfel
(16:32:24) arioj: Ari Ojanperä
(16:32:25) alexander_held: Alexander Held
(16:32:34) jjmo1: jjmo1: Jens Jørgen Mortensen
(16:32:38) Falco1: Falco Hueser
(16:32:48) lauri_lehtovaara: How about again with the institution?
(16:32:56) lauri_lehtovaara: Lauri Lehtovaara, University of Lyon 1
(16:32:58) askhl: Ask Hjorth Larsen, CAMD
(16:33:03) peter_klue: Peter J. Kluepfel (University of Iceland)
(16:33:04) rowue_: Rolf Würdemann, FMF
jjmo jjmo1
(16:33:21) arioj: Ari Ojanperä, Aalto University
(16:33:23) Falco1: Falco Hueser, CAMD
(16:33:23) jjmo1: jjmo1: Jens Jørgen Mortensen, CAMd
(16:33:25) shk2: Simon Kl?pfel (University of Iceland)
(16:33:36) alexander_held: Alexander Held, FMF Uni Freiburg, Germany, Phd student of Michael Walter and Michael Moseler
(16:34:35) jjmo1: Maybe the three TDDFT guys could start telling us what they are working on right now
(16:35:09) jjmo1: Then we could all ask questions and then move on to next project?
(16:35:16) lauri_lehtovaara: ok
(16:35:24) petarj forlod rummet.
(16:35:36) lauri_lehtovaara: but I'm not doing anything actively so...
(16:35:45) lauri_lehtovaara: ari?
(16:36:35) jjmo1: I know that you are working on merging the Ehrenfest branch to trunk - right?
(16:36:43) jjmo1: chlg: Are you here?
(16:37:05) arioj: Yes, exactly.
(16:37:35) lauri_lehtovaara: Ehrenfest molecular dynamics is TDDFT combined with classical dynamics for nuclei
(16:37:48) arioj: The work is almost finished - I would say we can merge the branch back to trunk within a week or so.
(16:38:28) jjmo1: Sounds good. Is this something ready for production for everybody?
(16:38:32) lauri_lehtovaara: Do you have some documentation how to use?
(16:39:46) arioj: Not yet, we definitely have to add some docs before it's ready for production use. Currently the code is difficult to use for anybode other than me and Christian.
(16:40:12) jjmo1: A tutorial or something would be nice. Shall we move on?
(16:40:21) lauri_lehtovaara: yea
(16:40:36) jjmo1: peter_klue: How are you doing?
(16:40:39) lauri_lehtovaara: let's try to keep these rolling quite fast
(16:40:47) peter_klue: great
(16:40:48) lauri_lehtovaara: what
(16:40:52) lauri_lehtovaara: 's the project?
(16:41:01) jjmo1: ODD
(16:41:02) peter_klue: self-consistent SIC
(16:41:08) lauri_lehtovaara: ODD?
(16:41:13) peter_klue: and other orbital-density dependent functionals
(16:41:16) peter_klue: ODD
(16:42:04) peter_klue: I'm currently working on OEP to SIC
(16:42:11) peter_klue: and consistent PAW potentials
(16:42:16) lauri_lehtovaara: will we also have TD-OEP/KLI?
(16:42:30) peter_klue: eventually, but not within the next year.
(16:42:33) lauri_lehtovaara: ok
(16:42:46) lauri_lehtovaara: i might have to work on this stuff during next year
(16:42:49) peter_klue: but I'd like to do that at some point
(16:43:07) lauri_lehtovaara: we have to talk when the time comes
(16:43:26) peter_klue: sure. But we are currently disconnected from the GPAW main code
(16:43:39) lauri_lehtovaara: ok, then we have to definitely talk :)
(16:44:04) jjmo1: Is there a chance that the two codes could be merged?
(16:44:50) peter_klue: yes, we build our code on the last GPAW release.
(16:45:18) peter_klue: so once we are done, we can merge again.
(16:45:28) jjmo1: Great. Should I say what I am working on or am I moving too fast?
(16:45:43) peter_klue: no fast is good.
(16:45:46) lauri_lehtovaara: agreed
(16:45:51) lauri_lehtovaara: at least for the first time
(16:46:04) naromero: yes, go ahead and tell us about the your work jjmo1
(16:46:13) jjmo1: I'm trying to add a planewave basis to GPAW
(16:46:34) jjmo1: The goal is to have something that is efficient for small systems
(16:46:50) lauri_lehtovaara: what is small? 10 or 100 atoms?
(16:47:06) jjmo1: And it will also be simpler to implement calculation of the stress tensor, which I would like to have soon
(16:47:30) jjmo1: 10-50 atoms - I think.
(16:47:34) naromero: have we given up on the stress tensor for the real-space code? Is it too hard to derive.
(16:47:49) naromero: PWs codes are super fast for small systems.
(16:47:56) jjmo1: I'm also working together with Marcin to make new PAW-setups
(16:48:40) jjmo1: For testing the new setups I do lots of calculations with only one atom!
(16:49:22) askhl: So do those calculations run faster presently with the planewave implementation?
(16:49:31) jjmo1: Some of the derivatives for stress are nasty in real-space, but very simple in reciprocal space
(16:49:57) rowue_: jjmo1: should we check out for the scalar-relativistic corrections
(16:50:01) jjmo1: So, my plan is to implement stress first with plane waves
(16:50:16) naromero: I can believe that. I have yet to see any real-space code with a stress tensor. Even with just norm-conserving pseudopotentials.
(16:50:45) jjmo1: Yes, they run faster already, but there are some more optimizations to do
(16:51:42) jjmo1: rowue_: Maybe you could say what you are working on?
(16:52:19) rowue_: had worked on the projector and grid issues as mentioned on the ML,
(16:52:35) ***chlg is reading scollback
(16:52:49) rowue_: now hunting for the error in sc-rel. setup, where you mentioned the sc-rel. corrections
(16:53:08) jjmo1: I'm hoping those corrections are not important!
(16:53:13) rowue_: if it's o.k. by you I would try to implement this in the old generator code and give you feedback
(16:53:24) rowue_: jjmo1: /me too
(16:53:28) jjmo1: That would be great!
(16:53:41) rowue_: jjmo1: then we go like this!
(16:53:55) jjmo1: chlg: Did you have anything to add about TDDFT?
(16:54:20) chlg: jjmo1: Not really. Ari covered everything I think
(16:54:31) chlg: we hope to have a merge by the end of the week
(16:54:53) jjmo1: And documentation next week :)
(16:55:18) jjmo1: askhl: What are you doing these days?
(16:55:20) chlg: Yeah that would probably be a good idea
(16:55:37) jjmo1: Apart from working on you Ph.d. defense
(16:55:44) askhl: I am optimizing the LCAO force expression for large systems.
(16:56:13) askhl: A number of force terms have never really been written for very large matrices, so a large amount of memory will be saved, and the calculations should also run a lot faster
(16:56:27) askhl: This hopefully will be done in a few days unless something nasty pops up
(16:56:39) askhl: After this there's a nonzero probability that I can work on the stresses
(16:56:54) jjmo1: Stress for LCAO?
(16:57:05) askhl: Also, I'm doing to make a debian package for ASE (and later GPAW) for the debian open science project in the near future
(16:57:35) askhl: stresses in the PW code. (Which we talked a bit about the other day)
(16:57:35) lauri_lehtovaara: I think that a very good idea
(16:58:00) jjmo1: alexander_held: Can you tell us about your work?
(16:58:16) askhl: (I'm done mentioning things I work on)
(16:58:25) alexander_held: I'm implementing a continuum solvent model
(16:58:39) alexander_held: this involves creating a new poisson solver
(16:58:46) jjmo1: What would you like to use that for?
(16:59:04) alexander_held: we are investigating ionic liquids e.g.
(16:59:15) alexander_held: and the bulk liquid could be modeled this way
(16:59:54) alexander_held: by introducing a dielectric constant
(16:59:59) jjmo1: What kind of liquid? Water?
(17:00:31) alexander_held: water would also work, but we think of ionic liquids as molten salts
(17:01:26) alexander_held: one could also calculate solvent free energies in water e.g.
(17:01:40) jjmo1: Interesting.
(17:01:48) alexander_held: here is the reference: http://jcp.aip.org/resource/1/jcpsa6/v131/i17/p174108_s1
(17:02:22) alexander_held: it has been implemented in quantum espresso, I think
(17:02:24) jjmo1: Elvar, who is not here, is working on QMMM, where the MM is a simple TIP3P model of water.
(17:02:52) alexander_held: good to know
(17:02:55) jjmo1: Don't know if this would be of interest for anyone
(17:03:25) jjmo1: naromero: I see you are working on a paper?
(17:03:43) naromero: yes, I finally gotten back to working on the GPAW HPC paper.
(17:04:02) naromero: It mostly describes our efforts in getting GPAW to scale on Blue Gene/P.
(17:04:29) naromero: I will attempt to finish it in the next couple of months if I don't get interrupted by the GPAW users at ANL.
(17:04:31) naromero: :)
(17:04:53) jjmo1: Falco1: Can you say a few words about GW?
(17:05:30) Falco1: it's working and I will upload a new version in the next days
(17:05:33) lauri_lehtovaara: real-space or LCAO?
(17:05:40) Falco1: I'll also write a documentation very soon
(17:05:43) Falco1: real-space
(17:05:54) lauri_lehtovaara: how large systems can you do?
(17:06:02) lauri_lehtovaara: is G0W0 or something fancier?
(17:06:08) Falco1: G0W0
(17:06:24) Falco1: it's still rather slow, so I'll have some optimization to do
(17:07:00) Falco1: right now, I'm only calculating for a few atoms per unit cell
fabiolima Falco1
(17:07:48) Falco1: once the planewave basis is available I'd like to use that
(17:07:57) Falco1: that will make it a lot faster
(17:08:58) jjmo1: Yes, that's another reason I'm working on that right now.
(17:09:21) jjmo1: kuismam_: fabiolima: janneb: marcin_dulak: shk2: Do you have anything to add?
(17:10:31) shk2: should one report small problems that show up sometimes at all, or is the focus more on implementing the new stuff?
(17:10:44) shk2: Could not get NO converged in LDA, when I tried once
(17:11:00) shk2: so I stopped using GPAW for now, was a bad first impression
(17:11:16) lauri_lehtovaara: It fits better to mailing list i guess
(17:11:24) naromero: I agree
(17:11:37) lauri_lehtovaara: I'm sure it can be solved
(17:11:40) shk2: good. then nothing to add
(17:11:49) jjmo1: Yes, I think there are people on gpaw-users that can help with NO
(17:12:05) lauri_lehtovaara: just one thing... I'm moving to Jyväskylä
(17:12:13) jjmo1: But it's also OK to ask here
(17:12:24) lauri_lehtovaara: and probably start doing some gpaw development again
(17:12:45) jjmo1: When are you moving?
(17:12:49) lauri_lehtovaara: 10.4.
(17:13:12) lauri_lehtovaara: I'll start working on metal clusters with organic ligands
(17:13:17) jjmo1: OK
(17:13:28) lauri_lehtovaara: TDDFT, electron dynamics
(17:13:41) jjmo1: Anything else we should talk about today?
(17:13:43) naromero: I have one thing to add about the HDF5.
(17:13:52) jjmo1: Go ahead
(17:14:44) naromero: The interface is quite robust and very portable. It is well-tested with the real-space code, but not very much with LCAO and TDDFT. I encourage people to test it and report bugs to the list.
(17:14:58) naromero: that's all
(17:15:05) askhl: Should we make a release of GPAW soon? Or is there too much going on
(17:15:31) naromero: It would be nice to get the set-ups finished before a new release.
(17:15:54) jjmo1: This could take time ...
(17:16:37) askhl: I would like a release reasonably soon (~new year) if possible
(17:16:38) jjmo1: Maybe a release right before new setups
(17:16:46) naromero: sure
(17:16:49) jjmo1: We can do that
(17:16:57) naromero: make that a 0.8.2 and then do 0.9 for the new set-ups
(17:17:37) jjmo1: What do you all think about our little meeting?
(17:17:51) jjmo1: When should we have the next one?
(17:17:58) lauri_lehtovaara: 10.1.2012?
(17:18:02) lauri_lehtovaara: or sooner?
(17:18:04) jjmo1: Who writes a summary?
(17:18:20) lauri_lehtovaara: I think it was very useful... at least for me.
(17:18:23) fabiolima: Nothing to accrescent
(17:18:24) askhl: Maybe we should meet before the next release
(17:18:31) askhl: to make sure things are in order and so on
(17:18:39) rowue_: perhaps once/month?
(17:18:40) askhl: or that can be done over mail, as well of course
(17:19:18) naromero: my one recommendation for the next meeting is that I enjoy hearing all the technical details, maybe have the next meetings focused on particular sub-set of topics. For example, perhaps just TDDFT.
(17:19:37) naromero: give people time to ask lots of questions
(17:19:44) naromero: just a thought
(17:19:52) naromero: once/month seems good
(17:20:04) askhl: I think this is a good way to have meetings, because it's not real-time. So people don't have to sit completely still all the time :). I'm up for a meeting in a month or so
(17:20:04) arioj: Agreed.
(17:20:29) askhl: A video meeting would be much too chaotic
(17:20:34) jjmo1: How about 20.12.2011?
(17:20:42) lauri_lehtovaara: ok
(17:20:47) askhl: Sounds fine
(17:21:02) chlg: if anything, a voice/video conference would be much more well-suited for a handfull of people discussing details, I think
(17:21:08) rowue_: perhaps in the structure main-topic/subtopics - the overview is also very valuable IMHO
(17:21:08) naromero: I will be on vacation, but don't let that stop you from meeting.
(17:21:15) rowue_: 20.12. would be o.k. by me
(17:21:21) alexander_held: 20.12. is fine for me too
(17:21:23) jjmo1: Should I write a summary to the ML's
(17:21:25) lauri_lehtovaara: I would like to have an overview of the projects everytime.
(17:21:40) rowue_: jjmo: if you would do this, it would be nice ;)
(17:21:50) fabiolima: accrescent gostaria of chamada do na um module error: from ase.calculators.jacapo.utils import bandstructure in version gapw of repository ubuntu
(17:22:18) fabiolima: Sorry would add an error in the call module from ase.calculators.jacapo.utils import bandstructure in version gapw of repository ubuntu
(17:22:43) askhl: Log file can be found here: http://www.student.dtu.dk/~ashj/opendir/gpawmeeting.log
(17:23:09) jjmo1: Excellent, I'll write a summary.
(17:23:12) chriso [4c0e301a@gateway/web/freenode/ip.76.14.48.26] trådte ind i rummet.
(17:23:30) lauri_lehtovaara: ok, great... are we more or less done?
(17:24:40) lauri_lehtovaara: Well, I'll have to go. Thanks, and bye.
(17:24:49) jjmo1: I'm gonna leave now, but I will see if someone talks to me on IRC, so just do that any time and I will try to answer any questions.
(17:25:11) lauri_lehtovaara forlod rummet (quit: Quit: lauri_lehtovaara).
(17:25:33) rowue_: o.k. - will leave also - cu
(17:25:35) askhl: Good bye
(17:25:35) Falco1: see you
(17:25:45) rowue_ forlod rummet (quit: Quit: leaving).
(17:25:50) alexander_held: see you
(17:25:52) Falco1 forlod rummet.
(17:25:54) alexander_held forlod rummet.
(17:25:59) peter_klue: bye
(17:26:03) peter_klue forlod rummet (quit: Quit: peter_klue).