Dear Paolo Thank you for reply. I didn't say that when I use 1 node each have 32 cores , I don't have any problem with run of pw1.in and pw2.in wiith nbnd=1024. But once I increase? the number of proccessors to 32 node each have 32 core this error happen.
Best regards Ali Kazempour Fritz-Haber-Institut fax : ++49-30-8413 4701 der Max-Planck-Gesellschaft Faradayweg 4-6 e-mail: kazempou at fhi-berlin.mpg.de D-14 195 Berlin-Dahlem / German --- On Thu, 6/10/10, Paolo Giannozzi <giannozz at democritos.it> wrote: From: Paolo Giannozzi <[email protected]> Subject: Re: [Pw_forum] error in gamma To: "PWSCF Forum" <pw_forum at pwscf.org> Date: Thursday, June 10, 2010, 4:44 AM kazempoor at ph.iut.ac.ir wrote: >? ? ? from? pzpotrf? : error #? ? ? ? 47 >? ? ???problems computing cholesky decomposition > > This is very strange because when I set nbnd=512 > the pw1.in and pw2.in run without any problem . it is not strange. The diagonalization algorithm in PWscf is expected to work for a number of bands that is much smaller than the dimension of the basis set. It usually works great if you calculate occupied bands and a few unoccupied bands. The more bands you add, the more you are likely to run into numerical trouble. The error here occurs in a Scalapack routine (pzpotrf). Finding out why and how is a highly nontrivial task Paolo -- Paolo Giannozzi, Democritos and University of Udine, Italy _______________________________________________ Pw_forum mailing list Pw_forum at pwscf.org http://www.democritos.it/mailman/listinfo/pw_forum -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20100610/7e4f37cf/attachment.htm
