Re: [SIESTA-L]
If I am not mistaken, the guide is written for intel-based cpus, but for sure it could be applicable for others. ESSL is ibm product for ibm-based machines only. in fact, only part of lapack is inlcluded; moreover, some lapack functions are different from ones in netlib distribution. On 11/17/06, Vasilii Artyukhov <[EMAIL PROTECTED]> wrote: Well, Sebastian's guide is pretty much all you need. Just an additional remark: check if you really need to install the libs. For instance, your cluster may have the ESSL package installed, which includes the most widely used math libraries, in particular, LAPACK and BLAS. If this is the case, you should use the available optimized libraries. 2006/11/15, Jess Kondor <[EMAIL PROTECTED]>: > > On 11/15/06, Vasilii Artyukhov < [EMAIL PROTECTED]> wrote: > > > > RTFM! Everything is said in the corresponding readme's of the > > libraries. First, you should install MPI, then ATLAS (it is required for the > > following), then MPIBLACS and SCALAPACK. Then, in theory, you can go for > > SIESTA. > > > > Don't forget to specify the "Add_" option when compiling the libraries > > and to configure the SIESTA makefile with --enable-MPI. > > > > In fact, you need also LAPACK (ATLAS contains only BLAS and small part > of LAPACK). Then, playing with different options of BLACS, you can get a > parallel version of SIESTA (in practice!). > > BTW, such strong recommendation as RTFM suggests much more expertise in > the compilation than the user has demonstrated. > > Good luck, > JK >
Re: [SIESTA-L]
Well, Sebastian's guide is pretty much all you need. Just an additional remark: check if you really need to install the libs. For instance, your cluster may have the ESSL package installed, which includes the most widely used math libraries, in particular, LAPACK and BLAS. If this is the case, you should use the available optimized libraries. 2006/11/15, Jess Kondor <[EMAIL PROTECTED]>: On 11/15/06, Vasilii Artyukhov <[EMAIL PROTECTED]> wrote: > > RTFM! Everything is said in the corresponding readme's of the libraries. > First, you should install MPI, then ATLAS (it is required for the > following), then MPIBLACS and SCALAPACK. Then, in theory, you can go for > SIESTA. > > Don't forget to specify the "Add_" option when compiling the libraries > and to configure the SIESTA makefile with --enable-MPI. > In fact, you need also LAPACK (ATLAS contains only BLAS and small part of LAPACK). Then, playing with different options of BLACS, you can get a parallel version of SIESTA (in practice!). BTW, such strong recommendation as RTFM suggests much more expertise in the compilation than the user has demonstrated. Good luck, JK
Re: [SIESTA-L] Au 6s pseudopotential
I believe that if you need some meaningful results, you should include both the semicore states and nonlinear core correction. As they are related to different things, you cannot just substitute one for the other. Well, you can, of course, but I doubt that it will do you any good. 2006/11/16, Marcos Verissimo Alves <[EMAIL PROTECTED]>: I didn't come to my office yesterday, so I am entering this discussion a bit late. Shouldn't it be possible to treat semicore states by including nonlinear core corrections as well? I know that it can be done for Pb, for example, in which you have to include 5d states as well (although I have not particularly succeeded at it, I know other siesta users - maybe Marivi Fernandez, but I am not sure - have). Is there any reason not to try nlcc instead of this "drastic measure" (to say so, since it can increase the computational burden a lot)? Marcos > You have to include the Au semicore states into your calculation if you > want > to get some meaningful results. It is useless to do calculations for the 6s > only, as 5d electrons also participate in the bonding and cannot be treated > as core (which is what you're trying to do). > > 2006/11/15, Haiying He <[EMAIL PROTECTED]>: >> >> Dear siesta users, >> >> I am working on a system including a large number of Au atoms (in bulk >> configuration). Does anyone have an Au pseudopotential with only 6s >> electron in the valence? >> >> I have tried with tm2 (improved Troullier-Martins), but without success >> so >> far. I cannot get a pseudopotential reproducing the equilibrium lattice >> constant of Au bulk. The total energy of the Au bulk continued to >> decrease >> far below the experimental equilibrium lattice constant. I noticed the >> generated pseudopotential is too "attractive", but I don't know how to >> overcome this. Any suggestion will also be greatly appreciated. >> >> Thanks in advance. >> >> Best regards >> >> Haiying >> >> >> >> >> >> >> -- >> >> >> Haiying He >> Department of Physics >> Michigan Technological University >> Houghton, MI 49931 >> > -- Dr. Marcos Verissimo Alves Post-Doctoral Fellow Condensed Matter and Statistical Physics Sector International Centre for Theoretical Physics Trieste, Italy I have become so addicted to vi that I try to exit OpenOffice by typing :wq!