[SIESTA-L] empty output in plstm
Dear SIESTA users, I am trying to make constant-height image of Si(111)-7x7 (not rectangular unit cell). After running plstm I got file with "0" bytes. The output of plstm is below. The values of cell vectors are strange. What can be the problem? Ruslan sie...@hpc-suvir1:~/siesta-2.0.2/Util/my-stm> ~/siesta-2.0.2/Util/my-stm/plstm < input ** Calculating STM image in Constant Height mode The STM image is obtained as the value of the charge at a given tip height Z =1.00 Bohr ** Reading grid data from file 7x7-DM.LDOS Cell vectors 5.182087913350189E+165 5.328740546986013E-315 0.000E+000 1.110537373234340E+242 -2.148645586157028E+218 5.327640732155088E-315 0.000E+000 0.000E+000 32387350528. Grid mesh:0 x 16 x 0 nspin = 200 Writing STM image in file7x7-DM.CH.STM
Re: [SIESTA-L] memory requirement
Hi walter, Siesta "eats" a lot of RAM, so the next time that you buy computers buy as much RAM as you can (at least more than 2gb per core). Your can reduce the amount of memory required using the options: DirectPhi .true. and SaveMemory .true. You save some memory at the expense of CPU time. Regarding, your previous mail today. if you have problems with mpi. you can use these options to use more "cores" when you compile: use FFLAGS= -w -O3 -mp1 -ax -DWXML_INIT_FIX and in you csh.cshrc include setenv OMP_NUM_THREADS 4 setenv MKL_NUM_THREADS 4 you can change 4 to 2 ,3 etc... the efficience depends on the calculation that you are running sometimes it is 400% 300% or 100% maybe the other users can help to clarify this. Best regards, pablo - Original Message - From: Walter Cañón To: SIESTA-L@listserv.uam.es Sent: Friday, May 15, 2009 5:35 PM Subject: [SIESTA-L] memory requirement Hi, again i have another question if i want to make calculation in system bigger as a cluster with 10 or more metalic centers like [Mn12O12(CH3CO2)16(H2O)4], you think that with a 1 terabyte hard disk of memory and 4 Gb of ram will be enough to run the calculation, if not, you can suggest me a practical configuration to these system. Thank you for your helpfull comments best regards to all. Walter Walter Alberto Cañón Mancisidor Químico Doctorado en Química Laboratorio de Fisicoquímica Inorgánica Facultad de Ciencias Químicas y Farmaceuticas Universidad de Chile wca...@ciq.uchile.cl Tel 56-2-9782987
Re: [SIESTA-L] Quad-core question?
Thanks Marcel. bye Marcel Mohr escribió: On Fri, 15 May 2009, Walter Cañón wrote: Hi, my name is Walter and i download recently the latest version of SIESTA. i know that the compilation to run siesta in parallel is quite difficult, so i want to know if i have a pc with two quad-core processors, i have to compile in parallel the two quad-core processors. I would like to compile some processors, too ;) But to use these 8 cores you have too compile in parallel and use mpiexec to run siesta. Cheers Marcel Thank you for your helpfull comments best regards to all. Walter Walter Alberto Cañón Mancisidor Químico Doctorado en Química Laboratorio de Fisicoquímica Inorgánica Facultad de Ciencias Químicas y Farmaceuticas Universidad de Chile wca...@ciq.uchile.cl Tel 56-2-9782987
Re: [SIESTA-L] memory requirement
Hi Walter It depends on the basis set used, Mesh Cutoff and the size of your supercell. I used siesta for the simulation of an organic-metallic system with 4 Fe centers: dzp & mesh 300 Ry. It takes 3Gb RAM. So, I think that you will need more than 4 Gb. best regards Ricardo Dr. Ricardo Faccio Prof. Adj. de Física Av. Gral. Flores 2124. CC 1157. CP 11800. Phone: + 598 2 924 9859 Fax: + 598 2 924 1906 Web: http://cryssmat.fq.edu.uy/ricardo/ricardo.htm - El 15/05/2009, a las 17:35, Walter Cañón escribió: Hi, again i have another question if i want to make calculation in system bigger as a cluster with 10 or more metalic centers like [Mn12O12(CH3CO2)16(H2O)4], you think that with a 1 terabyte hard disk of memory and 4 Gb of ram will be enough to run the calculation, if not, you can suggest me a practical configuration to these system. Thank you for your helpfull comments best regards to all. Walter Walter Alberto Cañón Mancisidor Químico Doctorado en Química Laboratorio de Fisicoquímica Inorgánica Facultad de Ciencias Químicas y Farmaceuticas Universidad de Chile wca...@ciq.uchile.cl Tel 56-2-9782987
[SIESTA-L] memory requirement
Hi, again i have another question if i want to make calculation in system bigger as a cluster with 10 or more metalic centers like *[Mn_12 O_12 (CH_3 CO_2 )_16 (H_2 O)_4 ]***, you think that with a 1 terabyte hard disk of memory and 4 Gb of ram will be enough to run the calculation, if not, you can suggest me a practical configuration to these system. Thank you for your helpfull comments best regards to all. Walter Walter Alberto Cañón Mancisidor Químico Doctorado en Química Laboratorio de Fisicoquímica Inorgánica Facultad de Ciencias Químicas y Farmaceuticas Universidad de Chile wca...@ciq.uchile.cl Tel 56-2-9782987
[SIESTA-L] help fiel for make.arch?
Just to make sure, is there a help file that describes in detail all the options and flags in make.arch file for a semiexperienced user? thanks JOnas
[SIESTA-L] orderN calculation stability
Hello, I have recently started to use siesta with a view to doing orderN calculations. So far all the calculations I have tried have become unstable in the first or second steps regardless of whether I manually tune the value of eta, set it to the value from a previous diagonalised calculation or use the chemical potential method (NB I would not expect the band gap to be small for these systems and the fermi level is about -3.7 eV). So I have a few questions which I have not been able to find answers for to see if I am doing something daft or impossible (Apologies if I've missed something obvious - I haven't studied DFT methods in great detail and I'm clutching at straws to find a way to optimise a particular set of crystal structures using DFT). i) I'm using the GGA potentials that were converted from AbInit - are these compatible with the orderN method ? ii) The system is expected to contract quite a lot as it is a crystal structure with the solvent removed - is it a really bad idea to try an relax such a system using the orderN method ? iii) The gradients in output file always seem to be very large compared to those I've seen in presentations (even though the number of electrons is not too bad at the end of the first cycle) - does this indicate from the outset that the method is unlikely to work with a particular structure ? Thank you for any help or suggestions, Jeremy
Re: [SIESTA-L] Quad-core question?
On Fri, 15 May 2009, Walter Cañón wrote: Hi, my name is Walter and i download recently the latest version of SIESTA. i know that the compilation to run siesta in parallel is quite difficult, so i want to know if i have a pc with two quad-core processors, i have to compile in parallel the two quad-core processors. I would like to compile some processors, too ;) But to use these 8 cores you have too compile in parallel and use mpiexec to run siesta. Cheers Marcel Thank you for your helpfull comments best regards to all. Walter Walter Alberto Cañón Mancisidor Químico Doctorado en Química Laboratorio de Fisicoquímica Inorgánica Facultad de Ciencias Químicas y Farmaceuticas Universidad de Chile wca...@ciq.uchile.cl Tel 56-2-9782987
