Top Page > Browsing

performance against Siesta, any impression? Date: 2011/05/25 00:23 Name: David Dear All, has anybody compared the (serial/parallel) performance of the Openmx3.5 against Siesta? Is there any major differences in speed for example?

Page: [1]

Re: performance against Siesta, any impression? ( No.1 ) Date: 2011/05/25 11:47 Name: T.Ozaki Hi, As long as I know, there is no systematic comparison between two codes with respect to computational (parallel) efficiency. If you are interested in that, please report your comparison to share the information among users. Regards, TO Re: performance against Siesta, any impression? ( No.2 ) Date: 2011/05/29 03:43 Name: David Dear Ozaki, for better comparison can you tell me how to speed up the following calculations, as much as i can. So far the same system is calculated faster with siesta (both compiled with gcc, and siesta uses its built in lapack) # Definition of Atomic Species Species.Number 1 <Definition.of.Atomic.Species Au Au6.0-s3p3d2 Au_PBE # DZP? Definition.of.Atomic.Species> # Atoms Atoms.Number 4 Atoms.SpeciesAndCoordinates.Unit Ang # Ang|AU|FRAC <Atoms.SpeciesAndCoordinates 1 Au 0.000000 0.000000 0.000000 9.5 9.5 2 Au 2.500000 0.000000 0.000000 9.5 9.5 3 Au 5.000000 0.000000 0.000000 9.5 9.5 4 Au 7.500000 0.000000 0.000000 9.5 9.5 Atoms.SpeciesAndCoordinates> Atoms.UnitVectors.Unit Ang # Ang|AU <Atoms.UnitVectors 10.0000 0.000000 0.00000 0.000000 20.000000 0.000000 0.000000 0.000000 20.000000 Atoms.UnitVectors> # SCF or Electronic System scf.XcType GGA-PBE # LDA|LSDA-CA|LSDA-PW|GGA-PBE scf.SpinPolarization off # On|Off|NC scf.ElectronicTemperature 300.0 # default=300 (K) scf.energycutoff 120.0 # default=150 (Ry) scf.maxIter 90 # default=40 scf.EigenvalueSolver Krylov # DC|GDC|Cluster|Band|Krylov scf.Kgrid 1 1 1 # means n1 x n2 x n3 scf.Mixing.Type Rmm-Diis # Simple|Rmm-Diis|Gr-Pulay|Kerker|Rmm-Diisk scf.Init.Mixing.Weight 9.30 # default=0.30 scf.Min.Mixing.Weight 0.004 # default=0.001 scf.Max.Mixing.Weight 0.400 # default=0.40 scf.Mixing.History 7 # default=5 scf.Mixing.StartPulay 6 # default=6 scf.criterion 1.0e-4 # default=1.0e-6 (Hartree) scf.lapack.dste dstevx # dstevx|dstedc|dstegr,default=dstevx Re: performance against Siesta, any impression? ( No.3 ) Date: 2011/05/29 21:42 Name: T.Ozaki Hi, I guess that the speed up ratio in the parallel computation may be limited to around 2 times for such a small system consisting of 4 atoms. Best regards, TO

Page: [1]