[CPMD-list] Diffrent energy results

Alessandro Contini alessandro.contini at unimi.it
Fri Feb 7 09:17:14 CET 2003


Hi,
I compiled both CPMD 3.5 and 3.7 with options IFC and IFC-MPI on a Dual ATHLON 
MP 1800+ workstation running Linux. I run some test jobs and I compared 
results with those obtained with a CPMD 3.5 executable compiled with pgf.
Here I report a summary of the logs:
3.5 IFC
------------------------------------------------------------------------------------
(K+E1+L+N+X)           TOTAL ENERGY =          -16.62003067 A.U.
 (K)                  KINETIC ENERGY =           11.83136464 A.U.
 (E1=A-S+R)     ELECTROSTATIC ENERGY =          -11.60976118 A.U.
 (S)                           ESELF =           12.63317171 A.U.
 (R)                             ESR =            0.75871391 A.U.
 (L)    LOCAL PSEUDOPOTENTIAL ENERGY =          -14.28070608 A.U.
 (N)      N-L PSEUDOPOTENTIAL ENERGY =            1.41910406 A.U.
 (X)     EXCHANGE-CORRELATION ENERGY =           -3.98003211 A.U.
          GRADIENT CORRECTION ENERGY =           -0.13742377 A.U.
 
-----------------------------------------------------------------------------------
 TOTAL TIME                            522.71              526.67
 ****************************************************************

       CPU TIME :    0 HOURS  8 MINUTES 47.29 SECONDS     
   ELAPSED TIME :    0 HOURS  8 MINUTES 51.66 SECONDS     
-------------------------------------------------------------------------------

3.5 IFC-MPI
-----------------------------------------------------------------------------
(K+E1+L+N+X)           TOTAL ENERGY =          -16.62001424 A.U.
 (K)                  KINETIC ENERGY =           11.83136938 A.U.
 (E1=A-S+R)     ELECTROSTATIC ENERGY =          -11.60975504 A.U.
 (S)                           ESELF =           12.63317221 A.U.
 (R)                             ESR =            0.75872076 A.U.
 (L)    LOCAL PSEUDOPOTENTIAL ENERGY =          -14.28070114 A.U.
 (N)      N-L PSEUDOPOTENTIAL ENERGY =            1.41910664 A.U.
 (X)     EXCHANGE-CORRELATION ENERGY =           -3.98003409 A.U.
          GRADIENT CORRECTION ENERGY =           -0.13742382 A.U.
 ----------------------------------------------------------------
 TOTAL TIME                            526.20              527.14
 ****************************************************************

       CPU TIME :    0 HOURS  8 MINUTES 50.67 SECONDS     
   ELAPSED TIME :    0 HOURS  8 MINUTES 52.16 SECONDS     
-----------------------------------------------------------------------------------

3.5 PGF-MPI
-----------------------------------------------------------------------------
 (K+E+L+N+X)         TOTAL ENERGY =             -16.59797708 A.U.
 (K)               KINETIC ENERGY =              11.83027741 A.U.
 (E=A-S+R)   ELECTROSTATIC ENERGY =             -11.58823544 A.U.
 (S)         ESELF                =              12.63317171 A.U.
 (R)         ESR                  =               0.75831421 A.U.
 (L) LOCAL PSEUDOPOTENTIAL ENERGY =             -14.27972295 A.U.
 (N)   N-L PSEUDOPOTENTIAL ENERGY =               1.41962562 A.U.
 (X)  EXCHANGE-CORRELATION ENERGY =              -3.97992171 A.U.
       GRADIENT CORRECTION ENERGY =              -0.13742161 A.U.
----------------------------------------------------------------
 TOTAL TIME                           1017.07             1017.88
 ****************************************************************

       CPU TIME :    0 HOURS 17 MINUTES  7.97 SECONDS     
   ELAPSED TIME :    0 HOURS 17 MINUTES  9.45 SECONDS     
-------------------------------------------------------------------------------

3.7 IFC
--------------------------------------------------------------------------------------
 (K+E1+L+N+X)           TOTAL ENERGY =          -16.62001419 A.U.
 (K)                  KINETIC ENERGY =           11.83134416 A.U.
 (E1=A-S+R)     ELECTROSTATIC ENERGY =          -11.60978788 A.U.
 (S)                           ESELF =           12.63317221 A.U.
 (R)                             ESR =            0.75868794 A.U.
 (L)    LOCAL PSEUDOPOTENTIAL ENERGY =          -14.28064257 A.U.
 (N)      N-L PSEUDOPOTENTIAL ENERGY =            1.41909650 A.U.
 (X)     EXCHANGE-CORRELATION ENERGY =           -3.98002440 A.U.
          GRADIENT CORRECTION ENERGY =           -0.13742357 A.U.
----------------------------------------------------------------
 TOTAL TIME                            425.18              425.62
 ****************************************************************

       CPU TIME :    0 HOURS  7 MINUTES  8.85 SECONDS     
   ELAPSED TIME :    0 HOURS  7 MINUTES 10.51 SECONDS     
--------------------------------------------------------------------------------------

3.7 IFC-MPI
------------------------------------------------------------------------------------
 (K+E1+L+N+X)           TOTAL ENERGY =          -16.62001414 A.U.
 (K)                  KINETIC ENERGY =           11.83132159 A.U.
 (E1=A-S+R)     ELECTROSTATIC ENERGY =          -11.60981854 A.U.
 (S)                           ESELF =           12.63317221 A.U.
 (R)                             ESR =            0.75865709 A.U.
 (L)    LOCAL PSEUDOPOTENTIAL ENERGY =          -14.28058847 A.U.
 (N)      N-L PSEUDOPOTENTIAL ENERGY =            1.41908670 A.U.
 (X)     EXCHANGE-CORRELATION ENERGY =           -3.98001543 A.U.
          GRADIENT CORRECTION ENERGY =           -0.13742335 A.U.
 ----------------------------------------------------------------
 TOTAL TIME                            523.03              523.95
 ****************************************************************

       CPU TIME :    0 HOURS  8 MINUTES 47.40 SECONDS     
   ELAPSED TIME :    0 HOURS  8 MINUTES 48.92 SECONDS     
------------------------------------------------------------------------------


Those benchmarks run all on a single processor.
The problem is that energy results show pretty marked differences, in 
particular between the PGF executable and IFC executables. 
IFC executables have been compiled with ifc 6.0, and the same for lapack, 
atlas and mpich-1.2.5.
Are those results normal? does anybody has a test suit for the code?

Any help would be really appreciated

Alessandro Contini

-- 
Dott. Alessandro Contini
Istituto di Chimica Organica "Alessandro Marchesini"
Facoltà di Farmacia Università degli Studi di Milano
Via Venezian, 21 20133 Milano
Tel. +390250314480  Fax +390250314476
http://users.unimi.it/istchimorg/pagconthtm.htm




More information about the CPMD-list mailing list