[Wien] Lapw1_mpi

[Florent Boucher]

Dear Peter,
I did the test with the "old" lapw1_mpi and it works.
Here are the performances. I choose a matrix that is not too big, just 
for test.
If you want a test on a larger system, just let me know.
Regards
Florent

      MATRIX SIZE 5908  

 2 CPU :
 
 Working with            1  x            2  Process Grid
 My ID =            0
 DESCPANEL, DESCHS, DESCZ (2):           0           1           1
 Time for al,bl:   0.207031250000000    
 nlo         644
 Time for los:    5.25390625000000    
 Time for LOOP 260:         71.0585937500000    
     HNS: Loop 140:      1.1 Loop 90:     16.5 matrix update:    
163.3loop 30:    15.4(wall time)
 Seclr4(Cholesky complete (CPU)) :              20.988     3275.091 Mflops
 Seclr4(Cholesky complete (WALL)) :             20.990     3274.894 Mflops
 Seclr4(Transform to eig.problem (CPU)) :       63.516     3246.690 Mflops
 Seclr4(Transform to eig.problem (WALL)) :      63.521     3246.440 Mflops
 Seclr4(Compute eigenvalues (CPU)) :           269.301     1020.993 Mflops
 Seclr4(Compute eigenvalues (WALL)) :          269.299     1020.998 Mflops
 Seclr4(Backtransform (CPU)) :                   5.832      888.766 Mflops
 Seclr4(Backtransform (WALL)) :                  5.832      888.793 Mflops
       TIME HAMILT (CPU)  =    76.7, HNS =   181.0, DIAG =   359.6
       TIME HAMILT (WALL) =    76.7, HNS =   181.0, DIAG =   359.6
 



 4 CPU :
Elapse Time : 355s
Memory : 4 x 398Mo = 1592 Mo

 Working with            2  x            2  Process Grid
 My ID =            0
 DESCPANEL, DESCHS, DESCZ (2):           0           1           1
 Time for al,bl:   0.208984375000000    
 nlo         644
 Time for los:    6.06445312500000    
 Time for LOOP 260:         45.3515625000000    
     HNS: Loop 140:      1.1 Loop 90:     16.6 matrix update:     
86.6loop 30:    15.5(wall time)
 Seclr4(Cholesky complete (CPU)) :              10.967     6267.876 Mflops
 Seclr4(Cholesky complete (WALL)) :             10.967     6267.665 Mflops
 Seclr4(Transform to eig.problem (CPU)) :       33.383     6177.298 Mflops
 Seclr4(Transform to eig.problem (WALL)) :      33.387     6176.576 Mflops
 Seclr4(Compute eigenvalues (CPU)) :           130.611     2105.132 Mflops
 Seclr4(Compute eigenvalues (WALL)) :          130.611     2105.138 Mflops
 Seclr4(Backtransform (CPU)) :                   3.006     1724.403 Mflops
 Seclr4(Backtransform (WALL)) :                  3.007     1723.913 Mflops
       TIME HAMILT (CPU)  =    51.9, HNS =   104.3, DIAG =   178.0
       TIME HAMILT (WALL) =    51.9, HNS =   104.3, DIAG =   178.0
 
     K=   0.50000   0.50000   0.50000            1
      MATRIX SIZE 5908  WEIGHT= 1.00  PGR:   

 8 CPU :
Elapse Time : 242s
Memory : 8 x 294Mo = 2352 Mo


 DESCPANEL, DESCHS, DESCZ (2):           0           1           1
 Time for al,bl:   0.208984375000000    
 nlo         644
 Time for los:    5.95117187500000    
 Time for LOOP 260:         34.2382812500000    
     HNS: Loop 140:      1.1 Loop 90:     16.6 matrix update:     
53.0loop 30:    15.5(wall time)
 Seclr4(Cholesky complete (CPU)) :               6.273    10957.075 Mflops
 Seclr4(Cholesky complete (WALL)) :              6.275    10954.326 Mflops
 Seclr4(Transform to eig.problem (CPU)) :       19.354    10655.200 Mflops
 Seclr4(Transform to eig.problem (WALL)) :      19.355    10654.167 Mflops
 Seclr4(Compute eigenvalues (CPU)) :            83.008     3312.388 Mflops
 Seclr4(Compute eigenvalues (WALL)) :           83.006     3312.453 Mflops
 Seclr4(Backtransform (CPU)) :                   2.012     2576.559 Mflops
 Seclr4(Backtransform (WALL)) :                  2.012     2576.640 Mflops
       TIME HAMILT (CPU)  =    40.6, HNS =    70.7, DIAG =   110.6
       TIME HAMILT (WALL) =    40.6, HNS =    70.7, DIAG =   110.6
 
     K=   0.50000   0.50000   0.50000            1

 16 CPU :
Elapse Time : 146s
Memory : 16 x 240Mo = 3840 Mo

 Working with            4  x            4  Process Grid
 My ID =            0
 DESCPANEL, DESCHS, DESCZ (2):           0           1           1
 Time for al,bl:   0.212890625000000    
 nlo         644
 Time for los:    4.41992187500000    
 Time for LOOP 260:         14.3320312500000    
     HNS: Loop 140:      1.1 Loop 90:     17.3 matrix update:     
30.3loop 30:    16.2(wall time)
 Seclr4(Cholesky complete (CPU)) :               3.400    20214.891 Mflops
 Seclr4(Cholesky complete (WALL)) :              3.401    20213.456 Mflops
 Seclr4(Transform to eig.problem (CPU)) :       11.016    18720.279 Mflops
 Seclr4(Transform to eig.problem (WALL)) :      11.017    18717.209 Mflops
 Seclr4(Compute eigenvalues (CPU)) :            46.498     5913.240 Mflops
 Seclr4(Compute eigenvalues (WALL)) :           46.498     5913.262 Mflops
 Seclr4(Backtransform (CPU)) :                   0.953     5438.230 Mflops
 Seclr4(Backtransform (WALL)) :                  0.953     5438.961 Mflops
       TIME HAMILT (CPU)  =    19.2, HNS =    48.0, DIAG =    61.9
       TIME HAMILT (WALL) =    19.2, HNS =    48.0, DIAG =    61.9

-- 
 --------------------------------------------------------------------------
| Florent BOUCHER                    |                                     |
| Institut des Matériaux Jean Rouxel | Mailto:Florent.Boucher at cnrs-imn.fr  |
| 2, rue de la Houssinière           | Phone: (33) 2 40 37 39 24           |
| BP 32229                           | Fax:   (33) 2 40 37 39 95           |
| 44322 NANTES CEDEX 3 (FRANCE)      | http://www.cnrs-imn.fr              |
 --------------------------------------------------------------------------

-------------- next part --------------
A non-text attachment was scrubbed...
Name: Lapw1_mpi_2.pdf
Type: application/pdf
Size: 18491 bytes
Desc: not available
Url : http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20031215/5c4c0cb1/Lapw1_mpi_2.pdf
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Lapw1_mpi_1.pdf
Type: application/pdf
Size: 19167 bytes
Desc: not available
Url : http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20031215/5c4c0cb1/Lapw1_mpi_1.pdf