ROMS profile report

General scientific issues regarding ROMS

Moderators: arango, robertson

Post Reply
Message
Author
User avatar
kate
Posts: 4091
Joined: Wed Jul 02, 2003 5:29 pm
Location: CFOS/UAF, USA

ROMS profile report

#1 Unread post by kate »

I was wondering if anyone else got puzzling results from the profiles built into ROMS. I ran the same job on both the IBM pwr4 and the Cray X1 systems. Admittedly, the ice stuff is in the wrong place.

Code: Select all

IBM:
 Nonlinear model elapsed time profile:

  Initialization ...................................        54.950  ( 0.0183 %)
  Reading of input data ............................       223.390  ( 0.0745 %)
  Processing of input data .........................       542.310  ( 0.1808 %)
  Processing of output time averaged data ..........     11962.020  ( 3.9875 %)
  Computation of vertical boundary conditions ......       946.300  ( 0.3154 %)
  Computation of global information integrals ......       639.480  ( 0.2132 %)
  Writing of output data ...........................     19655.350  ( 6.5520 %)
  Model 2D kernel ..................................     67543.550  (22.5152 %)
  2D/3D coupling, vertical metrics .................       929.770  ( 0.3099 %)
  Omega vertical velocity ..........................      1702.610  ( 0.5676 %)
  Equation of state for seawater ...................      5655.350  ( 1.8852 %)
  KPP vertical mixing parameterization .............     33417.750  (11.1396 %)
  3D equations right-side terms ....................      9002.220  ( 3.0008 %)
  3D equations predictor step ......................     26992.110  ( 8.9976 %)
  Pressure gradient ................................      4806.900  ( 1.6023 %)
  Harmonic mixing of tracers, geopotentials ........      7903.390  ( 2.6345 %)
  Harmonic sress tensor, S-surfaces ................      3875.650  ( 1.2919 %)
  Corrector time-step for 3D momentum ..............     18419.660  ( 6.1401 %)
  Corrector time-step for tracers ..................     13043.060  ( 4.3478 %)
                                              Total:    227315.820   75.7741

 Nonlinear model message Passage profile:

  Message Passage: halo exchange ...................     32606.690  (10.8692 %)
  Message Passage: data broadcast ..................        96.030  ( 0.0320 %)
  Message Passage: data reduction ..................        50.950  ( 0.0170 %)
  Message Passage: data gathering ..................     19070.720  ( 6.3571 %)
  Message Passage: data scattering..................       155.590  ( 0.0519 %)
  Message Passage: boundary data gathering .........     31535.280  (10.5121 %)
  Message Passage: point data gathering ............      1460.360  ( 0.4868 %)
  Bulk fluxes using NCEP ...........................     10358.960  ( 3.4531 %)
  Generate coefficients for ice dynamics solver.....     27213.810  ( 9.0715 %)
  Generate RHS for ice dynamics solver..............       236.690  ( 0.0789 %)
                                              Total:    122785.080   40.9295

Time output:

real   5h 13m 51.37s
user       0m  0.51s
sys        0m  2.31s

Cray: 
Nonlinear model elapsed time profile:

  Initialization ...................................       111.369  ( 1.0189 %)
  Reading of input data ............................       406.327  ( 3.7173 %)
  Processing of input data .........................       300.441  ( 2.7486 %)
  Processing of output time averaged data ..........      2065.267  (18.8940 %)
  Computation of global information integrals ......       102.608  ( 0.9387 %)
  Writing of output data ...........................     15116.464  (******* %)
  2D/3D coupling, vertical metrics .................       144.929  ( 1.3259 %)
  Omega vertical velocity ..........................      1729.715  (15.8242 %)
  Equation of state for seawater ...................       909.494  ( 8.3205 %)
  3D equations right-side terms ....................      1946.450  (17.8070 %)
  Pressure gradient ................................       910.191  ( 8.3268 %)
  Harmonic sress tensor, S-surfaces ................       708.939  ( 6.4857 %)
  Corrector time-step for tracers ..................      1404.420  (12.8483 %)
                                              Total:     25856.614  236.5478

 Nonlinear model message Passage profile:

  Message Passage: data broadcast ..................       154.671  ( 1.4150 %)
  Message Passage: data reduction ..................        14.826  ( 0.1356 %)
  Message Passage: data gathering ..................     10344.180  (94.6332 %)
  Message Passage: data scattering..................       265.362  ( 2.4276 %)
  Bulk fluxes using NCEP ...........................      2938.392  (26.8817 %)
  Generate coefficients for ice dynamics solver.....       379.307  ( 3.4701 %)
  Generate RHS for ice dynamics solver..............        99.862  ( 0.9136 %)
                                              Total:     14196.600  129.8768

Time output:

real   6h 44m 35.48s
user  93h 17m 24.35s
sys   12h 52m 36.35s

Both jobs are on 16 processors, both fit in the 8 hour queue. On BENCHMARK3 with no output, the Cray was substantially faster than the IBM.

Post Reply