New URL for NEMO forge!   http://forge.nemo-ocean.eu

Since March 2022 along with NEMO 4.2 release, the code development moved to a self-hosted GitLab.
This present forge is now archived and remained online for history.
ticket/1821/BranchTesting (diff) – NEMO

Changes between Initial Version and Version 1 of ticket/1821/BranchTesting


Ignore:
Timestamp:
2017-01-27T13:07:51+01:00 (7 years ago)
Author:
frrh
Comment:

--

Legend:

Unmodified
Added
Removed
Modified
  • ticket/1821/BranchTesting

    v1 v1  
     1 
     2 
     3Setting up and testing optimisations for #1821 
     4 
     5== Branch Creation == 
     6 
     7'''Here's how I created the branch:''' 
     8 
     9  * Created branch [log:branches/UKMO/dev_r5518_optim_GO6_alloc branches/UKMO/dev_r5518_optim_GO6_alloc] from head of package branch (r7573 at time of writing).  
     10 
     11  * Stripped out svn keywords. (I'm surprised we need to do this at all since we're creating a branch of a branch which has already been stripped - there are only a subset of files affected by this too... mostly AGRIF related)  
     12 
     13  * Merge in the optimisations from our MEDUSA based branch and add further changes in the same style which should be of benefit to GO6 more generally. Revisions r7581:r7602 show the actual code changes made.  
     14 
     15== Testing == 
     16 
     17'''Initial testing in a copy of the GO6 standard job.''' 
     18 
     19  * Took copy of standard suite u-ah494/trunk@29996 to u-aj380Optim to act as control run 
     20 
     21  * Created working copy of this to act as test run. u-aj380Optim 
     22 
     23  * Replaced branches/UKMO/dev_r5518_GO6_package@7573 in working copy with branches/UKMO/dev_r5518_optim_GO6_alloc@7602 
     24 
     25  * Ran both jobs for a 10 day NRUN with NEMO timings activated. Repeated both runs to try to iron out any random variations in run time on the XC40.  
     26 
     27  * Checking solver.stat at 10 days, we have bit comparison. (I've not done a rigorous comparison of restart files) 
     28 
     29  * Comparing NEMO timer output, it seems that there are indications that the optimised run may be ~1-2% faster overall, looking at total elapsed and total CPU, though these differences 
     30    lie well within the noise of variabilities in run time which can be 10% or more.    
     31 
     32  * Here's some output; 
     33 
     34  * Control run: 
     35{{{ 
     36Elapsed Time (s)  CPU Time (s) 
     37       593483.386   584478.452 
     38  
     39Averaged timing on all processors : 
     40----------------------------------- 
     41Section             Elap. Time(s)  Elap. Time(%)  CPU Time(s)  CPU Time(%)  CPU/Elap Max elap(%)  Min elap(%)  Freq 
     42sbc_ice_cice        0.1751989E+03   14.17           174.41      14.32        1.00         14.50     13.76     640.00 
     43tra_adv_tvd         0.7552396E+02    6.11            75.33       6.19        1.00          7.42      5.27     640.00 
     44tra_nxt             0.6302550E+02    5.10            62.89       5.17        1.00          7.65      2.14     640.00 
     45tra_ldf_iso         0.5789208E+02    4.68            57.77       4.74        1.00          5.08      4.06     640.00 
     46sol_pcg             0.5625327E+02    4.55            56.08       4.61        1.00          4.55      4.55     640.00 
     47dia_wri             0.4800178E+02    3.88            47.84       3.93        1.00          9.16      0.04     640.00 
     48tra_bbc             0.4565193E+02    3.69            45.39       3.73        0.99         11.38      0.09     640.00 
     49nonosc              0.4531063E+02    3.66            45.23       3.71        1.00          4.11      3.44    1280.00 
     50zps_hde             0.3959416E+02    3.20            39.35       3.23        0.99          7.05      0.08    1281.00 
     51ldf_slp             0.3823219E+02    3.09            38.15       3.13        1.00          3.25      2.98     640.00 
     52 
     53}}} 
     54 
     55  * Test Run: 
     56{{{ 
     57Total timing (sum) : 
     58 -------------------- 
     59Elapsed Time (s)  CPU Time (s) 
     60       583015.673   573647.291 
     61  
     62Averaged timing on all processors : 
     63----------------------------------- 
     64Section             Elap. Time(s)  Elap. Time(%)  CPU Time(s)  CPU Time(%)  CPU/Elap Max elap(%)  Min elap(%)  Freq 
     65sbc_ice_cice        0.1728438E+03   14.23           172.01      14.39        1.00         14.56     13.80     640.00 
     66tra_adv_tvd         0.7375908E+02    6.07            73.52       6.15        1.00          7.29      5.22     640.00 
     67tra_nxt             0.6276407E+02    5.17            62.63       5.24        1.00          7.68      2.33     640.00 
     68tra_ldf_iso         0.5655850E+02    4.66            56.48       4.73        1.00          4.86      3.43     640.00 
     69sol_pcg             0.4990999E+02    4.11            49.77       4.16        1.00          4.11      4.11     640.00 
     70dia_wri             0.4792860E+02    3.95            47.80       4.00        1.00          9.32      0.05     640.00 
     71tra_bbc             0.4585938E+02    3.78            45.57       3.81        0.99         11.72      0.10     640.00 
     72zps_hde             0.3905950E+02    3.22            38.80       3.25        0.99          7.08      0.09    1281.00 
     73ldf_slp             0.3817614E+02    3.14            38.10       3.19        1.00          3.30      3.03     640.00 
     74nonosc              0.3657865E+02    3.01            36.47       3.05        1.00          3.48      2.77    1280.00 
     75 
     76 
     77}}} 
     78