Changes between Initial Version and Version 1 of Working Groups/HPC/Mins_MP_2019_01_23


Ignore:
Timestamp:
2019-02-01T09:52:59+01:00 (19 months ago)
Author:
mcastril
Comment:

Legend:

Unmodified
Added
Removed
Modified
  • Working Groups/HPC/Mins_MP_2019_01_23

    v1 v1  
     1= Discussion of Development of a mixed-precision option within NEMO. Grenoble, France. 
     2 
     3[[PageOutline]] 
     4 
     5Last edition on '''[[Wikinfo(changed_ts)]]''' by '''[[Wikinfo(changed_by)]]''' 
     6 
     7 
     8== Participants 
     9 
     10* BSC: Miguel Castrillo, Oriol Tintó 
     11* CNRS: Gurvan Madec, Claire Levy, Sebastien Masson, Rachid Benshila 
     12* UKMO: Mike Bell 
     13* ECMWF: Nils Wedi 
     14* INRIA: Laurent Debreu 
     15* Ocean Next: Laurent Brodeau 
     16 
     17== Agenda 
     18 
     19- Presentation of the work by the BSC 
     20- Discussion   
     21 
     22== Minutes 
     23 
     24Oriol gave a brief explanation of the motivation and the methodology, describing the initial test used to assess the impact (using differences from simulations with different timesteps).  
     25 
     26Gurvan noted that the Asselin filter parameter should be modified when the timestep length is changed and that certain parts of the code would be expected to be sensitive to precision: for example the tri-diagonal solver in the vertical diffusion, the 2D barotropic mode, accumulation of truncation error in the 
     27surface pressure gradient and the vertical scale factors. Other aspects like biharmonic dissipation should not need high precision calculation.  
     28It was agreed that it would be interesting to see whether Oriol's results are consistent with these insights.  
     29 
     30Nils asked whether there was a concensus that mixed precision is worth investment of effort. It was agreed that it is because it can give nearly a factor of 2 reduction in cost. Recent simplification work in the communication interfaces should also assist the creation of single-precision versions of the communication routines.  
     31 
     32Nils asked what approach would be taken to implementation. It was agreed that code maintenance is a priority and mechanisms to do this using cpp/fpp substitutions were discussed. 
     33 
     34It is difficult to define "one size fits all" tests for mixed precision calculations. A useful approach is to explore the sensitivity of results for specific processes in idealised test cases.   
     35There was a consensus that tests to assess the accuracy of the results would not only be useful for the work on mixed precision, but also for the NEMO community in general. 
     36 
     37It was agreed to explore mixed precision further. To take things forward, Oriol will visit IPSL for a few days in March when Gurvan is back from Australia.