Changes between Version 2 and Version 3 of ticket/0677_mpp_rep


Ignore:
Timestamp:
2010-06-07T09:18:59+02:00 (10 years ago)
Author:
rblod
Comment:

Legend:

Unmodified
Added
Removed
Modified
  • ticket/0677_mpp_rep

    v2 v3  
    1111Implementation of both methods to get mpp reproducibility, one from ECMWF (key_mpp_rep1) and the other from DFO (key_mpp_rep2). The target is to choose one, thanks to my reviewer's advices, but athis time (7th of June), I made an intensive use of cpp keys to delimit clearly the both methods. 
    1212 
    13 Both are based on the Idea of self compensated summation, see the paper "Using Accurate Arithmetics to Improve Numerical Reproducibility and Stability in parallel applications, Yun He and Chris Ding, Journal of supercomputing, Vol 18, Number 3, pages 259-277, doi 10.1023/A1008153532043. 
     13Both are based on the Idea of self compensated summation, see the paper "Using Accurate Arithmetics to Improve Numerical Reproducibility and Stability in parallel applications, Yun He and Chris Ding, Journal of supercomputing, Vol 18, Number 3, pages 259-277, doi 10.1023/A1008153532043. 
     14 
     15We have: 
     16 
     17sum = a+b 
     18 
     19error = b + (a-sum) 
     20 
     21In the next addition, the error is first added back : 
     22 
     23(sum,error) = SCS(a,b) 
     24 
     25(sum1,error1) = SCS(sum,c+error) 
     26 
     27These methods have been implemented in a new module lib_fortran.F90 with a few additions in lib_mpp.F90. In the sake of simplicity, I implemented a glob_sum function which is either a standard one( SUM + CALL mpp_sum), either one of the otw methods and the switch is done in lib_fortran. 
     28 
     29Nota: I also used this branch to implement a SIGN function which overwrite the standard fortran one (key_nosignedzero) to keep the f90 behaviour. 
    1430 
    1531=== Testing ===