Slow down of reduction after running it through many loops

[Martyn Gigg]

This script demonstrates the problem (data is in systemtests). If you look at the timing of the reduction then it increases a lot between the first few runs and the last.

import mantid
from mantid.simpleapi import *
import dgreduce

iliad_setup=dgreduce.setup
iliad=dgreduce.arb_units
iliad_abs=dgreduce.abs_units
iliad_help=dgreduce.help

inst='mar'
iliad_setup(inst)
ext='.raw'
mapfile='mari_res'
#det_cal_file must be specified if the reduction sends out put to a workpsace #cal_file='MAR17578.raw'
cal_file='MAR11060.RAW'
#load vanadium file


for i in range(200):
        print '############' ,i,' ##################'
        run='11001'
        WB='MAR11060.RAW'
        ei = 12.5
        rebin_params = "-6.25,0.03125,11.875"
        BkgdRange = [11705.726218019521, 19500]

        w1=iliad(WB,run,ei,rebin_params,mapfile,det_cal_file=cal_file,norm_method='current')
        DeleteWorkspace('w1')

[Martyn Gigg]

Fix bug when loading duplicate instrument parameters.

The multimap was continuously growing as there was no way of checking if a given parameter for a component was already loaded. The storage is now a map with key <Component*,name> so that only unique component/named pairs will be stored. Refs #7743.

Changeset: c34a3597e9d84d01bc0619c4aa202c111e914f03

[Martyn Gigg]

Fix case when no log exists for parameter link.

Also, improved error message for LoadEmptyInstrument when running child algorithm. Refs #7743

Changeset: 48ed7f59d88050f2e0de14c178e86920a054b805

[Martyn Gigg]

Add equality operator to ParameterMap...

and use it in CheckWorkspacesMatch. This avoids changes in ordering within the map affecting whether the map is actually the same as another. Refs #7743

Changeset: 4b7aee01301d57493c72b16e0dcb28de8ec18efa

[Martyn Gigg]

Branch: bugfix/7743_reduction_looping_slow_down

Tester: Before merging the changes confirm to yourself that the processing of each iteration takes longer as the script keeps executing (it prints the loop time after each iteration).

Merge the changes and rerun the the script above. Confirm that each loop now takes roughly the same time, with the exception of the first that has to load the instrument.

Check that the systemtests have been run (i.e there has been at least one night between the changes going to develop and the point you are testing) and are passing.

[Russell Taylor]

After the changes there is no slowdown trend when running the above script. I am relying fairly heavily on the tests to validate the changes, but they look good and have certainly cleaned things up in certain places.

[Stuart Campbell]

This ticket has been transferred to github issue ​8588