Ticket #9241 (closed: fixed)
FindPeaksMD should be done in parallel
Reported by: | Vickie Lynch | Owned by: | Vickie Lynch |
---|---|---|---|
Priority: | major | Milestone: | Release 3.2 |
Component: | Framework | Keywords: | |
Cc: | Blocked By: | ||
Blocking: | Tester: | Alex Buts |
Description
Finding 1000 peaks takes several minutes, so it should use threads.
Change History
comment:2 Changed 7 years ago by Vickie Lynch
- Status changed from assigned to inprogress
Refs #9241 calculateCentroid is problem; try double
Changeset: 02193e7ffb7e7b7467df5dc7a09921bcc7d84615
comment:3 Changed 7 years ago by Vickie Lynch
Revert "Refs #9241 calculateCentroid is problem; try double"
This reverts commit 02193e7ffb7e7b7467df5dc7a09921bcc7d84615.
Changeset: f3f9fa0e736c30823a528d6c9da647052ab2a3b6
comment:4 Changed 7 years ago by Vickie Lynch
Refs #9241 parallel loop of centroid
Changeset: 5946a775ab1650b4069572b2192450dec5d442ae
comment:5 Changed 6 years ago by Vickie Lynch
Refs #9241 calculate centroid once
Changeset: 719a77fd0b9ebf23190c8cb1a57bff1022c93ac1
comment:6 Changed 6 years ago by Vickie Lynch
- Status changed from inprogress to verify
- Resolution set to fixed
From 13 seconds to less than 1 second for FindPeaksMD using data from system tests:
Load(Filename=r'TOPAZ_3132_event.nxs',OutputWorkspace='TOPAZ_3132_event',LoadMonitors='1') ConvertToMD(InputWorkspace='TOPAZ_3132_event',QDimensions='Q3D',dEAnalysisMode='Elastic',Q3DFrames='Q_sample',LorentzCorrection='1',OutputWorkspace='TOPAZ_3132_md',MinValues='-25,-25,-25',MaxValues='25,25,25',SplitInto='2',SplitThreshold='50',MaxRecursionDepth='13',MinRecursionDepth='7') FindPeaksMD(InputWorkspace='TOPAZ_3132_md',PeakDistanceThreshold='0.35325000000000001',MaxPeaks='1000',DensityThresholdFactor='100',OutputWorkspace='TOPAZ_3132_peaks')
comment:7 Changed 6 years ago by Alex Buts
- Status changed from verify to verifying
- Tester set to Alex Buts
comment:8 Changed 6 years ago by Alex Buts
- Status changed from verifying to closed
Merge remote-tracking branch 'origin/feature/9241_parallel_FindPeaksMD'
Full changeset: 9f62bdc01ae10c7a3c994bb0aefac5792d7f3b36
comment:9 Changed 6 years ago by Alex Buts
Controversial solution, though there is obvious improvement by prefetching centroid.
Nothing in parallel here though and centroid overhead is now on each box even when you do not generally care about it...
As no current system test is affected, I am accepting it until any problem arises.
comment:10 Changed 5 years ago by Stuart Campbell
This ticket has been transferred to github issue 10084