Memory leak in CalibrateRectangularDetectors

[Wenduo Zhou]

As running the python algorithm CalibrateRectangularDetectors, memory leak is observed. Using mantid's "File"->"Clear All Memory" does not release much memory either.

Script is as

CalibrateRectangularDetectors(Instrument="PG3", RunNumber = 13428, Extension = event.nxs", CompressOnRead = False, SmoothSummedOffsets = False, MaxOffset = 0.1, CrossCorrelation = False, PeakPositions = "0.4093, 0.4120, 0.4205, 0.4327, 0.4359, 0.4460, 0.4645, 0.4768, 0.4996, 0.5148, 0.5439,0.5640, 0.6029,.6305,.6864,.7280,.8182,.8916,1.0754,1.2610,2.0592", PeakWindowMax = 0.1, PeakFunction = "Gaussian", BackgroundType = "Linear", Binning = ".40, -.0004, 2.1", GroupDetectorsBy = "Group", FilterBadPulses = True, SaveAs = "calibration", OutputDirectory = "/tmp/")

[Peter Peterson]

Re #7506. Checkpointing small bits of code cleanup.

Changeset: dc27fc8e6ccf11ae584e710c803640be9a2b1571

[Wenduo Zhou]

Checkpointing work. Refs #7506.

Changeset: aaeec8343f0e254810674bf6e41c78c37b1d4cb8

[Wenduo Zhou]

Checkpointing work. Refs #7506.

Changeset: 56a273d3cb180759ab1f1be977e64912421b1e67

[Nick Draper]

Wenduo,

This ticket has not been updated today, and you did not send me a progress report when asked.

If I do not hear from you why this ticket should delay the release and how much longer it is going to take then I’ll move it to release 3.0 when I get into work Tuesday morning.

[Wenduo Zhou]

Fixed the memory leak. Refs #7506.

Memory leak is caused by mistakenly using ws.dataX(spectrum), which causes a copy operatoin on vector, other than ws.readX(spectrum). As it is not threading safe, the memory leak occurs.

Changeset: 567c58d0be0adc094312403a5388e2355a1fed82

[Wenduo Zhou]

Fixed the inconsistency in unit test. Refs #7506.

Changeset: 4cd8c6dc5710d1ebe0648b898e3a6bf6546dfa3b

[Wenduo Zhou]

Revert some change. Refs #7506_calibrate_mem_leak.

Changeset: 52bbaaba60e47548c00b95102501e8dc02c82b25

[Wenduo Zhou]

Fixed the memory leak. Refs #7506.

Memory leak is caused by mistakenly using ws.dataX(spectrum), which causes a copy operatoin on vector, other than ws.readX(spectrum). As it is not threading safe, the memory leak occurs.

Changeset: 5fd0e176fed6f3f2e5d1ab65c13b178dc4c2c926

[Wenduo Zhou]

Fixed a possible memory error. Refs #7506.

Changeset: 3ab7e48924c9476ad4abd49b6dbb6491e7b9d039

[Wenduo Zhou]

How to verify

Re-run the script, which is described in the ticket such that the memory leak can be observed.

CalibrateRectangularDetectors(Instrument="PG3", RunNumber = 13428, Extension = event.nxs", CompressOnRead = False, SmoothSummedOffsets = False, MaxOffset = 0.1, CrossCorrelation = False, PeakPositions = "0.4093, 0.4120, 0.4205, 0.4327, 0.4359, 0.4460, 0.4645, 0.4768, 0.4996, 0.5148, 0.5439,0.5640, 0.6029,.6305,.6864,.7280,.8182,.8916,1.0754,1.2610,2.0592", PeakWindowMax = 0.1, PeakFunction = "Gaussian", BackgroundType = "Linear", Binning = ".40, -.0004, 2.1", GroupDetectorsBy = "Group", FilterBadPulses = True, SaveAs = "calibration", OutputDirectory = "/tmp/")

You can see that the problem reported should not occur anymore.

[Nick Draper]

Sorry we don't have that file at ISIS, can this be tested at ORNL?

[Wenduo Zhou]

Correct a spelling error. Refs #7506.

Changeset: 44688b8851ce0903c828da420ccac58b5fcad81e

[Russell Taylor]

BIG decrease in memory use from the critical change here.

[Stuart Campbell]

This ticket has been transferred to github issue ​8351