Ticket #2038 (closed: fixed)

Opened 10 years ago

Last modified 5 years ago

MemoryLeak in LoadSNSEventNexus?

Reported by: Janik Zikovsky Owned by: Janik Zikovsky
Priority: major Milestone: Iteration 26
Component: Mantid Keywords:
Cc: pf9@… Blocked By:
Blocking: Tester: Vickie Lynch

Description

Memory use seems to grow by a few hundred MB when repeatedly loading PG3_1370

Change History

comment:1 Changed 10 years ago by Janik Zikovsky

  • Status changed from new to accepted

comment:2 Changed 10 years ago by Janik Zikovsky

(In [7955]) Refs #2038: Memory leak test.

comment:3 Changed 10 years ago by Nick Draper

Russell has some experience in tracking these memory leaks down.

comment:4 Changed 10 years ago by Janik Zikovsky

(In [7971]) Refs #2038: memory leak tester.

comment:5 Changed 10 years ago by Janik Zikovsky

(In [7972]) Refs #1973 and #2038: Fixed memory leak due to eventWorkspace->CopyDataFrom()

comment:6 Changed 10 years ago by Janik Zikovsky

(In [7979]) Refs #2038: Commentedout memory leak tester.

comment:7 Changed 10 years ago by Janik Zikovsky

(In [7982]) Refs #2038: Little things that might help memory management.

comment:8 Changed 10 years ago by Janik Zikovsky

  • Status changed from accepted to verify
  • Resolution set to fixed

(In [7985]) Fixes #1973 and #2038: fairly confident that there is no actual memory leak in Linux or Windows, and that the issue is with improper memory use reporting.

comment:9 Changed 10 years ago by Nick Draper

  • Milestone changed from Iteration 26 to Iteration 27

Bulk move of tickets to iteration 27, if your ticket is essential for Iteration 26 then move it back.

comment:10 Changed 10 years ago by Nick Draper

  • Milestone changed from Iteration 27 to Iteration 26

Sorry I didn't mean to move these ones reverting back to It 26

comment:11 Changed 10 years ago by Vickie Lynch

  • Status changed from verify to verifying
  • Tester set to Vickie Lynch

comment:12 Changed 10 years ago by Vickie Lynch

I looked at Memory History in the Ubuntu system monitor while repeatedly calling LoadEventPreNexus with PG3_1370_neutron_event.dat and the same OutputWorkspace. Before call 896MiB After 1st 2.2GiB After 2nd 3.5GiB After 3rd 4.8GiB After 4th 6.1GiB After workspace deleted 6.1 GiB

Since there are 60 million events in this file, the growth divided by the number of events is 22. This still looks like a memory leak to me.

comment:13 Changed 10 years ago by Janik Zikovsky

I believe this to be a Linux memory reporting issue rather than an actual memory leak. Valgrind does not find any significant leak (only a few KB); and the same algorithm under windows does not show this behavior, according to Russell.

comment:14 Changed 10 years ago by Vickie Lynch

  • Status changed from verifying to closed

I cleared the event vector and the memory reported did not change, so this may be a Linux issue.

comment:15 Changed 5 years ago by Stuart Campbell

This ticket has been transferred to github issue ​2885

Note: See TracTickets for help on using tickets.