Bug 582821 - Improve IO locality in MultiplePathsFromGCRootsComputerImpl bfs
Summary: Improve IO locality in MultiplePathsFromGCRootsComputerImpl bfs
Status: CLOSED MOVED
Alias: None
Product: MAT
Classification: Tools
Component: Core (show other bugs)
Version: 1.15   Edit
Hardware: PC Mac OS X
: P3 normal (vote)
Target Milestone: ---   Edit
Assignee: Jason Koch CLA
QA Contact:
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2023-12-31 15:01 EST by Jason Koch CLA
Modified: 2024-05-08 17:05 EDT (History)
1 user (show)

See Also:


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Jason Koch CLA 2023-12-31 15:01:28 EST

    
Comment 2 Andrew Johnson CLA 2024-01-03 14:44:47 EST
I think this is okay.
Another case of slow I/O is a HPROF file compressed with gzip (not so much with chucked GZIP as generated by the JVM, or by export HPROF with the chunked option). This can involve restarting the gunzip from a PosStream position, and depending on memory this might be more widely spaced than the 1MB chunked multi-memory gzip file.
Comment 3 Jason Koch CLA 2024-01-16 17:35:26 EST
Note that this changes the results presented in the tests because the sort() order changes the traversal order at each BFS layer. Removing the sort during each iteration confirms this.
Comment 4 Jason Koch CLA 2024-01-16 17:35:55 EST
the *regression test suite, that is
Comment 5 Eclipse Webmaster CLA 2024-05-08 17:05:48 EDT
This issue has been migrated to https://github.com/eclipse-mat/org.eclipse.mat/issues/46.