Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

add -report_max for duplicate errors to reduce overhead on constructing call stacks #897

Open
derekbruening opened this issue Nov 28, 2014 · 2 comments

Comments

@derekbruening
Copy link
Contributor

From zhao...@google.com on May 21, 2012 18:44:11

When there are too many duplicated errors, the overhead on constructing the call stack for verification is too high and should be avoided.
One example is 471.omnetpp, xref issue #862 .

Original issue: http://code.google.com/p/drmemory/issues/detail?id=897

@derekbruening
Copy link
Contributor Author

From bruen...@google.com on May 21, 2012 17:30:31

note that the current -report_max deliberately applies only to
unique errors, as tons of just one unique error shouldn't
preclude reporting other errors

Summary: add -report_max for duplicate errors to reduce overhead on constructing call stacks

@derekbruening
Copy link
Contributor Author

From rnk@google.com on May 21, 2012 20:18:23

If we assume that in normal operation 99% of errors are suppressed, this would be helped by issue #884 . If we had a trie of suppressed error stack prefixes, then we'd only have to unwind a few frames before we can tell that the report should be suppressed, rather than taking the whole stack. I doubt this would help omnetpp performance much because it is already doing a short unwind of only 12 frames.

If we implemented a shadow callstack, that would shift all of the work of creating the callstack from report time to code cache time. It would make this issue go away, but it would reduce our overall performance.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

1 participant