Bug 221826 - Regression in some 'JavaReplaceAll' performance tests
Summary: Regression in some 'JavaReplaceAll' performance tests
Status: RESOLVED WONTFIX
Alias: None
Product: JDT
Classification: Eclipse Project
Component: Text (show other bugs)
Version: 3.4   Edit
Hardware: PC Windows XP
: P3 normal (vote)
Target Milestone: 3.4 RC1   Edit
Assignee: Dani Megert CLA
QA Contact:
URL:
Whiteboard:
Keywords: performance, test
Depends on:
Blocks:
 
Reported: 2008-03-07 07:15 EST by Frederic Fusier CLA
Modified: 2008-05-16 05:19 EDT (History)
1 user (show)

See Also:

Attachments
Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description Frederic Fusier CLA 2008-03-07 07:15:55 EST 
Looking at I20080305-1100 performance results, it seems that JavaReplaceAll#test() and JavaReplaceAllWithQuickDiffTest#test() show a significant regression.

Here are the numbers:
 - JavaReplaceAll#test():                  between -6.2% and -18.9%
 - JavaReplaceAllWithQuickDiffTest#test(): between -5.6% and -15.8%


For these tests, it seems that something happened between I20071120-0800 and I20071127-0800 builds (not 100% sure as one of the test box does not show the same gap).
Comment 1 Dani Megert CLA 2008-05-15 09:53:36 EDT 
One part of the regression is caused by a fix which prevents data corruption. This is nothing we can change/improve unless brining back the bug again (bug 130843).

Second jump needs to be investigated.
Comment 2 Dani Megert CLA 2008-05-15 11:24:10 EDT 
Another little regression comes from the colored labels (20-40ms) in the Java Outline. This was not enabled in 3.3.
Comment 3 Dani Megert CLA 2008-05-16 05:19:24 EDT 
The second jump is due to a performance optimization (see bug 113247 for details): document and annotation access became faster and hence things rulers and quick diff are faster and start updating the UI before we reach stop() i.e. for the user it is/looks faster but our stop condition doesn't catch this, as we do not wait for all those things to be present.
Comment 4 Dani Megert CLA 2008-05-16 05:19:42 EDT 
I will comment this on the test. No plans to tweak the test at this point.