Community
Participate
Working Groups
See "http://fullmoon.ottawa.ibm.com/eclipse/downloads/drops/I20050324-1400/performance/win/org.eclipse.ui.tests.performance.OpenCloseEditorTest.testOpenAndCloseEditors_java().html" "http://fullmoon.ottawa.ibm.com/eclipse/downloads/drops/I20050324-1400/performance/win/org.eclipse.ui.tests.performance.OpenCloseEditorTest.testOpenAndCloseEditors_perf_basic().html" http://fullmoon.ottawa.ibm.com/eclipse/downloads/drops/I20050324-1400/performance/win/org.eclipse.ui.tests.performance.OpenCloseEditorTest.testOpenAndCloseEditors_perf_outline().html The problem seems to have been introduced between I20050315-1100 and I20050324-1400.
The test needed some work but the failure is legitimate. Sadly the time you saw was a change in the test machine so it doesn't give us as much to go on. I have updated and released the tests. Here are the results. R 3.0.2 ----- testOpenAndCloseEditors:perf_basic testOpenAndCloseEditors:perf_basic: setUp... Scenario 'org.eclipse.ui.tests.performance.OpenCloseEditorTest#testOpenAndCloseEditors:perf_basic()' (average over 100 samples): Used Java Heap: 6.09K Working Set: 28.4K Committed: 14.08K Working Set Peak: 0 Elapsed Process: 377 ms Kernel time: 136 ms Page Faults: 19 CPU Time: 367 ms GDI Objects: 0 testOpenAndCloseEditors:perf_basic: tearDown... ----- testOpenAndCloseEditors:perf_outline testOpenAndCloseEditors:perf_outline: setUp... Scenario 'org.eclipse.ui.tests.performance.OpenCloseEditorTest#testOpenAndCloseEditors:perf_outline()' (average over 100 samples): Used Java Heap: 7.15K Working Set: 1.84K Committed: -3031 Working Set Peak: 0 Elapsed Process: 422 ms Kernel time: 176 ms Page Faults: 10 CPU Time: 410 ms testOpenAndCloseEditors:perf_outline: tearDown... GDI Objects: 0 ----- testOpenAndCloseEditors:java testOpenAndCloseEditors:java: setUp... Scenario 'org.eclipse.ui.tests.performance.OpenCloseEditorTest#testOpenAndCloseEditors:java()' (average over 100 samples): Used Java Heap: 28.27K Working Set: 206.24K Committed: 195K Working Set Peak: 205.84K Elapsed Process: 3.82 s Kernel time: 925 ms Page Faults: 170 CPU Time: 3.81 s GDI Objects: 10 testOpenAndCloseEditors:java: tearDown... R3.1 ----- testOpenAndCloseEditors:perf_basic testOpenAndCloseEditors:perf_basic: setUp... LOCAL Open/Close Editor Scenario 'org.eclipse.ui.tests.performance.OpenCloseEditorTest#testOpenAndCloseEditors:perf_basic()' (average over 100 samples): Used Java Heap: 40.18K Working Set: 71.64K Committed: 91.88K Working Set Peak: 74.84K Elapsed Process: 482 ms Kernel time: 195 ms Page Faults: 161 CPU Time: 493 ms GDI Objects: 20 testOpenAndCloseEditors:perf_basic: tearDown... ----- testOpenAndCloseEditors:perf_outline testOpenAndCloseEditors:perf_outline: setUp... Scenario 'org.eclipse.ui.tests.performance.OpenCloseEditorTest#testOpenAndCloseEditors:perf_outline()' (average over 100 samples): Used Java Heap: 25.94K Working Set: 13.56K Committed: 66.12K Working Set Peak: 0 Elapsed Process: 482 ms Kernel time: 233 ms Page Faults: 60 CPU Time: 454 ms GDI Objects: 20 testOpenAndCloseEditors:perf_outline: tearDown... ----- testOpenAndCloseEditors:java testOpenAndCloseEditors:java: setUp... Scenario 'org.eclipse.ui.tests.performance.OpenCloseEditorTest#testOpenAndCloseEditors:java()' (average over 100 samples): Used Java Heap: 83.13K Working Set: 341.44K Committed: 350.28K Working Set Peak: 310.48K Elapsed Process: 3.12 s Kernel time: 1.33 s Page Faults: 881 CPU Time: 3.14 s testOpenAndCloseEditors:java: tearDown... GDI Objects: 20
Adding Bug 88050 as a blocking bug - this is what is making our test for open/close the java editor so bad
Opening all of org.eclipse.swt.internal.win32 with text rather than java editors resulted in the following heavy points TextEditor#createPartControl was 8249ms and is now 11220ms (23.6% of the time) TextEditor#doSetInput took 6911 ms. It was fast enough to not show up on my 3.0 trace. (14.5% of the time) HandlerAuthority#sourceChanged takes 11440ms. This was not in 3.0. (24% of the time) PresentablePartFolder#layoutContent took 4473ms (9.4% of the time). I think this is actually better than 3.0 though.
Moving to Stefan to investigate the failures for basic performance as Nick as less familiar with the changes in this space.
tod is this still an issue/important?
I reworked these early in 3.2 so they are more meaningful now. We should run against 3.1 again but the test itself is good.
Is this still a problem in 3.3? PW
Changes requested on bug 193523
There are other bugs covering this in 3.3 with more information. PW