Re: [platform-releng-dev] Performance re-testing on releases and 3.1 mil

[Tobias Widmer <tobias_widmer@xxxxxxxxxx>]

I have a question regarding the measure used in the assertions. If we do not explicitly specify whether to take CPU Time or Elapsed Process, which one is used as default?

 

Tobias

 

Dienstag, 12. April 2005 23:17
To: platform-releng-dev@xxxxxxxxxxx
cc:
From: Sonia Dimitrov <Sonia_Dimitrov@xxxxxxxxxx>
Subject: [platform-releng-dev] Performance re-testing on releases and 3.1   milestones

Starting this Friday, we would like to run performance tests on previous 
releases (starting at 2.1.3) and 3.1 milestones to generate a new history of 
performance measurements on four newly configured performance test machines. 
 These will consist of two identical "slower" machines (Intel P4 2 GHz, 512 
MB RAM) and two identical "fast" machines (Intel P4 3 GHz, 2 gig RAM) which will 
run Windows XP (SP1) and Red Hat Enterprise Linux WS 3.  This task will 
require approximately 80 hours to complete and if all goes smoothly, we expect 
to have all the re-testing done by Wednesday April 20.   

Please see 
http://dev.eclipse.org/viewcvs/index.cgi/%7Echeckout%7E/platform-releng-home/performance_test_plan.htm 
for further details and background. 

In order for us to proceed, we ask ALL teams using CPU Time in assertions 
and summaries (ie. fingerprints) to switch to using Elapsed Process in HEAD and 
perf_* streams of your tests where needed by 11pm (Ottawa time) Thursday, April 
14. 

Please respond to this list with 
questions or comments. 

Sonia