Performance/Meetings/2008 02
|
|
|---|
|
Quick Navigation Team Communication Activities |
| About this template |
Performance/Meetings/2008/02/28
Meeting Minutes
IRC Meeting of Sun Microsystems (StarOffice) with RedFlag2000 and the Community ??
Performance Project
Date: 2008/02/28
Time: 16:30– 17:24
Meeting No.:
Agenda:
(16:31:39) LiHeng: hello
(16:31:52) LiHeng: waiting....
(16:32:14) LiHeng: Everybody English now!
(16:32:28) kuangliang: Hello guys
(16:34:57) LiHeng: Did you revc the agenda?
(16:35:11) yugq: yes
(16:36:41) mhu: Hi all; sorry for being late.
(16:37:24) yugq: hi
(16:37:35) liangjun: mhu: hello
(16:37:51) LiHeng: hello !
(16:38:28) mhu: I hope, you're all well, and had good new year holydays.
(16:38:54) LiHeng: mhu:yes , thank you
(16:39:31) LiHeng: mhu:did you get the agenda?
(16:40:36) LiHeng: mhu:Do you have some topics you want ?
(16:40:40) mhu: LiHeng: yes, and now I've also read it. looks fine for me.
(16:41:06) mhu: No additional topics.
(16:41:32) LiHeng: let's begin from How to issue the profiling reports made by Valgrind to all developer.:)
(16:41:59) LiHeng: yugq:please:)
(16:42:04) yugq: OK
(16:43:10) yugq: The output data from valgrind profiling is difficult to read. But we can read cachegrind and callgrind output data with KCachegrind.
(16:44:04) yugq: Memcheck output data is better than callgrind data, but still bad.
(16:44:07) mhu: yes, i've used kcachegrind as well
(16:44:52) LiHeng: mhu:but kcachegrind can run only on the linux
(16:45:18) yugq: The problem is how to put these data to wiki to share with others and compare them with each other.
(16:45:42) mhu: LiHeng: yes, but valgrind is Linux only, too. Or do I misunderstand something?
(16:45:43) yugq: One way is using snapshot pictures.
(16:46:21) LiHeng: mhu:if we publish the data with analysis reports , can all developers read and use it to compare?
(16:46:55) mhu: Oh, you're speaking of the raw data?
(16:47:22) yugq: I think most of performance optimization achieves are available to other platform either.
(16:47:27) yugq: mhu: yes.
(16:47:46) mhu: yes, raw data will be difficult to interpret for other devolpers.
(16:48:30) yugq: how we compare data with every profiling and let others know that.
(16:49:35) LiHeng: so, do developers of OO need a tools to anlysis the raw data anywhere?
(16:50:18) mhu: I think, we need to "reduce" the (amount of) data to "some" numbers first, before we (a) present them, and (b) compare them with other profiling runs.
(16:50:58) mhu: ...where we still would need to define what "reduce" and "some" actually should mean.
(16:51:16) LiHeng: Are there many develops need raw data ?
(16:51:37) yugq: mhu: yes.
(16:52:15) mhu: LiHeng: the general developer should be able to follow our conclusioins, I think, but does not necessarily need to have all the same tools (only if she wants to reproduce our conclusions).
(16:53:21) LiHeng: In Beijing we will make anlysis report by the raw data with our point, but i think other developers would want to use the data for other views
(16:53:22) mhu: As to publishing the (maybe raw) data...
(16:53:28) yugq: Presenting key data is a good idea.
(16:54:24) mhu: ...I would think, that we could use the tools.openoffice.org website to upload the data, and then put a link to them into the wiki.
(16:54:32) LiHeng: yugq:yes; )
(16:55:03) LiHeng: mhu:ok
(16:55:30) yugq: mhu: ok
(16:55:58) mhu: yes, I think providing the data for others to reproduce is fine.
(16:56:20) mhu: s/reproduce/reproduce our conclusions/
(16:56:38) LiHeng: we will anlyze the raw data for one or two weeks , and publish both of reports and raw data on wiki
(16:56:57) mhu: okay, fine with me.
(16:57:52) mhu: If you want, you could send the raw data to me, so that I can join you and have a look at the data as well.
(16:57:53) LiHeng: mhu: did you watch the wiki page recently?
(16:58:13) mhu: LiHeng: no, sorry. I've not have a look at wiki.
(16:58:21) yugq: mhu: OK, we need your help :)
(16:58:53) LiHeng: mhu:please help us to check wiki, we do some change;)
(16:59:54) LiHeng: go topic 2?
(17:00:16) mhu: LiHeng: okay, I'm now looking at the wiki page. The top of the page already looks much cleaner than before, I think.
(17:00:35) mhu: yes, we can go to topic 2.
(17:00:59) LiHeng: thank you!
(17:01:17) LiHeng: 2.Start to document for profiling tools
(17:01:57) yugq: I have a question for this topic
(17:02:12) LiHeng: yugq:please:)
(17:02:25) LiHeng: yugq:sorry
(17:03:04) yugq: will the document for profiling tools be presented on tools project or performance project?
(17:04:53) yugq: mhu said the profiling data would be presented on tools project.
(17:06:00) mhu: oh, yes. I would suggest to place documents / data / ... onto the tools.openoffice.org (project) website, and descriptive text onto the wiki.
(17:06:18) yugq: mhu: OK, thanks!
(17:06:38) LiHeng: yugq:have more?
(17:06:41) mhu: If the documents are "only" text, they could go into the wiki, of course.
(17:07:18) mhu: ...but images, odt, pdf, ... are better uploaded to the project website.
(17:08:00) yugq: mhu: yes.
(17:08:11) yugq: LiHeng: no more
(17:09:03) LiHeng: mhu: i have send a email to matle to ask profiling-tools and source of them, but he did not reply
(17:09:41) mhu: LiHeng: okay, I'll talk to Malte, and ask him to reply (at least).
(17:10:25) LiHeng: mhu:thank you,
(17:11:20) LiHeng: we will make some profiling data on OO2.3.1with those tools, but did OO have some already?
(17:11:23) mhu: LiHeng: with "profiling tools and source" you mean "automated CWS performance test", yes?
(17:12:32) LiHeng: mhu: so yes ,but and the monitor built by Matle
(17:12:58) mhu: LiHeng: okay, yes Malte did build them.
(17:13:21) LiHeng: we hope to document for those tools
(17:13:58) mhu: ...and, no, I don't think that we save the results of the automated performance tests.
(17:13:59) LiHeng: and make report with them
(17:14:31) mhu: ...in particular not any details (like cachegrind logfiles).
(17:15:12) yugq: mhu: yes.
(17:15:15) mhu: LiHeng: yes, that sounds like a good plan (document tools, and make reports, as you said).
(17:15:38) LiHeng: mhu: we want to develop base those tools, and get more useful data
(17:16:21) mhu: yes, understood. I'll talk to Malte.
(17:17:16) LiHeng: and that , we will make report with both of common profile tools and private tools every time.
(17:18:01) LiHeng: mhu: publish the result ,and source only in Beijing and Berlin
(17:18:23) LiHeng: sorry Hamburg:)
(17:19:26) mhu: well, berlin is only ~280km from hamburg :-)
(17:19:40) yugq: :-D
(17:20:49) liangjun: :)
(17:20:50) LiHeng: :) so , please help us get the source as quickly as possible, we want to start this work at March 15,
(17:21:15) mhu: yes, I'll see what I can do.
(17:21:40) LiHeng: mhu:thank you very much!
(17:22:00) mhu: anything more for today ? (I am somewhat in a hurry, sorry)
(17:22:08) mhu: LiHeng: you are welcome.
(17:22:16) LiHeng: that all, for me
(17:22:25) LiHeng: other ones?
(17:22:29) yugq: no more, for me
(17:22:56) liangjun: no thing
(17:23:09) mhu: okay, I send email when I know more. Otherwise, see you next Thursday. Good bye.
(17:23:21) LiHeng: mhu: next time we will dicuss severl key point about performance of OO
(17:23:33) mhu: LIHeng: okay.
(17:23:35) LiHeng: see you next week,
(17:23:38) LiHeng: bye
(17:23:44) yugq: bye
(17:23:45) mhu: bye all.
(17:24:00) liangjun: bye
