Difference between revisions of "Performance/Meetings/2008 08"

Latest revision as of 23:12, 17 March 2010

Performance/Meetings/2009/08/28

Meeting Minutes
IRC Meeting of Sun Microsystems (StarOffice) with RedFlag2000 and the Community ??
Performance Project

Date: 2008/08/28
Time: 15:57– 16:58
Meeting No.:

Agenda:

(3:57:39 PM) liheng: kuangliang:please call guoyongshi to attend this meeting
(3:57:40 PM) liheng: :)
(3:58:50 PM) liheng: xiuzhi:guo yongshi will join us to improve performance of performance of OO
(3:59:08 PM) xiuzhi: liheng,guoys:welcome
(3:59:09 PM) guoys: :)
(3:59:27 PM) liheng: guoys::)
(3:59:45 PM) mhu: Hi all
(3:59:53 PM) xiuzhi: hi all
(4:00:00 PM) liangjun: hello all
(4:00:01 PM) liheng: Hi matthias
(4:00:04 PM) yugq: hi
(4:00:06 PM) guoys: hi
(4:00:46 PM) liheng: mhu:let me introduce new person Guo Yongshi (guoys)
(4:01:14 PM) liheng: he joined us to improve performance of OO
(4:01:43 PM) mhu: Hi guoys, nice to meet you.
(4:01:49 PM) guoys: me too
(4:01:59 PM) liheng: guoys:please introduce yourself :)
(4:02:25 PM) liheng: guoys: please introduce yourself
(4:02:28 PM) guoys: you know my english is poor
(4:03:30 PM) liheng: okay,
(4:03:55 PM) liheng: mhu:gouys has worked for CH2000 for 7 years,
(4:04:52 PM) mhu: okay, so gouys has quite some experience, I guess.
(4:05:01 PM) liheng: and he is a sinor architect of OO & Linux
(4:06:16 PM) liheng: he also participate in the RedOffice on MID Device
(4:06:33 PM) mhu: okay
(4:07:17 PM) liheng: he knew VCL and Framework very well too, so he can help us to build the big picture of OOo's performance
(4:07:49 PM) mhu: yes, I think he will be a valueable addition to the team.
(4:08:55 PM) liheng: Maybe two person will join us in next month, they will create new tools for performance:)
(4:09:54 PM) liheng: okay
(4:10:02 PM) mhu: okay, yes support for tooling sounds good
(4:10:43 PM) liheng: continue the discussion of tools for performance
(4:11:25 PM) liheng: I think we need 4 kinds tools,...
(4:11:56 PM) liheng: 1.Test (Benchmark,profiling) Tools
(4:12:13 PM) liheng: 2.Raw data storage
(4:12:21 PM) liheng: 3.statistical tools
(4:12:32 PM) liheng: 4.report and rule of development.
(4:13:21 PM) liheng: and for Teat tools, we sparate them to basic function testing and application function testing
(4:14:58 PM) liheng: application function testing tools like the tools Malte gave us, that can run some case ,to show macroscopic value
(4:15:41 PM) mhu: I agree, and think this list (with 4 kinds) looks okay
(4:17:07 PM) liheng: and basic function testing tools will mark inner logical and function to show all key part in macroscopic function
(4:17:10 PM) mhu: this is basically what we discussed last I time, if I remember correctly.
(4:19:30 PM) liheng: yes, so we discussed it and think for inner function we must use tool-code(like RTL_XXX_LOG and new support we will provide)...
(4:20:24 PM) mhu: yes, the RTL_LOG... facility is also part of what Malte did, right?
(4:21:04 PM) liheng: and for macroscopic value, we design a framework of tools to support a flexible tools set
(4:21:31 PM) mhu: okay, fine.
(4:22:07 PM) liheng: do we need open source of our new tools
(4:22:10 PM) liheng: ?
(4:23:44 PM) mhu: well, that is what you need to decide. If the tools should be of maximum value to all developers, it is most often useful to open source them...
(4:24:17 PM) xiuzhi: +1
(4:24:27 PM) mhu: services that you provide, e.g. "2. raw data storage", of course cannot and need not be open sourced
(4:25:22 PM) mhu: ...also, "4. reporting..." could be a service that you provide, and no code to open source...
(4:25:27 PM) liheng: If we can open source, we can support more difference kinds test for all developers, that sounds good
(4:26:36 PM) mhu: ...so, it depends on what it is. If it is a useful tool where source code is involved, I think open source is usually better. Other can then help to develop and maintain those tools.
(4:27:23 PM) liheng: okay,:), We think open source of test tools is necessary and that can it eary to maintain specail extension for every CWS in future
(4:27:37 PM) mhu: if it is more a "service" (or "software as a service", SaaS) so that it can be remotely accessed, it need not be opensource.
(4:27:57 PM) mhu: liheng: okay, then we agree :-)
(4:29:07 PM) liheng: we can do some work for SaaS for fans :), and open source to all developers
(4:31:15 PM) liheng: In the view of code level support, i think we need more support on categorization and sorting
(4:32:20 PM) mhu: "categorization and sorting" of what?
(4:32:27 PM) liheng: If we build a picture that descript some key path of some macroscopic value,we must check it individual
(4:33:46 PM) liheng: eg. this benchmark is show xml parsing in loading and another is for saving (4:34:04 PM) mhu: ah, yes. okay.
(4:35:19 PM) liheng: so, we maybe format the log file and build tool to fetch a set of data
(4:36:01 PM) liheng: and put these data into statistic tool
(4:36:24 PM) liheng: like T-Test
(4:36:45 PM) mhu: log file? RTL_LOG_...? yes, that is good. This needs LOG statements in the code that you want to analyze.
(4:37:23 PM) mhu: yes, I can look up the (T-Test) code that I wrote some years ago, and send it to you.
(4:39:09 PM) liheng: thank you
(4:39:28 PM) liheng: i remember it code in Java
(4:40:12 PM) mhu: yes, but as it was translated from FORTRAN, we translate it from Java to C/C++ also :-)
(4:40:48 PM) mhu: It is only number crunching. that works in every programming language that I speak :-)
(4:41:40 PM) liheng: okay, i know FORTRAN,Java,C++,perl very well, on problem
(4:41:49 PM) mhu: and the code always looks like FORTRAN ...
(4:42:37 PM) mhu: okay. it is also not so much code (few hundred lines, if I remember correctly) (4:42:39 PM) liheng: but, FORTRAN is very well for math , i very like it in college
(4:43:47 PM) mhu: same for me; so we may conclude our age from our knowledge of FORTRAN ? :-)
(4:44:47 PM) liheng: Okay, of course . Do you think we create statistic tools in Java is a good idea, that they can run cross OS
(4:45:55 PM) mhu: yes, that is why I rewrote it in Java; also to be easy re-used in a webapplication container, and so on.
(4:47:47 PM) mhu: also, we could turn the code into an OOo-extension, and Java makes that multi-platform also.
(4:48:34 PM) liheng: Yes,it's I want to do!, we can start from pair of Test tools and Statistic Tools, make data format easy to use
(4:48:53 PM) mhu: (if that statistical analysis code would make sense as an extension (to Calc for example)).
(4:49:25 PM) liheng: and create a serial of templates in Calc for report them
(4:49:55 PM) mhu: I think we could not agree more, fine :-)
(4:52:36 PM) liheng: next step, we are going to design the framework for appliction function tools and find the key path for old hotspot, load/save startup etc
(4:52:36 PM) liheng: .
(4:55:31 PM) liheng: mhu:that all for me
(4:55:45 PM) mhu: yes, that would be the next step (sorry for slow response; someone is simultaneoulsly debugging a crash in my office, and oinvolving me in dsicussions...)
(4:56:24 PM) mhu: okay, I dont have more for today, also.
(4:56:46 PM) liheng: Okay, see you next week!
(4:57:02 PM) xiuzhi: see you
(4:57:10 PM) mhu: okay, yes see you next week.
(4:57:15 PM) mhu: bye all
(4:57:18 PM) liheng: bye
(4:57:22 PM) liangjun: bye
(4:57:24 PM) guoys: bye
(4:57:36 PM) kuangl: bye-bye
(4:58:08 PM) yugq: bye

Go back