|
From: Bryan M. <om...@br...> - 2006-08-17 21:02:10
|
Josef, had a question from one of the guys at work today asking about a "compare two trees" feature that was planned for kcachegrind at some point. As I understand it, it's for showing the difference in performance between two runs, particularly for showing the changes between two versions of the software under test. Is this implemented/planned ? TIA, Bryan "Brain Murders" Meredith |
|
From: Josef W. <Jos...@gm...> - 2006-08-17 22:20:15
|
On Thursday 17 August 2006 23:01, you wrote: > Josef, > > had a question from one of the guys at work today asking about a > "compare two trees" feature that was planned for kcachegrind at some point. > > As I understand it, it's for showing the difference in performance > between two runs, particularly for showing the changes between two > versions of the software under test. > > Is this implemented/planned ? Implemented: no. Planned: Not really sure. I am not convinced that some kind of difference view is easier to understand than 2 DAGs near to each other. How would you do it? Josef > > TIA, > Bryan "Brain Murders" Meredith > > |
|
From: Bryan M. <om...@br...> - 2006-08-18 17:45:11
|
What he was suggesting was a simple difference - literally subtracting one set of results from another one (having carefully lined up various call trees etc. as required and where possible) It's main use would be to make any performance impacts / regressions from one software version to another _really_ stand out. They have tried two sessions side by side but the increase distributed across a set of functions was too small to stand out of the overall picture, although the impact of the actual difference in itself is quite measurable. It might be better to get Andy to join the list so he can explain it properly (he is in a different team to me so I'm not directly involved in the task). On the other hand, I know that we both feel cheeky for asking as neither of us could help to implement it. Bryan Josef Weidendorfer wrote: > On Thursday 17 August 2006 23:01, you wrote: >> Josef, >> >> had a question from one of the guys at work today asking about a >> "compare two trees" feature that was planned for kcachegrind at some point. >> >> As I understand it, it's for showing the difference in performance >> between two runs, particularly for showing the changes between two >> versions of the software under test. >> >> Is this implemented/planned ? > > Implemented: no. Planned: Not really sure. I am not convinced that > some kind of difference view is easier to understand than 2 DAGs near > to each other. How would you do it? > > Josef > > >> TIA, >> Bryan "Brain Murders" Meredith >> >> > > ------------------------------------------------------------------------- > Using Tomcat but need to do more? Need to support web services, security? > Get stuff done quickly with pre-integrated technology to make your job easier > Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo > http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642 > _______________________________________________ > Valgrind-developers mailing list > Val...@li... > https://lists.sourceforge.net/lists/listinfo/valgrind-developers > |
|
From: Josef W. <Jos...@gm...> - 2006-08-18 18:55:16
|
On Friday 18 August 2006 19:44, you wrote: > What he was suggesting was a simple difference - literally subtracting > one set of results from another one (having carefully lined up various > call trees etc. as required and where possible) It could be interesting. However, to visually stand out it probably would be good to have the call arcs drawn with different widths (that has to wait for the Qt4 port...). Wouldn't it be enough to have a difference column in the function list on the left? > in the task). On the other hand, I know that we both feel cheeky for > asking as neither of us could help to implement it. Thanks for any further ideas. I simply pick out the ones I like best ;-) Josef |