User:Rl

I'm riccardo lucchese a student attending the Università degli Studi di Padova (Italy). I'll be an intern at olpc during 30th June - 5th September. I'll be working on my proposal "Profiling the Sugar framework" mentored by Tomeu Vizoso.

email: firstname.lastname gmail com

irc: rl in #sugar

Done

14th July - 18th July

• Written some scripts to automate the instrumentation work on the XO

• Got more confident with sugar's code base

• Build some test-cases to do profiling work on (+ reports)

 1. `Activities Switching'
 2. `Views/Icons rendering'
 3. `Quick-tabbing'

• Investigated what could be done to solve perfs problems for the test cases at the previous point

• Started working on two more perfs reports

• Continued development of Picker/Grapher

 Bug fixes and added some features for future work on profiling
 sugar's boot timings.  

6th July - 11th July

• Continued development of Picker/Grapher

 (http://lists.laptop.org/pipermail/devel/2008-July/016290.html)
 Added some features

• Got more confident with sugar's code base

 Studied the activities-launching process

 Studied the code involved in activities switching

• Profiled activities switching using cProfile and Picker

 From the activity point of view, the task is dominated by the fact
 that they are saving their state even if it didn't changed
 (most of time goes in taking screenshots, copying buffers from the X server
 and scaling them to obtain the preview).
 
 The Journal and the DS also play a big role here (next things to look at)
 

• Investigate if something could be done to help with point 3 (from the activities point of view)

 One can bypass some gtk functions by directly working on the buffer returned
 by the X server but that would add some extra complexity for little gain
 as those operation are expensive anyway and activities shouldn't
 save their state (when it doesn't change) in the first place.
 

• Got in contact with Zach Riggle (gsoc student) that is developing `sugarbot'

 It would be great to use sugarbot also for perfs automated regression tests.

30th June - 4th July

• Discussed with my mentor what exactly our goals are

 Think at the kids !

• Discussed what tools are needed

 Most profiling tools we have already.

 We need one more tool to investigate if there are any non
 obvious interferences between processes
 (a bit like http://www.bootchart.org/images/bootchart.png).

 A script for parsing sugar's logger.debug calls so that we can automate 
 performance regression tests.

• Started developing the first one of those tools we want but we don't have

 It's almost finished, only needs some retouches

• Got more confident with sugar code base

 Ui particularly

• Searched trac for perfs problems

 Some were also suggested from devel@