[Sugar-devel] [Dextrose] [PATCH v5 sugar] Pulsing icon delayed by 5 seconds or so SL#2080

[Martin Dengler]

I'm sorry to harp on this, but I hope you can see it actually takes
quite some effort to communicate this stuff well...

On Thu, Oct 28, 2010 at 09:27:18PM +0530, Anurag Chowdhury wrote:
> I have attached the pulsingicon.py file

Thanks - very useful to understand what you're timing.

> in which I placed the benchmark script

which benchmark script?!  Another file you should include.

 to test the time taken by update function everytime
> it is called.
> I used the time.time() function to measure the time taken to process
> the update step.

There is a better way: use cProfile and gprof2dot.py.  You will get
graphs (and of course the raw data) like this:

http://www.martindengler.com/tmp/sl.o-2080/pulsingicon.py-stats-graph.png

...and then you could show us before and after your change on the XO
1.0 and XO 1.5.  Normally the relative time is most interesting, but
since we are also interested in the absolute running time I patched
gprof2dot.py to show that, and you can see the total time (in seconds)
consumed by each graph node and its subtree as the second-to-last
number shown in the node.

You can find the pulsingicon.py changes and the instructions I used to
make that in:
http://www.martindengler.com/tmp/sl.o-2080/README.txt

> And If you may compare the odd steps of both logs i.e the numbers in
> the lines having "seconds taken to update frame" appearing at the odd
> steps
> (3rd ,5th ,7th.....entries) and similarily compare  the even steps
> you may find the average time taken on XO-1.5 to run the update
> function was nearly 50% faster.

Odd steps?  Even steps?  Sounds scary.  Why?

Martin
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 190 bytes
Desc: not available
URL: <http://lists.sugarlabs.org/archive/sugar-devel/attachments/20101028/a20c9afa/attachment.pgp>