Re: Layer speedup

[Jiří Techet <techet gmail com>]

On Fri, Sep 17, 2010 at 10:25, Patrick Salecker <patty9999 gmx de> wrote:
> Am 25.08.2010 22:38, schrieb Jiří Techet:
>>
>> On Sun, Aug 22, 2010 at 11:12, Russell Strong<russell strong id au>
>>  wrote:
>>
>>>
>>> Hi,
>>>
>>> The following patch replaces marker sorting ( by latitude ) with
>>> clutter's set_depth to achieve the same result.
>>>
>>> Using a modified animated-marker demo with 484 markers, this patch
>>> reduces CPU using on my laptop from 80% to 19%.
>>>
>>> There is still more that can be done, but that's now in clutter, which
>>> appears to be sorting a list every time a depth is changed.
>>>
>>> Any thoughts?
>>>
>>
>> Nice! I've just applied it. I haven't measured the performance myself
>> but I believe your findings are correct and it's a nice simplification
>> as well. The only thing I have changed is that the code didn't take
>> into account the southern hemisphere where things are opposite - the
>> more you go to the south, the farther the markers should be. But this
>> bug was already present before your patch, I've just noticed it now.
>>
>> Thanks!
>>
>> Jiri
>>
>
> Will these patches also be in the next 0.6 release?
> I tried to add 3000 marker to a layer with libchamplain 0.6.1 and it took
> unusable 110 seconds.
> After applying the patch it decreased to only 8 seconds.

There are no critical bugs in 0.6 that need to be fixed so actually I
didn't plan to make more 0.6 releases.

Regarding the performance I'm now wondering if we should make the
ordering at all if it's so expensive - the 8 seconds for 3000 points
are quite a lot. I'll do some benchmarking myself and decide
afterwards.

Thanks,

Jiri


> _______________________________________________
> libchamplain-list mailing list
> libchamplain-list gnome org
> http://mail.gnome.org/mailman/listinfo/libchamplain-list
>