Performance issues

[Zero]

Hi,
After long time i am back to mobile mapping project with tile5.By
the way, I managed to get a working version of wms layer.
it doesnt seem very stable but useable, i can share it if you want.
The thing i want to discuss is the performance issue of the latest
builds, i use v0.4.0. its faster then any mobile javascript mapping
library. I also compared its performance with the openlayers after
their codesprint. v0.4.0 has some bugs, rendering issues but it is
fast and smooth . I upgraded my project to 0.9.4.1, codebase is more
elegant but its suffering. Pan and zoom are flickering. I tried
0.9.3.1 too.

Are there any issues or you know about this situation.

[Damon Oehlman]

Absolutely interested in seeing what you have done with a WMS layer.  If you have done the ground work it should be pretty simple for me to integrate that into the source as an extra engine.

Have you tried the latest branch in master?  I would be interested how that goes around performance, although I haven't done any side by side benchmarks my impression is that it is a little faster than the previous versions.  While I think my intentions were good around unifying the various timer events with the "COG.Loopage" functionality it isn't quite as fast as simple setTimeout calls.  

Also the latest version in master implements the requestAnimationFrame hooks instead of setTimeout so it will be even faster on browsers that support it (unfortunately I no mobile browsers I have tried do yet, but soon, hopefully).

Is there a particular device that you are working with, if so I might be able to have a bit more of a look here and see what I can find.  Performance is something I am very keen to do the best job of as possible (hence many of the changes over time) so I'm very interested in getting it right.

Cheers,

Damon.

--
Damon Oehlman
http://distractable.net/

[Damon Oehlman]

One other thing that I should mention is that there are definitely a couple of tricks that could be done to further improve performance, but it would mean trading off against other features (such as how well transparency was supported).

These aren't things that I would enable by default, but perhaps a configuration switch when creating a map could be specified to enable "high performance" mode if it's something that is desired.

With regards to this point having an understanding on people's general use cases would be great, i.e: 

• non-transparent tiles, drawn overlays only (with no transparency)
• non-transparent tiles, but transparent image markers (and other overlays)
• transparent tiles
  

Cheers,

Damon.

[Baris Goral]

• Well, i havent tested on latest branch but i will soon and share my results. I'm developing my application on an ipad. In my opinion performance is the key point for tile5. 

• I use nontransparent tiles, transparent markers and overlays.

• I tested my wms engine with demis.nl it didnt work :) as expected, we use reverse axis metric projected wms server. demis is different so tile scheme is different too. I think i can implement a general approach and share with you.

you can look an unpublised demo at http://bit.ly/hbvvb2 

Thanks

Baris

[Damon Oehlman]

Hey Baris,

Thanks for that extra info - I might have a look at what I can do about implementing some of those optional performance improvements.  Given you are using non-transparent tiles there is definitely an opportunity to experiment with various canvas drawing modes to increase performance.

I'll aim to have something up in the dev branch (https://github.com/sidelab/tile5/tree/dev) a little later today that you will be able to do some tests on.

I might even try and put a performance test suite together as I would definitely like to gather performance data where possible.

Cheers,

Damon.