Thanks for the pull! So the main performance win here is calling factories later and 2 times less checks for extending bounds, right?

Yes, and the removing of Math.max and Math.min has also improved it a bit.

I just remember trying to optimize by removing Math.min/max calls (in some other code) and then coming to conclusion that it makes almost no difference. But maybe I'm wrong. Generally we'd need to find a compromise between performance and simplicity/readability/terseness of the code for this to get into core.

I updated jperfs page and the Math.min/max version is in green :

The biggest difference is with Internet Explorer. For the other browsers, it something like 10-15% slower.

Do you want a pull request with the Math.min/max version ?

This is a significant performance pit when using markerclusterer. I'd love this to be pulled into 0.8.

@danzel ^^

Yep, I like this :) We use extend A LOT in MarkerCluster.

I saw your re-factoring on JsPerf with properties checks instead of instanceof checks. I opted for instanceof because it's faster on iOS7 and equivalent with a desktop computer :-)

Cool, yeah I just wanted to check for myself :)

OK, going to make an attempt at simplifying this a bit soon and then merge.

OK, managed to simplify the code a bit with 665e766 and merged manually: https://github.com/Leaflet/Leaflet/commits/master

Thanks again, great work!

Awesome, next release is going to be intense!

@danzel wanna measure perf improvement in MarkerCluster?

Will do Monday when I'm on a real PC

Oh, thanks for the merge. I'm happy to know I have 20 lines of code in this wonderful project :-)

Looks like a minor improvement

Numbers are the time to cluster 50000 markers.
Top numbers are chrome. Bottom are IE11.
% is the the % time the profiler says is spent in this function (measured separately)

Still a nice improvement.