Fixes extreme slowness of Geometry.rayIntersection.

[ChrisDenham]

You are all going to love this pull request!

It fixes an extreme inefficiency in 'Geometry.rayIntersection' caused by using the 'indices' property rather than the '_indices' member. This was making three copies of the entire index array for EVERY triangle.

Ray intersections with complex models is now actually usable.

Enjoy!

Chris.

[makc]

ouch

https://github.com/ChrisDenham/Alternativa3D/blob/5b16652f83d8c8f8da394ece5a27671dfad0fe4c/src/alternativa/engine3d/resources/Geometry.as#L109

and here I thought that positionBuffer.readFloat()-s were the problem :) 👍

[gonchar]

thank you! really helpful

[ChrisDenham]

Yeah, me too. I was beginning to think about implementing a mechanism to retain/cache a copy of Vector array so readFloat didn't need to bother converting 4 byte floats to 8 byte 'Number'/double every time. Then I thought..... hey hold on a minute.... surely that's not soooo bad, there's a single microprocessor instruction to convert float to double!

[makc]

@ChrisDenham there is whole branch with this mechanism

[ChrisDenham]

@makc Thanks. I had a vague recollection of you mentioning something about this before in relation to ray intersection, but couldn't find the reference. Interesting to note that that 'VectorGeometry' branch would 'proportionally' benefit even more from this pull request than the 'master' branch.

[makc]

...it is almost 200 commits behind the master. not going to happen :)

[gonchar]

also my pull request is about vector geometry :) #59