Speedup Matrix inverse

[Corey-Maler]

Remove function call gives up to 30% of speedup. See comparison https://jsfiddle.net/xqs0st3c/

[bhouston]

Nice.

[mrdoob]

Should we do the same with Matrix4.getInverse()?

[Corey-Maler]

Ow, I didn't expect a much speed up in Matrix4, but tests (https://jsfiddle.net/xqs0st3c/1/) shows that in Matrix4 it is even faster (on my computer 1500ms vs 1000ms thats ~35%).

P.S. In google chrome on Mac OS.

But code is little ugly now here.

[mrdoob]

Thanks!

[Usnul]

for those that are curious why this happens. There are 2 read passes over elements array in the original implementation (one in main body of getInverse and one in multiplyScalar), also if you do an "instanceof" check instead of "elements.length === 9" you get numbers more like these:

Matrix3 test
 T…E.Matrix3 {elements: Float32Array[9]}
old: 438.133ms
new: 347.131ms

which is more like a 20% saving and not 30, still super useful though :)