Fast sine/cosine for ARMv7+NEON

I also posted this on StackOverflow so people could test it on real hardware.

Here is a function I wrote to compute sines and cosines really fast on the NEON vector FPU (iPhone 3GS, Pandora, BeagleBoard, …). I used a method described by Nicolas Capens and the fact that cos(x)=sin(x+90deg). Nils Pipenbrinck was kind enough to time it on his BeagleBoard and produced these results. Your mileage may vary.