Compressed Quats

gistfile1.cpp

// To store a quat in 64 bits, we find which element is the largest and use 2 bits to
// store which one it is. We now know the other 3 elements fall in the range
// of [-kOneOverRootTwo, kOneOverRootTwo]. We scale that range across 20/21/21 bits
// and store each. When extracting, we use the fact that the quat was normalized 
// to compute the 4th element.

...

// To store a quat in 32 bits, we find which element is the largest and use 2 bits to
// store which one it is. We now know the other 3 elements fall in the range
// of [-kOneOverRootTwo, kOneOverRootTwo]. We scale that range across 10 bits
// and store each. When extracting, we use the fact that the quat was normalized 
// to compute the 4th element.