The original implementation used euler-to-quat and then quat-to-matrix (which is how D3DXMath implemented this method as well). The result used 48 vector ops.
Doing this computation in closed-form directly as euler-to-matrix only takes around 19 vector ops.
The original implementation used euler-to-quat and then quat-to-matrix (which is how D3DXMath implemented this method as well). The result used 48 vector ops.
Doing this computation in closed-form directly as euler-to-matrix only takes around 19 vector ops.