forked from mirrors/0ad
b10534d224
transpose. This assumes orthogonal transformations (which we have exclusively, as far as I can tell), but is significantly faster, because calculating the inverse is slow. This was SVN commit r3041.