My bet is that he is using a continuos transformation and the guys who first did this (at least as far as I know) reproduced and completed the original transfor

My bet is that he is using a continuos transformation and the guys who first did this (at least as far as I know) reproduced and completed the original transformation grid that Escher did. http://escherdroste.math.leidenuniv.nl/index.php?menu=symmetry http://escherdroste.math.leidenuniv.nl/index.php?menu=symmet... The transformation is a conformal map so it should not deform angles. Other explanation is that the fragment shader is losing information in the process and that float numbers are not precise enough. Still this is very impressive