Issues on zooming
romiras opened this issue · 2 comments
romiras commented
evanw commented
You're seeing the limits of floating-point precision on your GPU. It's a fundamental limitation of my approach (compiling the equation to GLSL). GPUs prefer speed over accuracy so floating-point precision breaks down pretty quickly. The pattern will look different on different GPUs because it's implementation-dependent.
I considered trying to compile equations into code that emulates double-precision numbers using tuples of two single-precision numbers to avoid this limitation but that seemed hard.
romiras commented
Thanks for explanation.