Closed xiaowuc1 closed 5 years ago
I feel like the "fast path" for this (10e-6?) should use regular floats for speed, and switch to Decimal
when needed.
Adding a flag that defaults to fast
and uses floats in that case, and the Decimal
otherwise seems like a reasonable approach.
Most floating-point checkers do not depend on this, however it can avoid issues when the precision guarantee needed is very tight.