It is unclear to me if this "optimization" can be done with Chebyshev but at least it gives the wrong values now (see the added test I made by just asserting that the optimized min computation gave the same result as the one from scratch) so disable it for now.
Have to benchmark that pulling this computation out into its own function didn't make things worse.
It is unclear to me if this "optimization" can be done with Chebyshev but at least it gives the wrong values now (see the added test I made by just asserting that the optimized min computation gave the same result as the one from scratch) so disable it for now.
Have to benchmark that pulling this computation out into its own function didn't make things worse.