It is pretty well known that Coalton numerics are not very performant, even in monomorphic. The main reason is that methods don't get inlined, causing a lot of consing of double-float code. We might get a lot of mileage out of Lisp-inlining some choice methods in the standard library until another solution is achieved.
This issue is to track floating-point numerical performance, and closing this issue is gated on reaching performance parity with Lisp for similar code.
stylewarning
commented
1 month ago
It is pretty well known that Coalton numerics are not very performant, even in monomorphic. The main reason is that methods don't get inlined, causing a lot of consing of double-float code. We might get a lot of mileage out of Lisp-inlining some choice methods in the standard library until another solution is achieved.
This issue is to track floating-point numerical performance, and closing this issue is gated on reaching performance parity with Lisp for similar code.