I suspect it's really slowing down MantleDiver, which means it currently does not scale well. We should profile to confirm though.
I also suspect a lot of it is due to lambda/helper function specialization, so start sprinkling @nospecialize everywhere inside the macro implementation and see how it improves.
I suspect it's really slowing down MantleDiver, which means it currently does not scale well. We should profile to confirm though.
I also suspect a lot of it is due to lambda/helper function specialization, so start sprinkling
@nospecialize
everywhere inside the macro implementation and see how it improves.