02:28
@mappo Unless you're running CHECK in a tight loop, my hunch says probably not.
It's true that the localized dfns need to be fixed at every call to CHECK, but I'd wager that the allocation overhead isn't disproportionately higher than whatever else CHECK is doing.
(Though my grasp of interpreter overhead stuff is a bit shaky)
Is it feasible for you to directly test? I'd be tempted to grab hard numbers: a←⍳1e5 ⋄ cmpx 'origCHECK¨a' 'noDfnsCHECK¨a' or whatever.
