limit check pushed inside loop
Stephen Weeks
MLton@sourcelight.com
Wed, 9 Aug 2000 11:28:01 -0700 (PDT)
> O.K. I only mentioned it because from a performance point of view, a
> limitCheck with zero bytes can simply compare frontier with limit without
> some calculation (whether it be via leal or movl/addl).
I wouldn't have thought it would make a big difference, but ... I just ran
"even 750000000" with the zero byte check and with the n byte check. Here are
the running times.
zero byte 8.5
n byte 10.4
That's a huge difference. I guess we should put limit checks on loop exits.
Actually, I guess we only really want them if the loop isn't already allocating
anyways, but I don't see an easy way to do that in the current world.