A factor of 2.+. (We could provide a type system, take away recursion, and replace it with a structural induction form. It would be impossible to write infinite loops.)
On Nov 10, 2010, at 12:41 PM, John Clements wrote: > > On Nov 10, 2010, at 6:50 AM, Matthias Felleisen wrote: > >> >> Your reasoning is correct. What's the performance hit? > > In order to try it, I used the legendary 12.4.2, and... uh, it seg faulted. > > Okay, I submitted a bug report on that. Moving right along: > > Testing in the simplest way, here's what I got for permute of size 9: > > > Debugging enabled, no loop protection: > cpu time: 2520 real time: 2539 gc time: 1554 > cpu time: 1504 real time: 1523 gc time: 542 > cpu time: 1515 real time: 1534 gc time: 552 > > Debugging enabled, with loop protection: > cpu time: 4496 real time: 4532 gc time: 1940 > cpu time: 3653 real time: 3694 gc time: 1079 > cpu time: 3692 real time: 3730 gc time: 1112 > > Debugging disabled, no loop protection: > cpu time: 2181 real time: 2200 gc time: 1651 > cpu time: 1132 real time: 1152 gc time: 597 > cpu time: 1143 real time: 1163 gc time: 603 > > Debugging disabled, with loop protection: > cpu time: 3458 real time: 3494 gc time: 2007 > cpu time: 2434 real time: 2471 gc time: 972 > cpu time: 2424 real time: 2461 gc time: 1001 > > ... so it's quite a bit slower. Note that a whole bunch of this slowdown in > the "debugging enabled" category is presumably due to the errortrace > annotation of the macro expansion. I'm guessing that folding this into the > errortrace expansion would reduce its overhead quite a bit. > > John > _________________________________________________ For list-related administrative tasks: http://lists.racket-lang.org/listinfo/dev
