This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Loop optimizer patch


Ulrich Weigand wrote:

Mark,



It makes sense to me -- but I'm concerned that it's introducing a new case of quadratic behavior in the compiler.



Well, there are already several other routines in loop.c that are quadratic in the number of movables: ignore_some_movables, combine_movables, move_movables.


None of these have any sort of loop bound; do you still think it
is necessary to introduce one for force_movables?

I think so. We really don't want quadratic behavior in the compiler, except where absolutely necessary. Sometimes you can't get really get around it, but I'm not familiar enough with the code to which you're referring to know whethere we should be bounding those other cases or not. You can pick a bigger value than 5 if you want for the default, but I think there should be a bound.

Thanks,

--
Mark Mitchell
CodeSourcery, LLC
(916) 791-8304
mark@codesourcery.com


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]