Re: Graph coloring, patents

In article <1989Nov15.010209.11362@esegue.segue.boston.ma.us> Preston Briggs <preston@rice.edu> writes:
>I don't know which is more effective. I suspect (with no evidence) that
>Chow's method will be better on constrained machines (not many registers) and
>Chaitin's better otherwise. Neither seems worthwhile without some global
>optimization to provide grist for the mill.


Chow's method is always at least as good as Chaitin's graph coloring
(and sometimes better). It is likely that it will only be significantly
better on machines *without* lots of registers. That's because both of
these methods try to avoid spills. If you have a zillion registers,
you never spill with either method, so they're equal in that case.


// rfg

Return to the comp.compilers page.
Search the comp.compilers archives again.