|basic question on register allocation firstname.lastname@example.org (2008-02-20)|
|Re: basic question on register allocation email@example.com (Max Hailperin) (2008-02-24)|
|Re: basic question on register allocation firstname.lastname@example.org (2008-02-25)|
|Re: basic question on register allocation email@example.com (2008-02-28)|
|Date:||Thu, 28 Feb 2008 21:50:42 -0800 (PST)|
|Posted-Date:||29 Feb 2008 01:23:09 EST|
Thanks both for your input.
My misunderstanding was that I was thinking once it is determined that
a variable is spilled, all the interferences go away. I re-read my
compiler book (Muchnick's book) and indeed there is a loop that
repeats when allocation fails after spiling.
I've read Hack's work recently that is based on the fact that
interference graphs in SSA form are chordal and the chromatic number
of them can be determined in polynomial time. It is very interesting
to see there are still ways to improve on existing solutions to an old
problem like register allocation.
Return to the
Search the comp.compilers archives again.