This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Some thoughts and quetsions about the data flow infrastracture
- From: kenner at vlsi1 dot ultra dot nyu dot edu (Richard Kenner)
- To: zadeck at naturalbridge dot com
- Cc: gcc-patches at gcc dot gnu dot org, iant at google dot com, seongbae dot park at gmail dot com, vmakarov at redhat dot com
- Date: Wed, 14 Feb 2007 09:41:16 EST
- Subject: Re: Some thoughts and quetsions about the data flow infrastracture
- References: <45D0CF0D.3010708@redhat.com> <45D0D621.9060602@codesourcery.com> <45D13DB1.1020601@redhat.com> <45D14269.5050901@codesourcery.com> <45D1E67D.2090501@redhat.com> <571f6b510702130911m18e9dd00j18a522be5192735a@mail.gmail.com> <45D20FFC.7000906@redhat.com> <ab3a61990702131309q662e59c8y623d690cf767e26f@mail.gmail.com> <45D22DB8.5020102@redhat.com> <ab3a61990702131342s709d18e4id994c64e5a47fded@mail.gmail.com> <45D23F9F.4070409@naturalbridge.com> <10702141343.AA07027@vlsi1.ultra.nyu.edu> <45D31426.8020700@naturalbridge.com>
> Actually i really did mean changing. The style of changes that
> something like combine does where it creates a bunch of subtrees and
> tries them on and then backs them out is going to be difficult to
> handle. You are going to end up churning up the df representation quite
> a bit.
Not really. All those changes are purely internal and shouldn't result
in any DF changes whatsoever.
I think this is what you're saying below, but the model I'd have is that
combine is a blackbox that takes two or three insns and either says "I
can't do anything with this" or "here are one or two replacement insns".
In the latter case, the DF infrastructure would update itself for that
replacement.