This is the mail archive of the
mailing list for the GCC project.
Re: [tree-ssa] Do alias analysis after SSA. Improvements to PR8361.
- From: Diego Novillo <dnovillo at redhat dot com>
- To: Jeff Law <law at redhat dot com>
- Cc: "gcc-patches at gcc dot gnu dot org" <gcc-patches at gcc dot gnu dot org>
- Date: Thu, 08 Jan 2004 09:35:49 -0500
- Subject: Re: [tree-ssa] Do alias analysis after SSA. Improvements to PR8361.
- Organization: Red Hat Canada
- References: <200401080312.i083COGx001633@speedy.slc.redhat.com>
On Wed, 2004-01-07 at 22:12, email@example.com wrote:
> Very cool. Presumably the next step is using the nature of the SSA
> graph to provide a rudimentary context sensitive alias analysis?
Ah, yes. Now that I re-read my message I never quite specified what I
had in mind. Yes, the next step is to walk use-def chains at pointer
dereference sites to collect points-to information.
In principle, the SSA web gives us points-to information for free. It
all depends on how much walking we're willing to do on the use-def
links. We could collect context sensitive points-to information by
keeping alias information tied to SSA names. Alternately, we can
aggregate that into context insensitive sets per DECL.
> Or do you have other infrastructure that has to go in first?
The only other piece of infrastructure is the use-def chain walker,
which is just a few lines of code. The big change was moving alias