This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: SSA for Trees - PRE implementation
On Jan 8, 2004, at 2:46 PM, Sidney Bennett wrote:
From: "Daniel Berlin" <dberlin@dberlin.org>
I've recently attempted an SSA PRE implementation based on the
Kennedy, et.al. paper and met with significant difficulties.
Non-surprising :P
Daniel,
Quick question then, did you find the SSA PRE implementation any
better than
the lazy code motion (LCM) and Morel-Renvoise algorithms that are
already
part of GCC?
Besides the fact that they work on SSA and a completely different level?
Our existing implemenations are in the backend, working on RTL.
the SSAPRE implementation works on trees, in our middle end.
In the end, will it have been any easier to implement, will it
be any faster or produce better code?
Certainly not easier to implement, but it should produce faster and
better code.
The straight line strength reduction i'm working on will also improve
code.
I guess the real question is ... Do
you see it eventually replacing the either existing PRE pass?
Not unless we don't produce any partially redundant expressions in the
backend.
Sid