Related articles |
---|
copy propagation? ppanda@Rational.Com (Partha Sarathi Panda) (2001-02-12) |
Re: copy propagation? apiron@ulb.ac.be (Anthony PIRON) (2001-02-15) |
From: | Anthony PIRON <apiron@ulb.ac.be> |
Newsgroups: | comp.compilers |
Date: | 15 Feb 2001 00:42:07 -0500 |
Organization: | VUB/ULB Computing Centre |
References: | 01-02-055 |
Keywords: | optimize |
Posted-Date: | 15 Feb 2001 00:42:07 EST |
Hi,
It's seem simple may be I'm wrong.
if you have got something like:
a0=a1
...
=a0
you can replace every occurrence of a0 by a1.
e.g.
a0=a1 // now deadcode (if no side effects on the a0 variable)
suppress this instruction
...
=a1
if you obtain for some phi node all the same right-hand parameters do
the same transformation for this phi node
e.g.
a0=a1 ...
\ /
a2=phi(a0,a1)
...
= a2
becomes to
a0=a1 ... // now dead code (a0-a1)
\ /
a2=phi(a1,a1) // now dead code
...
= a1
Theses things can be done during sparse representation building. And the
dead code is never introduced (immediately removed).
Pariah Sarathi Panda wrote:
> How can copy propagation be done using sparse data-flow information
> and without doing register renaming?
What's the signification of "without register renaming". There is
always renaming in copy propagation.
--
Anthony PIRON - Researcher
Universite Libre de Bruxelles, Department of Computer Science
CP 212 - Boulevard du Triomphe, B-1050 Bruxelles, Belgium
Tel. +32 2 650 50 55, Fax +32 2 650 56 09, GSM +32 478 530 022
Email: apiron@ulb.ac.be Web: http://www.ulb.ac.be/di/ssd/apiron
Return to the
comp.compilers page.
Search the
comp.compilers archives again.