This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.

Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]
Other format:	[Raw text]

Re: [PATCH] Fix PR middle-end/28690, indexed load/store performance + reload bug

From: Paolo Bonzini <paolo dot bonzini at lu dot unisi dot ch>
To: Ian Lance Taylor <iant at google dot com>
Cc: Peter Bergner <bergner at vnet dot ibm dot com>, GCC Patches <gcc-patches at gcc dot gnu dot org>
Date: Tue, 05 Dec 2006 15:24:54 +0100
Subject: Re: [PATCH] Fix PR middle-end/28690, indexed load/store performance + reload bug
References: <20061205050808.GA13002@vervain.rchland.ibm.com> <457572D1.8070801@lu.unisi.ch> <m3irgq8lwh.fsf@localhost.localdomain>

And in fact I realized that if you fix LEGITIMIZE_MEMORY_ADDRESS, you
probably don't need this patch anyhow.

I am not sure. Actually, I am pretty sure the canonicalization which is the basic idea of this patch pays off on other platforms. But to do it properly, we first need to make sure GCC does not needlessly pessimize code that, when canonicalized this way, would undo the other canonicalization in swap_commutative_operands_with_target (for example). Then we can do much more, for example by boosting the precedence of REGs vs. MEMs (something that right now pays off IIRC on PowerPC but -- I'm sure about this -- loses on x86).

Paolo

References:
- [PATCH] Fix PR middle-end/28690, indexed load/store performance + reload bug
  - From: Peter Bergner
- Re: [PATCH] Fix PR middle-end/28690, indexed load/store performance + reload bug
  - From: Paolo Bonzini
- Re: [PATCH] Fix PR middle-end/28690, indexed load/store performance + reload bug
  - From: Ian Lance Taylor

Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]