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 PR37053: Move tweaks of commutative precedence to target hook

From: Paolo Bonzini <bonzini at gnu dot org>
To: Ulrich Weigand <uweigand at de dot ibm dot com>
Cc: Peter Bergner <bergner at vnet dot ibm dot com>, Maxim Kuvyrkov <maxim at codesourcery dot com>, gcc-patches <gcc-patches at gcc dot gnu dot org>, Ian Lance Taylor <iant at google dot com>, Luis Machado <luisgpm at linux dot ibm dot com>, rdsandiford at googlemail dot com
Date: Wed, 05 Aug 2009 17:24:26 +0200
Subject: Re: [PATCH] Fix PR37053: Move tweaks of commutative precedence to target hook
References: <200908051457.n75EvnNp013352@d12av02.megacenter.de.ibm.com>

I'm wondering if we cannot make a simple local fix to the problem
that "validate_replace_rtx" cannot be completely undone by doing
another "validate_replace_rtx".  We should be able to keep the
logic in reload_as_needed as-is, but simply use the existing
change-group management routines to cancel or confirm the change
after we did the extra test ...

The patch below implements this option.  Any comments?  Can you
verify this fixes the original problem?

If it works (which I'd be pretty confident of), I can only feel stupid. :-)

Paolo

References:
- Re: [PATCH] Fix PR37053: Move tweaks of commutative precedence to target hook
  - From: Ulrich Weigand

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