This is the mail archive of the
mailing list for the GCC project.
Re: [PATCH] Splitting of virtual definitions (aka. V_MUST_DEF)
- From: Diego Novillo <dnovillo at redhat dot com>
- To: Brian Booth <bbooth at redhat dot com>
- Cc: GCC-Patches <gcc-patches at gcc dot gnu dot org>, Richard Henderson <rth at redhat dot com>
- Date: Tue, 08 Jun 2004 08:09:54 -0400
- Subject: Re: [PATCH] Splitting of virtual definitions (aka. V_MUST_DEF)
- Organization: Red Hat Canada
- References: <firstname.lastname@example.org>
On Mon, 2004-06-07 at 09:45, Brian Booth wrote:
> At the moment, VDEFs are used to mark statements that either define a
> non-GIMPLE register or could possibly define one. This patch splits up
> these two cases. It replaces the VDEF with two conceptually different
> constructs: V_MAY_DEF and V_MUST_DEF. These definitions are added to
> statements that may define a non-GIMPLE register and statements that
> have a killing definition of one respectively. Another way of looking at
> it is that the V_MAY_DEF is simply VDEF with a new name while V_MUST_DEF
> is a VDEF without the implied use of its operand.
Thanks Brian. I've started looking at the patch now.
> WITH_PATCH WITHOUT_PATCH
> real 48m24.896s 48m18.499s
> user 37m57.030s 37m37.460s
> sys 6m10.790s 6m5.690s
Could you test the patch against POOMA
(http://www.tat.physik.uni-tuebingen.de/~rguenth/gcc/monitor-summary.html) and PR8361? Both in compile time and generated code.