This is the mail archive of the
mailing list for the GCC project.
Re: [tree-ssa] SWITCH_EXPR lowering
- From: Andrew MacLeod <amacleod at redhat dot com>
- To: Richard Henderson <rth at redhat dot com>
- Cc: Jan Hubicka <jh at suse dot cz>, Zdenek Dvorak <rakdver at atrey dot karlin dot mff dot cuni dot cz>, gcc-patches <gcc-patches at gcc dot gnu dot org>, Jeff Law <law at redhat dot com>, Diego Novillo <dnovillo at redhat dot com>
- Date: 01 Nov 2003 15:25:58 -0500
- Subject: Re: [tree-ssa] SWITCH_EXPR lowering
- References: <1067017615.14175.3094.camel@p4><20031024175758.GB17291@atrey.karlin.mff.cuni.cz><1067030697.14175.3476.camel@p4><20031024213156.GA22864@atrey.karlin.mff.cuni.cz><1067032171.21257.3504.camel@p4> <1067032627.14404.3520.camel@p4><20031025221955.GA16508@atrey.karlin.mff.cuni.cz><20031026080601.GA31855@redhat.com><20031101141341.GA7363@atrey.karlin.mff.cuni.cz><20031101155500.GE14974@kam.mff.cuni.cz> <20031101195019.GA11339@redhat.com>
On Sat, 2003-11-01 at 14:50, Richard Henderson wrote:
> On Sat, Nov 01, 2003 at 04:55:00PM +0100, Jan Hubicka wrote:
> > I was just playing around the idea of replacing these low level control
> > branches by specialized nodes, like COND_JUMP taking arugment of
> > condition and two CFG edges. In such was we avoid a lot of unnecesary
> > memory references (the goto statements and labels).
But then we are adding uneccessary clutter to the IL. For the moment I
think we're better off simply sticking with the GENERIC COND_EXPR and
leaving the gotos the way they are. I dont like the idea of adding yet
another COND_EXPR type tree node that has the same information but in a
different format. I dont think COND_EXPR's are the root of our memory
problems, and although you might save a bit, in the overall scheme I
dont think you save enough to make it worthwhile.
Ditching CE nodes and other mechanisms yet to be determined are more
likely to be profitable avenues.