This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] Teach VRP to derive ranges from SWITCH_EXPR (fix PR21258)
On Thu, 12 Apr 2007, Diego Novillo wrote:
> Richard Guenther wrote on 04/12/07 10:35:
>
> > Bootstrapped and tested on x86_64-unknown-linux-gnu. I will apply
> > this to mainline tomorrow.
>
> Did you compare compile times vs optimization as I asked you on IRC?
There's a 1% slowdown for compiling tree.i with an unoptimized, checking
enabled stage1 compiler.
The patch will be on the C++ tester tonight to gather
more numbers. GCC of course has lots of switch statements but also
benefits most from this optimization.
> > *************** vrp_visit_cond_stmt (tree stmt, edge *ta
> > *** 4759,4766 ****
> >
> > *taken_edge_p = NULL;
> >
> > ! /* FIXME. Handle SWITCH_EXPRs. But first, the assert pass needs to
> > ! add ASSERT_EXPRs for them. */
> > if (TREE_CODE (stmt) == SWITCH_EXPR)
> > return SSA_PROP_VARYING;
> >
> > --- 4887,4893 ----
> >
> > *taken_edge_p = NULL;
> >
> > ! /* FIXME. Handle SWITCH_EXPRs. */
> > if (TREE_CODE (stmt) == SWITCH_EXPR)
> > return SSA_PROP_VARYING;
>
> Why not handle SWITCH_EXPR here?
I think that's an orthogonal problem -- will the propagator engine
be able to handle SWITCH_EXPR if I just touch this piece of VRP?
Richard.
--
Richard Guenther <rguenther@suse.de>
Novell / SUSE Labs
SUSE LINUX Products GmbH - Nuernberg - AG Nuernberg - HRB 16746 - GF: Markus Rex