This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: predicate aware uninitialized variable analysis
- From: Richard Guenther <rguenther at suse dot de>
- To: Xinliang David Li <davidxl at google dot com>
- Cc: Andrew Pinski <pinskia at gmail dot com>, Richard Guenther <richard dot guenther at gmail dot com>, GCC Patches <gcc-patches at gcc dot gnu dot org>
- Date: Tue, 12 Oct 2010 11:26:49 +0200 (CEST)
- Subject: Re: predicate aware uninitialized variable analysis
- References: <522e93240907221419y2d8dbdbdta7343148338688cc@mail.gmail.com> <q2r522e93241004131223scbd7cc8fx283be7d728f7ef58@mail.gmail.com> <h2k84fc9c001004140344of207d33av33ab3a99e58d9fdd@mail.gmail.com> <AANLkTi=yfeEdyU2n3HL7cnoEa7E8bmbY7y8+ESzeAj_P@mail.gmail.com> <AANLkTinEULS8moG8PKHEAW4Kd5NnB87UfiZT6Z7jR_S0@mail.gmail.com>
On Mon, 11 Oct 2010, Xinliang David Li wrote:
> On Mon, Oct 11, 2010 at 11:14 PM, Andrew Pinski <pinskia@gmail.com> wrote:
> > On Wed, Apr 14, 2010 at 5:44 PM, Richard Guenther
> > <richard.guenther@gmail.com> wrote:
> >> Can you instead of returning a sbitmap from compute_uninit_opnds_pos
> >> use an unsigned int mask (with proper checks for PHIs with too many
> >> operands which you don't want to handle because of complexity
> >> reasons anyway)?
> >
> > Looks the proper checks for PHIs with too many operands which you
> > don't want to handle was never done. ?And the current proposal on the
> > table to fix that does not skip them but rather just truncates them.
>
> To make it stable across hosts, I can change it back to use unsigned
> and do the skipping as originally suggested. I will wait a little
> longer for more comments (Richard) before making that change just in
> case -- in particular which one is more preferred: using wider type
> vs. consistency across hosts.
Well, you can simply use unsigned HOST_WIDEST_FAST_INT and skip for
more than 32 PHI arguments (HOST_WIDEST_FAST_INT is at least 4 bytes
on all supported hosts).
Richard.