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] Make VRP recognize range tests (unsigned)a + CST <= CST2


On Mon, 12 Feb 2007, Jeffrey Law wrote:

> On Fri, 2007-02-09 at 10:32 +0100, Richard Guenther wrote:
> > On Thu, 8 Feb 2007, Diego Novillo wrote:
> > 
> > > Richard Guenther wrote on 02/06/07 07:56:
> > > > This patch makes VRP recognize range and anti-range tests.  It does so
> > > > by allowing more flexible ASSERT_EXPRs (which is neccessary for the
> > > > anti-range case) like
> > > > 
> > > >  ASSERT_EXPR <a_1, (unsigned)a_1 + 5 > 13>
> > > > 
> > > > and derive ranges and anti-ranges from such directly.
> > > > 
> > > I don't really like where this is going.  VRP does not need more complexity.
> > > Though it triggers 30,000 times more, are there any perceivable benefits in
> > > runtime or code size?
> > 
> > See my followup mail - no earthshaking improvements.  But one would need
> > to test with bounds-checking code.
> I must have missed the follow-up.  What exactly do you mean by 30k
> triggers?  Is that 30k asserts generated or 30k transformations using
> the result of the new asserts?  The latter would definitely be in the
> "wow" category.

Heh, no ;)  It generates 30k asserts.  The followup said the patch
results in 0.003% overall size reduction of stage2 gcc/*.o.  I expect
this to slightly improve with followup patches improving the handling
of anti-ranges.

Richard.

-- 
Richard Guenther <rguenther@suse.de>
Novell / SUSE Labs


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