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 RFA: Increase support for restrict qualifier


Richard Henderson <rth@redhat.com> writes:

> On Sat, Oct 01, 2005 at 09:53:53PM -0700, Ian Lance Taylor wrote:
> > > I'd hope it's possible to write a testcase for the optimization 
> > > improvements in the present patch as well.
> > 
> > I don't know.  The real optimization improvement is in the scheduler
> > which will no longer see dependencies which it formerly saw.
> 
> It's usually easier to write aliasing test cases with copy/constant
> propagation and dead code elimination.  The general form is
> 
> 	x = 0;
> 	y = 1;
> 	if (x != 0)
> 	  link_error ();
> 
> where x and y are the two things you want to affirm do not alias.

I thought about that, but I think that will effectively wind up
testing the tree layer.  At least, it will do so eventually if not
now.  My patch was really aimed at the RTL layer.  It would be quite
possible to pass the test above while still failing to consider alias
analysis when doing instruction scheduling.

So, I dunno.

Ian


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