------- Comment #11 from dean at arctic dot org 2008-02-19 17:42 -------
Subject: Re: [4.0/4.1/4.2/4.3 Regression] jump to
middle of loop on entry with using old version of an variable
On Mon, 19 Feb 2008, xinliangli at gmail dot com wrote:
> Note that assignment of s0 = s in the loop is mostly dead except for the last
> occurence. So it should be optimized into:
>
> do {
> s += *d;
> ++d;
> } while (s < g);
> v[0] = (s-*(d-1));
fwiw the real loop i pulled this fragment from has a bit more dependent on
s0 ... i just used this fragment because it showcased a particular problem
i observed in the full loop. i can't include the full loop because it's
in proprietary code :(
-dean
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27986
