Re: [RFC] 69526 - ivopts candidate strangeness

2016-03-20 Thread Richard Biener
On Thu, Mar 17, 2016 at 5:31 PM, Robin Dapp wrote: > The attached patch is a first and somewhat hideous attempt to fix the > missed optimization discussed here: > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69526 > > where (paraphrasing) a spurious > n = n - 1 > n = (sizetype) n > n = n + 1 > is

[RFC] 69526 - ivopts candidate strangeness

2016-03-20 Thread Robin Dapp
The attached patch is a first and somewhat hideous attempt to fix the missed optimization discussed here: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69526 where (paraphrasing) a spurious n = n - 1 n = (sizetype) n n = n + 1 is being created. The patch tries to avoid this on tree level as well a