On Wed, Jan 23, 2019 at 09:23:53AM +0100, Richard Biener wrote:
> On Tue, 22 Jan 2019, Jakub Jelinek wrote:
>
> > Hi!
> >
> > SCEV can analyze not just integral/pointer IVs, but (scalar) float ones as
> > well. Calling build_int_cst on such types results in ICE, build_zero_cst
> > works. Though the loop invariant PHI IVs, if we represent them as using
> > +0.0 step, aren't correct if honoring signed zeros, as + 0.0 will make
> > a +0.0 out of -0.0.
> >
> > Bootstrapped/regtested on {x86_64,i686,powerpc64{,le}}-linux, ok for trunk?
>
> OK.
Thinking about it, if we honor SNaNs, we should punt too, even if we are not
honoring signed zeros, because sNaN + 0.0 artificially added there will
raise an exception, where previously the code could not raise it.
Fixed thusly, bootstrapped/regtested on x86_64-linux and i686-linux, ok for
trunk?
2019-01-24 Jakub Jelinek <ja...@redhat.com>
PR tree-optimization/88964
* gimple-loop-interchange.cc (loop_cand::analyze_induction_var): Also
punt if HONOR_SNANS (chrec).
--- gcc/gimple-loop-interchange.cc.jj 2019-01-23 09:34:53.739433699 +0100
+++ gcc/gimple-loop-interchange.cc 2019-01-23 12:53:35.215848790 +0100
@@ -690,8 +690,8 @@ loop_cand::analyze_induction_var (tree v
{
/* Punt on floating point invariants if honoring signed zeros,
representing that as + 0.0 would change the result if init
- is -0.0. */
- if (HONOR_SIGNED_ZEROS (chrec))
+ is -0.0. Similarly for SNaNs it can raise exception. */
+ if (HONOR_SIGNED_ZEROS (chrec) || HONOR_SNANS (chrec))
return false;
struct induction *iv = XCNEW (struct induction);
iv->var = var;
Jakub
