RFR(XS): 8028198: SIGSEGV in PhaseIdealLoop::build_loop_late_post

Igor Veresov igor.veresov at oracle.com
Tue Nov 12 15:19:15 PST 2013

Looks good.


On Nov 12, 2013, at 5:30 AM, Rickard Bäckman <rickard.backman at oracle.com> wrote:

> Hi,
> can I have this change reviewed please?
> In some cases the optimizer decided to split a MathExact node through a
> Phi creating an invalid pattern where there is a Phi between the
> MathExact and its projections. The fix is to not allow split_thru_phi
> for MathExact nodes.
> Webrev: http://cr.openjdk.java.net/~rbackman/8028198/
> Bug: https://bugs.openjdk.java.net/browse/JDK-8028198
> Thanks
> /R

More information about the hotspot-compiler-dev mailing list