RFR(S): 8229483: Sinking load out of loop may trigger: assert(found_sfpt) failed: no node in loop that's not input to safepoint

Nils Eliasson nils.eliasson at oracle.com
Tue Sep 24 07:54:50 UTC 2019


Still good!

// Nils

On 2019-09-24 09:43, Roland Westrelin wrote:
> Thanks for the review, Nils.
>
> Valhalla testing found a similar issue: the load is first sunk into an
> inner strip mined loop where there's an anti dependent store, then the
> store becomes a candidates to be moved after the loop and finally the
> load is scheduled in the outer strip mined loop and verification code
> fires. The logic I proposed before doesn't help because the load is
> first moved into the inner loop. Here is a fix that covers both cases:
>
> http://cr.openjdk.java.net/~roland/8229483/webrev.01/
>
> Roland.


More information about the hotspot-compiler-dev mailing list