RFR(M): 8223363: Bad node estimate assertion failure

Patric Hedlin patric.hedlin at oracle.com
Tue May 28 16:48:31 UTC 2019


Missing webrev with test-case for JDK-8223502:

Webrev: http://cr.openjdk.java.net/~phedlin/tr8223502/


Best regards,
Patric

On 2019-05-28 12:35, Patric Hedlin wrote:
> Dear all,
>
> I would like to ask for help to review the following change/update:
>
> Issue:  https://bugs.openjdk.java.net/browse/JDK-8223363
> Webrev: http://cr.openjdk.java.net/~phedlin/tr8223363/
>
> 8223363: Bad node estimate assertion failure
>
>     Tightening node budget estimates. New loop clone estimate 
> attempting to
>     approximate complex loop cases (slightly) more accurately. Still 
> ad-hoc.
>
>
> Also addressed:
>
> Issue:  https://bugs.openjdk.java.net/browse/JDK-8223502
> Issue:  https://bugs.openjdk.java.net/browse/JDK-8224648
>
> 8223502: Node estimate for loop unswitching is not correct: 
> assert(delta <= 2 * required) failed: Bad node estimate
> 8224648: assert(!exceeding_node_budget()) failed: Too many NODES 
> required! failure with ctw
>
>
> Testing: hs-tier1..7, hs-precheckin-comp, Lucene (CTW)
>
> Caveat:  Multiple failures present in several tiers.
>
>
> Best regards,
> Patric
>


More information about the hotspot-compiler-dev mailing list