[11u] RFR: 8059241: C2: Excessive RemoveUseless passes during incremental inlining

[Lemmond, Dan]

I've applied the patches for both of the bugs in the bugtail and both of them apply cleanly. If this gets approved, they'll need to be pushed together.

JBS Issues:
https://bugs.openjdk.java.net/browse/JDK-8221592
https://bugs.openjdk.java.net/browse/JDK-8223581

I'll make note of this when I make the "[11u] Fix Request" comment.

Roland, if you need anything from me please let me know.

Thanks,

Dan
 

On 3/26/21, 2:42 AM, "jdk-updates-dev on behalf of Andrew Haley" <jdk-updates-dev-retn at openjdk.java.net on behalf of aph at redhat.com> wrote:

    CAUTION: This email originated from outside of the organization. Do not click links or open attachments unless you can confirm the sender and know the content is safe.



    On 3/26/21 9:21 AM, Roland Westrelin wrote:
    >
    >>> I’d like to request a review for this backport of JDK-8059241
    >>> JBS: https://bugs.openjdk.java.net/browse/JDK-8059241
    >>> Webrev: http://cr.openjdk.java.net/~dlemmond/8059241/webrev.11u.00/
    >>
    >> Ouch. We'd need a C2 specialist to have a very close look at the risk
    >> of backporting this.
    >
    > This should mostly affect workloads that make heavy use of
    > invokedynamic. That alone would decrease the risk. The fix has been
    > integrated a while ago so it's well tested. I would say the change
    > itself is clearly not trivial but fairly low risk.

    Okay. Will you review this, please? And please consider the likely
    performance gain v.s churn and risk.

    --
    Andrew Haley  (he/him)
    Java Platform Lead Engineer
    Red Hat UK Ltd. <https://www.redhat.com>
    https://keybase.io/andrewhaley
    EAC8 43EB D3EF DB98 CC77 2FAD A5CD 6035 332F A671