On Thu, 25 Jun 2026 20:58:25 GMT, Dean Long <[email protected]> wrote:
>> David Simms has updated the pull request with a new target base due to a >> merge or a rebase. The pull request now contains 2847 commits: >> >> - Merge branch '8317277' into jep401_sub_review_8317278 >> - Merge remote-tracking branch 'valhalla/lworld' into 8317277 >> - 8387024: [lworld] Remove the mention of mutexes from the specification >> text of value classes >> >> Reviewed-by: alanb, dholmes, liach >> - 8387054: [lworld] BACKOUT JDK-8375694 C2: Dead loop constructed with >> CastPP in late inlining >> >> Reviewed-by: dholmes >> - 8386904: [lworld] ProblemList tests that fail due to JDK-8375645 on all >> platforms >> 8387050: [lworld] ProblemList two >> compiler/valhalla/inlinetypes/TestValueClasses.java sub-tests in Xcomp mode >> >> Reviewed-by: dholmes >> - 8386638: [lworld] Port JEP 401 to PPC64 >> >> Co-authored-by: Richard Reingruber <[email protected]> >> Reviewed-by: dbriemann, rrich, lfoltan >> - Merge >> >> Merge jdk-28+3 >> - 8386815: [lworld] Reduce the number of constructors in IdentityException >> >> Reviewed-by: liach, alanb >> - 8386626: [lworld] Only enable valhalla IR tests in >> ScalarReplacementWithGCBarrierTests.java for aarch64 and x86 >> >> Reviewed-by: mhaessig, mchevalier >> - 8386718: [lworld] C2: Integer overflow in arraycopy scaling for flat >> arrays >> >> Reviewed-by: qamai >> - ... and 2837 more: https://git.openjdk.org/jdk/compare/915efc50...5bc36893 > > src/hotspot/share/adlc/formssel.cpp line 850: > >> 848: strcmp(_matrule->_opType,"SafePoint" )==0 || >> 849: strcmp(_matrule->_opType,"Halt" )==0 || >> 850: strcmp(_matrule->_opType,"CallLeafNoFP")==0) > > CallLeafNoFP has its own match_edge() override, so this change looks like an > optimization. Shouldn't we add other missing Call nodes here too, to make > the optimization complete? That change comes from: https://mail.openjdk.org/archives/list/[email protected]/thread/CNM2Y3TPOGJCTOGKZOW3QQWVT2GAYOHQ/ (thanks @TobiHartmann for digging that out) This part: > C2 doesn't support calling to an address that is loaded in a register. I hacked support for it from the existing CallLeafNoFP node. It's somewhat ugly but I would say it's good enough for now. (still good enough 10 years later, it seems) Removing it causes some c2 crashes. Let me add a comment. ------------- PR Review Comment: https://git.openjdk.org/jdk/pull/31122#discussion_r3518671727
