[aarch64-port-dev ] RFR: Bulk Shenandoah integration, 2019-09-09

[Andrew John Hughes]

On 08/09/2019 19:34, Aleksey Shipilev wrote:
> Hi,
> 
> Please review the routine integration of Shenandoah updates to aarch64-port/jdk8u-shenandoah. Since
> this involves the cross-repository pull, the webrev is hard to generate without including merge
> fluff. Current integration basically picks up everything here:
>   https://builds.shipilev.net/patch-openjdk-shenandoah-jdk8/hotspot/
> 
> Only hotspot is needed to update. Everything else is in perfect sync already.
> 
> Changesets to be pushed:
>   https://cr.openjdk.java.net/~shade/shenandoah/merges/aarch64-port-8u-20190909/changesets.txt
> 
> Squashed patch to cross-check what is being pushed:
>   https://cr.openjdk.java.net/~shade/shenandoah/merges/aarch64-port-8u-20190909/all-diffs.patch
> 
> Tour of changes:
> 
> *) assembler_x86.(cpp|hpp), macroAssembler_x86.cpp, templateTable_x86_64.cpp removes
> jccb/jmpb_if_possible, and returns affected chunks to upstream state.
> 
> This code was added for Shenandoah store checking code long time ago, and was never used after
> Shenandoah parts were removed. See the current difference against upstream here, we are reverting
> some of those:
> 
> https://builds.shipilev.net/patch-openjdk-jdk8-aarch64/hotspot/src/cpu/x86/vm/assembler_x86.hpp.sdiff.html
> 
> https://builds.shipilev.net/patch-openjdk-jdk8-aarch64/hotspot/src/cpu/x86/vm/assembler_x86.cpp.sdiff.html
> 
> https://builds.shipilev.net/patch-openjdk-jdk8-aarch64/hotspot/src/cpu/x86/vm/templateTable_x86_64.cpp.sdiff.html
> 
> https://builds.shipilev.net/patch-openjdk-jdk8-aarch64/hotspot/src/cpu/x86/vm/macroAssembler_x86.cpp.sdiff.html
> 
>  *) x86_64.ad reverts chunks to upstream state. immU8 was used to match Shenandoah gc-state checks,
> and now handled elsewhere. Whitespace changes were the leftovers from previous removals and/or
> backports. See the current difference against upstream here, we are removing some of those:
>   https://builds.shipilev.net/patch-openjdk-jdk8-aarch64/hotspot/src/cpu/x86/vm/x86_64.ad.sdiff.html
> 
>  *) c1_LIR.(cpp|hpp) protects lir_shenandoah_wb, like it does in 11+
> 
>  *) shenandoahRootProcessor.(cpp|hpp), shenandoahSynchronizerIterator.(cpp|hpp), synchronizer.cpp,
> thread.cpp: removes ShenandoahFastSyncRoots, refactors it a bit, and reverts ObjectSynchronizer to
> upstream state.
> 
> This one is actually a subtle bugfix disguised as the reversal to upstream state, see:
>   https://mail.openjdk.java.net/pipermail/shenandoah-dev/2019-July/010124.html
> 
> See the current difference against upstream here, we are reverting some of those chunks:
> 
> https://builds.shipilev.net/patch-openjdk-jdk8-aarch64/hotspot/src/share/vm/runtime/synchronizer.cpp.sdiff.html
> 
> https://builds.shipilev.net/patch-openjdk-jdk8-aarch64/hotspot/src/share/vm/runtime/synchronizer.hpp.sdiff.html
> 
> https://builds.shipilev.net/patch-openjdk-jdk8-aarch64/hotspot/src/share/vm/runtime/thread.cpp.sdiff.html
> 
>  *) arguments.cpp: drops UseNUMAInterleaving in Shenandoah-specific initialization block. It is in
> shenandoahArguments.cpp in other releases, but 8u is not luxurious enough to have GC-specific
> argument handling yet.
> 
>  *) bitMap.inline.hpp: since JDK-8211926 landed via backports, this assert should be generic, and
> not Shenandoah-specific. This change reverts it to upstream state. See the current difference with
> upstream, we are reverting the assert to upstream state:
> 
> https://builds.shipilev.net/patch-openjdk-jdk8-aarch64/hotspot/src/share/vm/utilities/bitMap.inline.hpp.sdiff.html
> 
> Testing: Linux {x86_64, aarch64} hotspot_gc_shenandoah, regular build and tests on all platforms
> 

Looks good to me. Good to see the diff from upstream being reduced.

Thanks,
-- 
Andrew :)

Senior Free Java Software Engineer
Red Hat, Inc. (http://www.redhat.com)

PGP Key: ed25519/0xCFDA0F9B35964222 (hkp://keys.gnupg.net)
Fingerprint = 5132 579D D154 0ED2 3E04  C5A0 CFDA 0F9B 3596 4222
https://keybase.io/gnu_andrew