RFR (S) 8078438: Interpreter should support conditional card marks (UseCondCardMark)
Aleksey Shipilev
aleksey.shipilev at oracle.com
Thu Apr 23 22:22:28 UTC 2015
Hi,
I would like to propose a patch to support UseCondCardMark support in
interpreter. At this point, only C2 has the proper support for it. C1
changes went into hs-comp today.
Bug:
https://bugs.openjdk.java.net/browse/JDK-8078438
Webrev:
http://cr.openjdk.java.net/~shade/8078438/webrev.01/
Along with the actual change, I also cleaned up the store_check split in
MacroAssembler -- it does not seem to be doing anything. AArch64 would
probably need a mirror fix.
Testing:
- Linux/x86_64/release build
- microbenchmarks and eyeballing the generated interpreter stubs
http://cr.openjdk.java.net/~shade/8078438/perf.txt
(Having problems running vm.quick against vanilla jdk9-dev -- it's been
a while I ran it, so it should be an environmental problem -- would
appreciate if anyone is able to run the change with -Xint
-XX:(+-)UseCondCardMark on their setups)
Thanks,
-Aleksey.
More information about the hotspot-runtime-dev
mailing list