AArch64: Status

Christian Thalinger christian.thalinger at oracle.com
Thu Mar 10 04:08:27 UTC 2016

> On Mar 9, 2016, at 2:19 AM, Roland Schatz <roland.schatz at oracle.com> wrote:
> On 02/27/2016 12:03 AM, Christian Thalinger wrote:
>> Now that https://bugs.openjdk.java.net/browse/JDK-8150561 is integrated into hs-comp the Graal AArch64 port can run (almost) all bytecode unit tests successfully.  I say almost because there are two that are failing:
>> Tests run: 1374,  Failures: 2
> Nice!
>> These two fail because of the way AArch64Address works today.  In my opinion AArch64Address knows too much about low-level details of AArch64 addressing modes.  Fixing these two failures is non-trivial so I’m postponing this.
>> The next most important thing to fix is deoptimization.  Right now we crash every time we hit an uncommon trap or deoptimize for some other reason.  Here is a stack trace for a deoptimization:
> IMO, now would be a good time to implement the test/compiler/jvmci/code tests for AArch64. They should catch most problems with code installation and debug info, which might be a reason for deopt failures.

Right.  Any volunteers? :-)

More information about the graal-dev mailing list