http://hg.openjdk.java.net/aarch32-port/jdk9-arm3264 merged

[Bob Vandette]

> On Oct 12, 2016, at 12:34 PM, Andrew Haley <aph at redhat.com> wrote:
> 
> On 12/10/16 16:48, Bob Vandette wrote:
> 
>> I’d like to have a discussion soon on the future direction of the
>> JDK 9 version of the aarch32 port relative to Oracle’s contribution.
>> 
>> Can we shoot for a phone call or email exchange discussion on the
>> first week of November?  This should give everyone over 2 weeks to
>> take a look at our contribution.
>> 
>> I realize that JDK 9 is still under development but if we’d like
>> integrate an ARM implementation into JDK 9.1, we should start the
>> process pretty soon.
> 
> OK, we should talk about that.
> 
> Re the AArch64 port: I've gone quiet because I'm checking it out, but
> it's tricky and it's taking far longer than I'd expected.
> 
> It turns out to be extremely hard to get benchmarks running on JDK9.
> SPECjvm2008 doesn't run unless it's altered, and the reporting
> restrictions mean that any results can't be published.  Only a subset
> of the DaCapo benchmarks run.  DaCapo doesn't have the nice harness
> that SPECjvm has, so getting reasonably stable measurements is
> difficult.  At least SciMark runs reasonably stably.

Yes, benchmarking is an ART.  There are so many variables and you need to 
select a benchmark that best matches the type of application you are interested
in.  For AArch64, I don’t see these small microbenchmarks like specJVM2008
all that meaningful but running hadoop and specJBB takes considerable effort
to setup and run to get reliable results.  We also have the reporting issue
that gets in the way of comparing unreleased products.

Bob.

> 
> Andrew.