Request for review: 8004076: Move _max_locals and _size_of_parameters to ConstMethod for better sharing
Coleen Phillimore
coleen.phillimore at oracle.com
Wed Dec 5 14:01:20 PST 2012
Hi Jiangli,
This looks great! It's also bug 6329051. I had two questions. Can
you run -Xshare:dump on 64 bits to see if the SharedReadOnly size is
still big enough? JPRT will run -Xshare:dump for you on 32 bit
platforms. The other is there is strange indentation in vmStructs
around the lines you changed in the webrev, which may or may not be an
unintended change.
Thanks!
Coleen
On 12/5/2012 3:38 PM, Jiangli Zhou wrote:
> Hi Coleen and others,
>
> Please review the webrev for moving the Method::_max_locals and
> Method::_size_of_parameters to ConstMethod:
>
> http://cr.openjdk.java.net/~jiangli/8004076/webrev.00/
>
> Tested with jprt, runthese and vm.quick.testlist. No noticeable
> performance impact with the change.
>
> ==============================================================================
>
> logs.maxlocal_baseline.spec.1:
> Benchmark Samples Mean Stdev Geomean Weight
> specjvm98 8 546.91 12.58
> ==============================================================================
>
> logs.maxlocal.spec.1:
> Benchmark Samples Mean Stdev %Diff P
> Significant
> specjvm98 8 547.44 12.13 0.10
> 0.933 *
> ==============================================================================
>
>
> ==============================================================================
>
> logs.maxlocal_baseline.jbb:
> Benchmark Samples Mean Stdev Geomean Weight
> specjbb2005 3 56898.90 481.78
> ==============================================================================
>
> logs.maxlocal.jbb:
> Benchmark Samples Mean Stdev %Diff P
> Significant
> specjbb2005 3 56901.46 613.98 0.00
> 0.996 *
> ==============================================================================
>
>
> Thanks,
> Jiangli
More information about the hotspot-runtime-dev
mailing list