[8u] Request for enhancement backport approval for 8144019
Hiroshi H Horii
HORII at jp.ibm.com
Fri Apr 14 16:35:48 UTC 2017
I'm sorry that my request doesn't follow the guideline to backport
> Please publish the actual webrev as it applies to 8
I would like to add links to webrev that Michi created.
> and follow the enhancement request approval template at:
[a rationale for why the enhancement should be backported]
C1 and Tiered Compilation provide faster startup for JVMs. In recent
popular workloads, such as Hadoop and Spark (via Yarn), multiple JVMs
concurrently run with their short life cycles. C1 and Tiered Compilation
significantly shorten execution time of these workloads. For example, the
above changes provide 15% improvement of execution time of Spark (2.1.0)
q16 on POWER8 3.325MHz with 1.0 scale factor.
We have ran SPECjbb2015, Spark TPC-DS queries, and jitreg tests. We
confirmed SPECjbb2015 and Spark worked well. Michi will report the details
of jitreg tests.
Most of changes in shared codes have effects on only ppc64 and ppc64le.
src/share/vm/c1/c1_Compilation.hpp: 1 line changed: 0 ins; 0 del; 1 mod;
src/share/vm/c1/c1_GraphBuilder.cpp: 10 lines changed: 9 ins; 0 del; 1
mod; 4460 unchg
src/share/vm/c1/c1_IR.cpp: 2 lines changed: 1 ins; 0 del; 1 mod; 1404
src/share/vm/c1/c1_IR.hpp: 7 lines changed: 6 ins; 0 del; 1 mod; 357 unchg
src/share/vm/c1/c1_LIR.hpp: 2 lines changed: 0 ins; 0 del; 2 mod; 2506
src/share/vm/c1/c1_LIRGenerator.cpp: 10 lines changed: 8 ins; 0 del; 2
mod; 3673 unchg
src/share/vm/c1/c1_LIRGenerator.hpp: 6 lines changed: 5 ins; 0 del; 1 mod;
src/share/vm/c1/c1_RangeCheckElimination.hpp: 7 lines changed: 6 ins; 0
del; 1 mod; 243 unchg
src/share/vm/c1/c1_Runtime1.cpp: 8 lines changed: 0 ins; 7 del; 1 mod;
src/share/vm/runtime/deoptimization.cpp: 10 lines changed: 2 ins; 0 del; 8
mod; 2074 unchg
src/share/vm/runtime/deoptimization.hpp: 5 lines changed: 1 ins; 0 del; 4
mod; 385 unchg
src/share/vm/runtime/globals.hpp: 3 lines changed: 3 ins; 0 del; 0 mod;
Changes in src/cpu/ppc will add C1 for ppc64 and ppc64le. We believe that
impacts on Interpreter and C2 are limited on ppc64 and ppc64le.
Please tell me if I'm still missing information to request an approval of
Rob McKenna <rob.mckenna at oracle.com> wrote on 2017/04/15 00:13:57:
> From: Rob McKenna <rob.mckenna at oracle.com>
> To: Hiroshi H Horii/Japan/IBM at IBMJP, volker.simonis at gmail.com
> Cc: Tim Ellison <Tim_Ellison at uk.ibm.com>, Michihiro Horie/Japan/
> IBM at IBMJP, jdk8u-dev at openjdk.java.net, "Doerr, Martin"
<martin.doerr at sap.com>
> Date: 2017/04/15 00:15
> Subject: Re: [8u] Request for enhancement backport approval for 8144019
> As I understand it the "webrev" here does not constitute the fix as it
> will be pushed to 8.
> Please publish the actual webrev as it applies to 8 and follow the
> enhancement request approval template at:
> Given that this change actually does affect shared code we will need a
> particular focus on the 2nd bullet point in that template.
> To be explicit: this change is not approved for 8u-dev yet.
> On 06/04/17 05:31, Rob McKenna wrote:
> > Approved on the understanding that Volker is happy with it. Volker,
> > would you mind commenting?
> > -Rob
> > On 07/04/17 01:06, Hiroshi H Horii wrote:
> > > Dear all,
> > >
> > > Could you please approve the backport of the following ppc64 change
> > > jdk8u-dev?
> > >
> > > 8144019: PPC64 C1: Introduce Client Compiler
> > >
> > > Bug: https://bugs.openjdk.java.net/browse/JDK-8144019
> > > Webrev:
> > > Review:
> > > http://mail.openjdk.java.net/pipermail/hotspot-compiler-dev/
> > > URL: http://hg.openjdk.java.net/jdk9/hs/hotspot/rev/4a24de859a87
> > >
> > > Though the original patch includes many changes in shared, we
> will be able
> > > to reduce them to less than 100 lines of changes in shared.
> > >
> > > Regards,
> > > Hiroshi
> > > -----------------------
> > > Hiroshi Horii, Ph.D.
> > > IBM Research - Tokyo
> > >
More information about the jdk8u-dev