Bug ID: JDK-8148604 JEP 280, Switch to more optimal concatenation strategy

JDK-8148604 : JEP 280, Switch to more optimal concatenation strategy

Type: Sub-task
Component: core-libs
Sub-Component: java.lang

Priority: P2
Status: Resolved
Resolution: Fixed

Submitted: 2016-01-29
Updated: 2016-06-23
Resolved: 2016-05-19

Versions (Unresolved/Resolved/Fixed)

The Version table provides details related to the release that this issue/RFE will be addressed.

Unresolved : Release in which this issue/RFE will be addressed.
Resolved: Release in which this issue/RFE has been resolved.
Fixed : Release in which this issue/RFE has been fixed. The release containing this fix may be available for download as an Early Access Release or a General Availability Release.

To download the current JDK release, click here.

JDK 9
9 b120Fixed

Related Reports

Blocks :	JDK-8150463 - StringConcat MH_INLINE_SIZED_EXACT should skip storage initialization
Blocks :	JDK-8146801 - Allocating short arrays of non-constant size is slow
Blocks :	JDK-8144590 - JEP-JDK-8085796: Indify String Concatenation - Performance Plan
Blocks :	JDK-8151516 - test/tools/javac/TestIndyStringConcat depends on runtime JDK details
Blocks :	JDK-8150465 - Unsafe methods to produce uninitialized arrays
Blocks :	JDK-8153265 - [TESTBUG] compiler/whitebox/ForceNMethodSweepTest should not assume asserts are benign
Blocks :	JDK-8153277 - [TESTBUG] gc/arguments/TestMaxMinHeapFreeRatioFlags is too sensitive for stray allocations in verifyRatio
Blocks :	JDK-8148605 - jdk.compiler should be compiled with Indy String Concat enabled
Relates :	JDK-8158851 - MH.publicLookup() init circularity, triggered by custom SecurityManager with String concat and -limitmods java.base

Description

We are now flying with a default concat strategy that dubs javac's StringBuilder::append chains without any hints. We need to consider switching to another scheme by default.

Comments

MH_INLINE_SIZED_EXACT wins on all tests: http://cr.openjdk.java.net/~shade/8148604/perf/ Results are in ns/op, and encoded as: (m|s) - multi/single thread, (c1|c2) -- C1 or C2 compiled, (inline|indyWithConstants) -- indy bytecode flavor. See e.g. C1 and C2 performance with different strategies and single-threaded workload: http://cr.openjdk.java.net/~shade/8148604/perf/s-c1-indyWithConstants.log http://cr.openjdk.java.net/~shade/8148604/perf/s-c2-indyWithConstants.log

18-05-2016

JDK-8146801 is known to improve both BC_SB_SIZED_EXACT and MH_INLINE_SIZED_EXACT strategies a lot. Need that patch in mainline before re-doing performance work.

28-02-2016