Bug ID: JDK-8186027 C2: loop strip mining

JDK-8186027 : C2: loop strip mining

Type: Enhancement
Component: hotspot
Sub-Component: compiler
Affected Version: 10

Priority: P4
Status: Resolved
Resolution: Fixed

Submitted: 2017-08-09
Updated: 2024-08-27
Resolved: 2017-11-28

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 10
10 b36Fixed

Related Reports

Duplicate :	JDK-8154302 - C2 removes safepoint poll from loop
Duplicate :	JDK-8177704 - Regression in SPECjvm2008.sparse-large because of changes to marking cycle in JDK-8017744
Relates :	JDK-8192762 - LoopNode::verify_strip_mined() fails with "assert failed: only phis"
Relates :	JDK-8193597 - sun/nio/cs/TestStringCoding.java fails intermittently with getBytes(csn) failed -> GBK
Relates :	JDK-8196296 - Lucene test crashes C2 compilation
Relates :	JDK-8197563 - assert(is_Loop()) crash in PhaseIdealLoop::try_move_store_before_loop()
Relates :	JDK-8202747 - C2: assert(mode == ControlAroundStripMined && use == sfpt) failed: missed a node
Relates :	JDK-8202950 - C2: assert(found_sfpt) failed: no node in loop that's not input to safepoint
Relates :	JDK-8208275 - C2 crash in Node::add_req(Node*)
Relates :	JDK-8229450 - C2 compilation fails with assert(found_sfpt) failed: no node in loop that's not input to safepoint
Relates :	JDK-8231550 - C2: ShouldNotReachHere() in verify_strip_mined_scheduling
Relates :	JDK-8234350 - assert(mode == ControlAroundStripMined && (use == sfpt \|\| !use->is_reachable_from_root())) failed: missed a node
Relates :	JDK-8235452 - Strip mined loop verification fails with assert(is_OuterStripMinedLoop()) failed: invalid node class
Relates :	JDK-8253756 - C2 CompilerThread0 crash in Node::add_req(Node*)
Relates :	JDK-8192977 - LoopStripMiningIter is set to 1000 if UseCountedLoopSafepoints is enabled
Relates :	JDK-8195117 - Performance regressions from JDK-8186027
Relates :	JDK-8247838 - PostLoopMultiversioning is broken and generates incorrect result
Relates :	JDK-8322631 - Enable loop strip mining for all GCs
Relates :	JDK-8247307 - C2: Loop array fill stub routines are not called
Relates :	JDK-8194993 - Loop Strip Mining has some leftover debugging code
Relates :	JDK-8196294 - LoopStripMiningIterShortLoop is set to zero by default
Relates :	JDK-8236759 - ShouldNotReachHere in PhaseIdealLoop::verify_strip_mined_scheduling
Relates :	JDK-8194914 - Compilation fails with "node not on backedge" in OuterStripMinedLoopNode::adjust_strip_mined_loop
Relates :	JDK-8282045 - When loop strip mining fails, safepoints are removed from loop anyway
Relates :	JDK-8229994 - assert(false) failed: Bad graph detected in get_early_ctrl_for_expensive
Relates :	JDK-8257575 - C2: "failed: only phis" assert failure in loop strip mining verification
Relates :	JDK-8270886 - Crash in PhaseIdealLoop::verify_strip_mined_scheduling
Relates :	JDK-5014723 - implement "strip mining" loop optimization
Relates :	JDK-8241486 - G1/Z give warning when using LoopStripMiningIter and turn off LoopStripMiningIter (0)
Relates :	JDK-8241492 - Strip mining not working for test/hotspot/jtreg/compiler/c2/Test6850611.java
Relates :	JDK-8293996 - C2: fix and simplify IdealLoopTree::do_remove_empty_loop
Relates :	JDK-8194992 - Null pointer dereference in MultiNode::proj_out related to loopexit()
Relates :	JDK-8197931 - Null pointer dereference in Unique_Node_List::push of node.hpp:1510
Relates :	JDK-8199784 - PhaseIdealLoop::place_near_use() might return wrong control with loop strip mining
Relates :	JDK-8205107 - assert(c->Opcode() == Op_SafePoint) failed: broken outer loop
Relates :	JDK-8220374 - C2: LoopStripMining doesn't strip as expected
Relates :	JDK-8195118 - LoopStripMining causes performance regression

Description

We've implemented loop strip mining as a way to balance pause time and throughput in the Shenandoah repo. We've been using it for several months and it is stable as far as we can tell. We intend to propose it for jdk 10.

Comments

[~dnsimon] no plan to take care of it at this point. Maybe later.
26-03-2020
[~roland] do you have plans to port/add strip mining to Graal? I imagine that will be important for efficiently running Shenandoah with Graal.
25-03-2020
URL: http://hg.openjdk.java.net/jdk/jdk/rev/f913f6dba2d3 User: jwilhelm Date: 2017-12-07 09:00:22 +0000
07-12-2017
URL: http://hg.openjdk.java.net/jdk/hs/rev/f913f6dba2d3 User: kvn Date: 2017-11-28 18:02:13 +0000
28-11-2017
new webrev: http://cr.openjdk.java.net/~roland/8186027/webrev.02/
21-11-2017
SPECjvm2008-Sparse.small-G1 show significant regression - about 10%. [~roland] Roland, can you verify? Thanks, Vladimir
21-11-2017
Latest webrev: http://cr.openjdk.java.net/~roland/8186027/webrev.01/
27-10-2017