JDK-8231720 : Some perf regressions after 8225653
  • Type: Bug
  • Component: hotspot
  • Sub-Component: compiler
  • Affected Version: 11,14
  • Priority: P3
  • Status: Resolved
  • Resolution: Fixed
  • Submitted: 2019-10-01
  • Updated: 2022-02-06
  • Resolved: 2019-10-10
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 11 JDK 13 JDK 14
11.0.10-oracleFixed 13.0.4Fixed 14 b19Fixed
Related Reports
Relates :  
JDK-8230552 - Provide information when hitting a HaltNode for architectures other than x86
Relates :  
JDK-8225653 - Provide more information when hitting SIGILL from HaltNode
Description
At least 5 promo benchmarks got regressions in 14-b14
Comments
Fix request (13u): I would like to backport this fix to 13u as a followup for JDK-8225653. The original change applies cleanly.
08-06-2020
Fix Request (11u). Patch fixes a performance issue with JDK-8225653 (speeds up stop() unless ShowMessageBoxOnError is on), applies cleanly. This backport is the second in a series of three that starts with JDK-8225653 and ends with JDK-8022574.
23-05-2020
URL: https://hg.openjdk.java.net/jdk/jdk/rev/1b76d17440a0 User: thartmann Date: 2019-10-10 08:39:09 +0000
10-10-2019
I ran a small benchmark locally and it looks like Tobias' guess is right: With __ ud2() instead of __ stop() in x86.ad: Benchmark Mode Cnt Score Error Units Transform.transform thrpt 15 319.670 ± 6.630 ops/min With __ stop() in x86.ad: Benchmark Mode Cnt Score Error Units Transform.transform thrpt 15 305.140 ± 11.541 ops/min I will have a closer look at that.
02-10-2019
ILW = Small performance regression, several benchmarks, no workaround = MMH = P3 This might be a side effect of more debug assembly being emitted: https://hg.openjdk.java.net/jdk/jdk/rev/fafba5cf3546#l1.8 Although that code is never executed, it might affect instruction cache behavior. But that's just a guess.
02-10-2019