JDK-8269598 : Regressions up to 5% on aarch64 seems due to JDK-8268858
  • Type: Bug
  • Component: hotspot
  • Sub-Component: compiler
  • Affected Version: 18
  • Priority: P1
  • Status: Closed
  • Resolution: Fixed
  • OS: linux
  • CPU: aarch64
  • Submitted: 2021-06-29
  • Updated: 2022-02-08
  • Resolved: 2021-07-15
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 18
18 b07Fixed
Related Reports
Relates :  
JDK-8268858 - Determine register pressure automatically by the number of available registers for allocation
Description
Seen in 18-b3 retriage and by individual CI build experiments.
Comments
[~dlong] Thanks for your information. Previously I already found performance degradation with -XX:+PreserveFramePointer in Alibaba Flink Nexmark benchmark. It's caused by inefficient codes generated by RA. That's why I begin working on register pressure. I already created JDK-8270808 [1] to track it. Meanwhile I will also track your mentioned regression issue together with it. Hope they're caused by the same reason.
16-07-2021
[~jzhu] Please try SPECjvm2008 compress on x86/x64 with -XX:+PreserveFramePointer. We are seeing a regression after JDK-8268858.
15-07-2021
Changeset: 7c234919 Author: Joshua Zhu <jzhu@openjdk.org> Committer: Yi Yang <yyang@openjdk.org> Date: 2021-07-15 01:57:31 +0000 URL: https://git.openjdk.java.net/jdk/commit/7c2349191893f07d133ff94a0be4472948f7920b
15-07-2021
I've found the root cause and will fix it next week.
09-07-2021
I can reproduce this issue. It's caused by extra spilling for c2 version of function "Compressor::compress". I'll try to figure out the root cause.
02-07-2021
Thanks Eric. I will firstly reproduce this regression.
01-07-2021
[~jzhu] We have 2 ARM platforms with slightly different CPUs. We observed similar perf regressions on both with some SPECjvm2008 benchmarks: SPECjvm2008-Compress-G1 : -5.23% SPECjvm2008-Compress-ParGC : -5.56% SPECjvm2008-MPEG-G1 : -1.49% SPECjvm2008-XML.transform-ParGC : -1.91% SPECjvm2008 is a free download from www.spec.org. I hope you can reproduce this result on your hardware.
30-06-2021
ILW = performance regression, one platform = HMH = P1
30-06-2021