JDK-8222675 : [Graal] 19% regression in promo benchmark in b15
  • Type: Bug
  • Component: hotspot
  • Sub-Component: compiler
  • Affected Version: 13,14
  • Priority: P4
  • Status: Resolved
  • Resolution: Cannot Reproduce
  • CPU: x86_64
  • Submitted: 2019-04-17
  • Updated: 2020-02-06
  • Resolved: 2020-02-06
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 15
15Resolved
Related Reports
Relates :  
JDK-8221341 - Update Graal
Description
We suspect this is due to 
https://github.com/oracle/graal/issues/942

but I will track it down to the OpenJDK commit and update this bug.
Comments
I re ran this over 13-b10 to 14-b5 and its all similar now.
06-02-2020
I am going to re-run the trend line on this on OCI in a new report from 13-b10 into some 14 builds to see what is going on if anything.
05-02-2020
Doug, could you please have a look or re-assign? Thanks.
29-10-2019
OK I will assign it to myself for now to remind me.
06-06-2019
This should be pulled in by the next Graal Update (JDK-8223807). [~ecaspole], could you please verify that it solves the issue?
06-06-2019
This should be fixed by https://github.com/oracle/graal/commit/07a13a12e45b0567e074f13dec0793ca1e5506c1. When matching a memory operation into a multiply we were using the non-avx variant of the instruction which made the register allocation a bit worse, and also ended up mixing SSE and AVX.
15-05-2019
Command line: -server -XX:+UseG1GC -XX:+UseLargePages -XX:+UseCountedLoopSafepoints -XX:+IgnoreUnrecognizedVMOptions -XX:LoopStripMiningIter=1000 -XX:+UnlockExperimentalVMOptions -XX:+EnableJVMCI -XX:+UseJVMCICompiler
15-05-2019
Eric, I'm assigning this to you for now. Please un-assign or re-assign to Labs once it's clear which change introduced this. Thanks.
18-04-2019
ILW = Performance regression (introduced in JDK 13 b15), benchmark with Graal as JIT, no workaround = MLH = P4
18-04-2019