-
Bug
-
Resolution: Won't Fix
-
P3
-
None
-
10
| Issue | Fix Version | Assignee | Priority | Status | Resolution | Resolved In Build |
|---|---|---|---|---|---|---|
| JDK-8198916 | 10.0.1 | Alexander Harlap | P3 | Closed | Won't Fix |
With 10-b40, SPECjvm2008-MonteCarlo-G1 is showing 2%-4% regressions across all platforms (Linux-x64, Windows-x64, Solaris-SPARC, Mac-x64). Platform with the clearest indication (ie, largest regression size) is Solaris-SPARC, but regressions appear across all platforms.
We also see the same regressions sizes in 10.0.1-b2.
The regressions reproduce in:
• Triage (ie, re-run on the same machine)
• Re-triage (ie, run on a different machine)
The regressions hold through at least b42.
JVMswap results indicate the cause is in the JVM. Assigning to hs/gc as a starting point ~ MonteCarlo is primarily impacted by compiler and GC, and in this case, we're not seeing the regression in the ParGC config (SPECjvm2008-MonteCarlo-ParGC).
We also see the same regressions sizes in 10.0.1-b2.
The regressions reproduce in:
• Triage (ie, re-run on the same machine)
• Re-triage (ie, run on a different machine)
The regressions hold through at least b42.
JVMswap results indicate the cause is in the JVM. Assigning to hs/gc as a starting point ~ MonteCarlo is primarily impacted by compiler and GC, and in this case, we're not seeing the regression in the ParGC config (SPECjvm2008-MonteCarlo-ParGC).
- backported by
-
JDK-8198916 10-b40 (and 10.0.1-b2) performance regression in SPECjvm2008-MonteCarlo-G1
-
- Closed
-