JDK-8283058 : Regression ~6% on Renaissance-ScalaKmeans on Windows in b13
  • Type: Bug
  • Component: hotspot
  • Sub-Component: compiler
  • Affected Version: 19
  • Priority: P3
  • Status: Resolved
  • Resolution: Not an Issue
  • OS: windows
  • CPU: x86_64
  • Submitted: 2022-03-11
  • Updated: 2022-04-04
  • Resolved: 2022-04-04
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 19
19Resolved
Related Reports
Relates :  
JDK-8278296 - Generalize long range check transformation
Description
Previously ScalaKmeans has shown bimodal performance but in b13 it is always on the slow side of the bimodal range and stayed consistently like that for several CI builds.
 
These run on Wnidows Server 2016 on OCI bm2.52 on Skylake cores.
Comments
The performance has sometimes returned to the earlier faster scores after b13 and continues bimodal.
04-04-2022
Further work on ScalaKmeans shows that it went bimodal on Windows since 19-b6 with JDK-8276673, when it started to fluctuate up to 23% faster than the earlier JDK-19 scores.
04-04-2022
I have been able to repro this sometimes in 772 a few builds before the 776 which is JDK-8278296 .
14-03-2022
JDK-8278296 optimizes long range of the shape built by the intrinsic for Preconditions.checkIndex(long index, long length, ..). That pattern could surface as the result of some c2 transformation but that seems very unlikely. I added a ShouldNotReachHere() in PhaseIdealLoop::transform_long_range_checks() and ran the benchmark. The ShouldNotReachHere() didn't fire. I doubt JDK-8278296 has something to do with regression.
14-03-2022
[~roland], please take a look.
12-03-2022
ILW = smallish perf regression; one benchmark, 1 platform; no workaround = MMH = P3
12-03-2022
I think this regression happens starting from the build with this change JDK-8278296.
11-03-2022