Uploaded image for project: 'JDK'
  1. JDK
  2. JDK-8167319

Peformance regression by +OptoScheduling option in SPARC

    XMLWordPrintable

    Details

    • Type: Enhancement
    • Status: Closed
    • Priority: P4
    • Resolution: Won't Fix
    • Affects Version/s: 9, 10
    • Fix Version/s: tbd
    • Component/s: hotspot
    • Labels:
    • Subcomponent:
    • CPU:
      sparc
    • OS:
      generic

      Description

      'testArrayStoreCheckRT1' and 'testArrayStoreCheckRT2' in JMH benchmarks have performance regression in SPARC when they work with enabled OptoScheduling option. In SPARC, running those benchmarks with disabled OptoScheduling give around 4x throughput performance compared to running with enabled OptoScheduling option. Generated code with enabled OptoScheduling shows that 8 store word instructions and 8 store byte instructions are interleaved in code sequence while
      disabled OptoScheduling gives non-interleaved store instructions code sequence which groups store word instructions store byte instructions, separately.

        Attachments

          Activity

            People

            Assignee:
            skim Sung Jin Kim
            Reporter:
            skim Sung Jin Kim
            Votes:
            0 Vote for this issue
            Watchers:
            6 Start watching this issue

              Dates

              Created:
              Updated:
              Resolved: