• msolace@alien.topB
    link
    fedilink
    English
    arrow-up
    1
    ·
    1 year ago

    which just shows the scheduler is wrong. which people who cared to put the effort in already did manually with lasso. the only missing piece is random main kernel threads jumping on to p cores. AMD scheduler isn’t perfect either. And both companies are going big/little. so plenty of room to keep improving.

    • CascadiaKaz@alien.topB
      link
      fedilink
      English
      arrow-up
      1
      ·
      1 year ago

      correction: it shows that Intel Thread Director is wrong, and that the scheduler shouldn’t trust it.

      • SkillYourself@alien.topB
        link
        fedilink
        English
        arrow-up
        1
        ·
        1 year ago

        Thread Director doesn’t do any directing, it’s a a set of new registers the OS scheduler is supposed to read for feedback on how well a thread is running on a core. If APO can do it right, it means the scheduler is wrong.

        15.6 HARDWARE FEEDBACK INTERFACE AND INTEL® THREAD DIRECTOR

        Intel processors that enumerate CPUID.06H.0H:EAX.HW_FEEDBACK[bit 19] as 1 support Hardware Feedback Interface (HFI). Hardware provides guidance to the Operating System (OS) scheduler to perform optimal workload scheduling through a hardware feedback interface structure in memory.

        • CascadiaKaz@alien.topB
          link
          fedilink
          English
          arrow-up
          1
          ·
          1 year ago

          facepalm are you daft?

          how the scheduler gets information from ITD doesn’t change what ITD does.