Re: Orange Pi 6 Plus works (15.1-RELEASE)
Sulev-Madis Silber <[email protected]> Tue, 30 Jun 2026 03:56:50 +0300
| Newsgroups | gmane.os.freebsd.devel.arm |
|---|---|
| Message-ID | <[email protected]> |
On June 30, 2026 3:27:18 AM GMT+03:00, Kevin Day <kevin@your=2Eorg> wrote: > > >> On Jun 29, 2026, at 5:37=E2=80=AFPM, Sulev-Madis Silber <freebsd-arm-fr= eebsd-org097@ketas=2Esi=2Epri=2Eee> wrote: >>=20 >>> 0-1,6-11 are fast, and 2-5 are slow=2E However, it seems like the sche= duler is just randomly picking a core, so performance varies wildly dependi= ng on which one you hit=2E >>>=20 >>> kern=2Esched=2Etopology_spec: <groups> >>> <group level=3D"1" cache-level=3D"3"> >>> <cpu count=3D"12" mask=3D"fff,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0">0, 1, 2, = 3, 4, 5, 6, 7, 8, 9, 10, 11</cpu> >>> </group> >>> </groups> >>>=20 >>>=20 >>> Is there an easy fix for this? Linux gets it right, but I'm really unf= amiliar with the low level of arm board support to understand how/where it'= s learning the topology=2E >>>=20 >>> -- Kevin >>>=20 >>>=20 >>=20 >> i'm surprised it works at all >>=20 >> iirc, no multiple speed cpu systems have support in fbsd so far, i have= n't even heard it being planned, so booting current is not likely to help e= ither >>=20 >> the more hw experienced guys(/gals?) can add additional info here or co= rrect me or add any roadmaps > >Ahhh you're right, sorry everyone=2E I thought I read that ULE gained sup= port for Alder Lake's E-cores, but I guess that was only proposed not done = yet=2E > >-- Kevin > > i'm not member of every ml but i know efnet/#bsdmips irc channel has knowi= ng people in it, should you have time to dev this possibly same or other people read it here i looked into commit logs and found nothing on cix p1, except false positi= ves on pcix also found <https://wiki=2Efreebsd=2Eorg/Scheduler/Hybrid> but can't comme= nt much about stuff written in it to me it looks more like that this subproject is in some pre-code stage=2E= hopefully it didn't just fizzle out basically we need user, tester, and dev=2E can you be all of those 3 or ju= st 2 first parts? i, most of time, am the first 2 those are my observations so far about assymetric multiprocessing on freeb= sd