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