Re: Orange Pi 6 Plus works (15.1-RELEASE)

Sulev-Madis Silber <[email protected]> Tue, 30 Jun 2026 04:08:00 +0300
Newsgroups gmane.os.freebsd.devel.arm
Message-ID <[email protected]>

On June 30, 2026 3:56:50 AM GMT+03:00, Sulev-Madis Silber <freebsd-arm-fre=
ebsd-org097@ketas=2Esi=2Epri=2Eee> wrote:
>
>
>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-f=
reebsd-org097@ketas=2Esi=2Epri=2Eee> wrote:
>>>=20
>>>> 0-1,6-11 are fast, and 2-5 are slow=2E However, it seems like the sch=
eduler is just randomly picking a core, so performance varies wildly depend=
ing 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 un=
familiar 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 hav=
en't even heard it being planned, so booting current is not likely to help =
either
>>>=20
>>> the more hw experienced guys(/gals?) can add additional info here or c=
orrect me or add any roadmaps
>>
>>Ahhh you're right, sorry everyone=2E I thought I read that ULE gained su=
pport 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 know=
ing 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 posit=
ives on pcix
>
>also found <https://wiki=2Efreebsd=2Eorg/Scheduler/Hybrid> but can't comm=
ent 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 j=
ust 2 first parts?
>
>i, most of time, am the first 2
>
>those are my observations so far about assymetric multiprocessing on free=
bsd
>

<https://forums=2Efreebsd=2Eorg/threads/kernel-support-for-heterogeneous-c=
ores=2E98526/>

possibly everyone found this all already

perhaps this could be tied with laptop support, as it's most likely mostly=
 used for mobile platforms and less elsewhere

tho with power prices nowadays, even servers like this=2E tho, i think big=
 platforms use symmetric cpus and just scale them for power if usage is low=
=2E also saves on cooling=2E but i question the low time on Servers (TM)