Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 10 Jul 2019 04:49:57 +0200
From:      Marcin Wojtas <mw@semihalf.com>
To:        Greg V <greg@unrelenting.technology>
Cc:        freebsd-arm <freebsd-arm@freebsd.org>, Denis Polygalov <dpolyg@gmail.com>,  Robert Crowston <crowston@protonmail.com>
Subject:   Re: raspberry pi 4
Message-ID:  <CAPv3WKcUuVg4Naj6AEF24eY6932diBDmwr2CDvXGOO5Q=PYynQ@mail.gmail.com>
In-Reply-To: <5935C316-22B8-49F5-BDFC-F0152C3F4DA9@unrelenting.technology>
References:  <20190709161243.GC4904@mon.zyxst.net> <HZPxf8oyosxDF2kVxJHXYBDY9ULZtF5VHU8FnEslTS9JS-dMsA1G61OnXEHmL0xUVPqZTeF2Q_Z9F58Su81uDDiX86do5d3mqFG7q4teJlw=@protonmail.com> <CAHxjC0-VJmQK=feqAb2H9sSAwHXo8=KTYr3Os72WBB58SaoiMg@mail.gmail.com> <5935C316-22B8-49F5-BDFC-F0152C3F4DA9@unrelenting.technology>

next in thread | previous in thread | raw e-mail | index | archive | help
=C5=9Br., 10 lip 2019 o 03:54 Greg V <greg@unrelenting.technology> napisa=
=C5=82(a):
>
> On July 10, 2019 3:52:43 AM GMT+03:00, Denis Polygalov <dpolyg@gmail.com>=
 wrote:
> >ROCK64 was released a year ago and it is better than RPi 4 now.
>
> Definitely not better in terms of processing power. RPi 4 has quad Cortex=
-A72 cores! A72!!
>
> Also, RPi 4 no longer uses a custom broadcom interrupt controller, it has=
 a GICv2, so it's definitely less stupid than the older ones.
>
> And it's probably the only SoC with an *official* open source GPU driver.=
 Not that the reverse engineered ones are bad, but this is pretty cool. App=
arently Mesa V3D is the *only* driver, and no one would ever get a blob.
>
> >ROCKPro64 even more better and have PCIExpress slot - unique
> >and most desirable feature on ARM boards.
>
> Unfortunately, it has a craptastic Synopsys Designware PCIe host controll=
er. And from what I understand, the worst version of it =E2=80=93 no ECAM m=
ode (probably?), tiny BAR size, no chance of running a GPU. Well, that's wh=
at people say on the internet at least.
>
> (For some context, the Marvell MACCHIATObin (Armada 8040) also has a Desi=
gnware controller, but it actually supports ECAM. It does have a hilarious =
bug =E2=80=93 some devices (in my experience, only ones recognized as "lega=
cy" by edk2) are multiplied, i.e. one device appears in all slots, because =
the hardware does not filter packets properly. Still, it works, and I have =
a GPU running on FreeBSD!)

3 cents on Armada8040 PCIE IP. Only endpoints that appear on bdf 0:0:0
are multiplied on all bus0 positions, e.g. e1000 card - this is why
bus0 requires special treatment (in both modes).  Other than that,
when initially configured to ECAM mode it behaves in a fully generic
way.

Regards,
Marcin

>
> >I just bought one of them while ago and tried to
> >boot FreeBSD on it without success
>
> Well, these days it should work =E2=80=93 CPU clock control, sdcard and E=
thernet at least. I have a patch on phabricator that enables USB3 on the RO=
CKPro64 (RK3399) =E2=80=93 might help ROCK64 too, haven't tested though.
> _______________________________________________
> freebsd-arm@freebsd.org mailing list
> https://lists.freebsd.org/mailman/listinfo/freebsd-arm
> To unsubscribe, send any mail to "freebsd-arm-unsubscribe@freebsd.org"



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAPv3WKcUuVg4Naj6AEF24eY6932diBDmwr2CDvXGOO5Q=PYynQ>