Re: [gentoo-mips] Re: What kind of MIPS hardware do we have? (Looking to restart Gentoo/MIPS, need info)

[Matt Turner]

On Sun, Jul 11, 2010 at 7:53 AM, Panagiotis Christopoulos
<pchrist@g.o> wrote:
> On 23:13 Sat 10 Jul     , Matt Turner wrote:
>> To summarize the hardware collection
>>  - 5 MIPS III/Loongson little-endian STMicroelectronics systems (Lemote, Gdium)
>>  - 11 MIPS IV big-endian SGI systems
>>  - 3 MIPS IV little-endian Cobalt systems
>>  - 2 MIPS64 selectable-endian Broadcom systems
>>
>> What do you guys think?
> Hi Matt,
>
>  Sorry for my interference, but I want to write some random thoughts
> that come and go, for while now. It's good that you asked about what
> hardware we have these days. Btw, I have two SGI O2, a r5k and a r10k.
> The r10k is useless for linux (unless something hanpened inside the last
> ~2 years that I served in the army and was away). The r5k is in good
> condition but *extremely slow*. Once upon a time, when the mips stable
> keyword wasn't dropped yet, I wanted to become a mips arch tester, but
> found it extremely difficult to work with my r5k, cause it was very very
> slow as I said before. I tried to find cheap and better mips hardware
> but with no luck. Those days, Lemote just started to exist, and I
> couldn't find any reseller in Europe. The SGI stopped direct support of
> o2/origin hardware(unless I'm wrong), which makes it difficult for
> someone to find spare parts in case something happens to a SGI machine
> (unless you 're in US).
>  From the side of the gentoo linux developer/arch tester/(at least from
> my side), we really need new, good and fast hardware in order to work
> productively. For a big project, such the Gentoo Mips Project, we need
> more people with good and probably also different hardware. But! For
> example, it's very difficult to do any gentoo development(ebuild
> writing/testing, arch testing, compiling stuff, even to build stages with
> catalyst), with embedded hardware or with old SGI hardware.
>  So, my question is, what hardware can a gentoo dev get these days, in
> order to work productively? And another question, a project such as the
> gentoo MIPS project, should see where the future goes and not to live
> in the past, supporting old hardware crap, so where does the MIPS future
> go anyway? Will it go to embedded devices only? Will it also go to
> desktop systems/production servers?
>
> (I really believe that the gentoo MIPS project died the last years cause
> people didn't have the right hardware to work)
>
> --
> Panagiotis Christopoulos ( pchrist )
>    ( Gentoo Lisp Project )

I absolutely agree that the lack of usably fast hardware is the biggest problem.

While installing software on the O2 is slow enough as it is, having
some kind of quality-assurance from Gentoo/MIPS in the form of
ACCEPT_KEYWORDS="mips -~mips" would definitely make installing
software on the O2 less painful. Believe me, I've been there.
Compiling some package on a 300MHz R5000 takes hours, and then you
find out that it's clearly _never_ been tested when it breaks your
entire system so you've got to start the whole process over with a
different version.

I think the future of Linux/MIPS for regular users lies with
Lemote/GDium hardware. (We're probably going to not worry about things
like Octeon, for instance, since they're so expensive.) You'll notice
that many MIPS kernel developer work for companies using MIPS in
embedded environments though.

SGI systems are cheap and available though, so I think we should make
an effort to support them. The SGI Origin 300, with quad 600 MHz
R14ks, 4 MB L2 cache each, and 4 GB RAM is quite a fast system.
Unfortunately, the port has never been finished. O300s are quite cheap
(< 300 USD) and would make excellent Linux/MIPS systems, if only the
kernel port were completed.

I hope I've answered your questions. Please let me know.

Thanks,
Matt