Stephen M. Rumble wrote:
> Quoting Garrett D'Amore <garrett_damore@tadpole.com>:
>
>>> The real solution, I imagine, is to port i386's i8254 timecounter
>>> support, as that is what IP12 uses.
>>
>> Well, if there is an i8254 in there, then yes, writing a tc framework
>> for it would be very, very good.  I don't have hardware to test
>> against.  Do you?  (I might be willing to do the work if you will test.)
>
> None of my IP12 machines work anymore and they're all in another
> country at the moment anyhow. pooka@ or sekiya@ may be able to test,
> but we three are the only ones I know to have run the code at all. If
> you want to commit something that's just compile-tested, I won't
> disapprove. IP20 _might_ have an i8254 as well, but we have no use for
> it. If it does, someone may be able to test against it instead.
>
> I'm not worried about IP12 having fantastic timekeeping, though. It's
> probably not worth your effort if it just runs as it used to. If you
> go about it anyhow, perhaps it would be important to keep in mind that
> SGI wired a 1MHZ master clock to TIMER2, which ticks the other two
> timers accordingly. At least, that's my recollection of a few years
> ago. There may be comments in the code, but I'm unsure.
>
>> I can make test kernels for GENERIC_IP32, GENERIC_IP2x, and GENERIC_IP12
>> available, if people want them to test with.
>
> People on the list have been very helpful in trying out GENERIC_IP2x
> kernels. If you were to make one avaiable and put out a request,
> someone may oblige. I'm sure that if there were any introduced
> problems with the MIPS3 machines somebody will catch them before the
> next branch.
>
> Thanks again for taking care of this.
>
> Steve

Really only the MIPS3 machines are a concern.  They are in HEAD now. 
(The only reasons they are a concern is that they use the mips3_cp0
counter.)  The older MIPS1 systems don't get excellent time keeping, but
they never had it either.  If no one is running them much anymore, then
it probably isn't worth my effort to fix that. :-)  (The timekeeping
isn't "inaccurate", its just that its accurate resolution is limited to
the granularity of the system clock -- e.g. 100Hz or whatever is
configured for the system.  After that it just increments once for each
time microtime() or whatever is called.)

-- 
Garrett D'Amore, Principal Software Engineer
Tadpole Computer / Computing Technologies Division,
General Dynamics C4 Systems
http://www.tadpolecomputer.com/
Phone: 951 325-2134  Fax: 951 325-2191