Subject: Re: Time travel..
samuel@ibrium.se
Date: Thu Sep 20 2001 - 15:52:19 MDT
On Thu, Sep 20, 2001 at 12:50:37PM -0700, Timothy A. Seufert wrote:
> At 1:48 PM -0500 9/20/01, Matt Christian wrote:
>
> >I know this was addressed for Adam but I have been having a similar
> >problem. I have a Power Computing PowerCenter Pro (originally 210 MHz
> >604e) upgraded to a 390 MHz G3 via an XLR8 accelerator card.
>
> What is the card's bus frequency and PLL multiplier setting? At 390
> MHz in a PowerCenter Pro, it seems likely that it's 60 MHz * 6.5.
> (This is important -- see below.)
>
> >> What does the "Dec frequency" line in the mol-log say?
> >
> >Mine states:
> >DEC frequency: 14 MHz, 140:10 mticks/usec
>
> PowerPCs have a high precision built-in timer called the
> "decrementer", which decrements at 1/4 the bus frequency on most PPCs
> (including the G3 and the 604e). If MOL and the MacOS running inside
> of MOL think the decrementer runs at 14 MHz, and it is actually
> running at 60/4 = 15 MHz, you've definitely got a source of clock
> drift.
>
One thing to try. Add a line
timebase_frequency: 0x017C0F93
to /etc/molrc. Try some different values and see
what happens:
0x00400F93 (4 MHz),
0x00e4e1c0 (15 MHz),
0x01312d00 (20 MHz),
0x017C0F93 (25 MHz),
0x02160ec0 (35 MHz),
0x02faf080 (50 MHz),
0x05f5e100 (100 MHz)
At least in the newworld setting, this parameter is the *only*
one that affects the operation of MacOS (I've tried modifying
the values of suspicious OF-properties without seeing any effects
whatsoever). Even a missing "clock-frequency" property should
be harmless.
A bogus "timebase-frequency" property is serious.
If it is completely missing, mol will calibrate using
gettimeofday (this is a relatively new feature).
Note: The above is true for 0.9.60. VBL interrupts and
all internal timers rely solely on the timebase_frequency
value nowdays. This is not true true for earlier version
of MOL.
/Samuel
----------------------------------------------------------
E-mail <samuel@ibrium.se> WWW: <http://www.ibrium.se>
Phone/fax: (home) +46 8 4418431, (work) +46 8 7908174
----------------------------------------------------------
This archive was generated by hypermail 2a24 : Thu Sep 20 2001 - 15:02:19 MDT