> Easy.. get the count register before and after the loop and check how
> many clock ticks it took, divide by number of loops...
Well, I have to correct myself. I didn't think of the fact that the BogoMIPS
printk() accounts for the fact of the loop having two instructions. So my
PICA's results and the SNI's results are correct but the ones that I on
the Tyne are strange. Oh well, I did only the first steps with MIPS CPUs
and the Linux port on that thing ...
> btw, mips shared lib stuff now runs really good on OpenBSD. the X11 bin
> dir shrinked to about 20% of what it was before. :-)
Who wonders. I know that the stuff will shrink without end. But now that
I have a 32m MIPS box the static binaries don't hurt that extremly anymore.
I could have done the shared library stuff quite some time ago but I want
to avoid working with a binutils snapshot. You know how these guys from
the FSF and Cygnus think about publishing snapshots. You now how all these
people'd like to get these binaries ...
I don't track the BSD flavourism very closely, so why did you switch to
> now, if i had a big endian machine besides the pica... :-)
I have one now; just the documentation for the big endian firmware from SNI
is missing. Well, all documentation is missing for this machine; what I
did is just based on reverse engineering and looking at the output of the
configuration tools of NT ... Finally all the code that I wrote for the
Tyne and rPC helped me alot so that doing the job in just two days was
It's just anoying that switching the SNI box between byteorders takes five
minutes and is a dangerous procedure. A power failure and the firmware
EEPROM has to be replaced physically. Time to brew something faster -
but this is more for the hardware guys ...
Anyway, the fun thing with the SNI box is that the SNI guys probably
believe that I didn't unpack the machine yet or so :->