Gabriel Paubert wrote:
If you're going to reimplement the delays using the time base, why not use the KeyLargo timer instead?And tie the bus on repeated I/O trying to read the timer, fighting perhaps with another processor on SMP? No thanks.
The pci/33 bus is much slower than the arbus/maxbus/whatever you want to call it; furthermore, there's a maximum length you can "tie" the bus. This is not really a problem.
It runs at the same frequency (18.432MHz) on all machines, so that'll save a multiply ;) Too bad for the non-Macs, but else you'll have to special-case the 601 (and maybe others) anyway.I don't know how you'll avoid a multiply unless you want to pass delays in strange units. It's better to have the
The compiler can fold it. Segher