7 Nov
2007
7 Nov
'07
6:58 p.m.
[...]
951). Of course, I obviously still have a ways to go.
As evidenced by
the completely garbled text, I have yet to get the PERQ's unique (and
extremely complicated) RasterOp hardware emulated to any significant
Good luck, you'll need it :-). The rasterop hardware is conceptually very
simple, and practically very complicated. The timing of just about
everything is critical, as you must know by now, the machine has a
4-state counter (basically counting through the 4 16-bit words that make
up a 'quadword', the memory hardware and video shift registers being 64
bits wide, whereas the path to the CPU is 16 bits wide). It's possible,
although not really documented, to do a read-modify-write cycle on a
quadword _and the data transfers overlap_ -- IIRC there's a 2 clock delay
between them. And that's exactly what the rasterop hardware does.
The Rasterop microcode is intrictate to say the least....
I asusme you've read the schematics and the techref. Have you seen the
Rasterop microcode? And the filksong?
degree. And after that I have a ton of hardware
details to work out.
But this is a major milestone, and I'm so excited that I just had to
share my excitement with someone... I never thought I'd get anywhere
near this far with the emulation, given how complicated the PERQ is.
And if I've seen so far, it's because I've been standing on the
shoulders of giants. Or one giant, at least -- our own Tony Duell who's
been ever-so-patient with my questioning and who probably knows more
about the PERQ hardware than the original designers did...
That much I really doubt... Mind you, even if I say so myself, I think I
understnad the hardware fairly well. And I have written microcode...
-tony