Hi Michael, et al,

Micheal Huth wrote:
> [...]
> BTW: Wolfgang I got your REU.

oh yeah and meanwhile you managed to do the die shots,
stitch all the parts together and create a first full
screen version of the REU's controller chip die:

     http://mail.lipsia.de/~enigma/neu/6581.html#mos8726r1

Well, I took the "Low quality JPG image" ;-) and tried
to identify some building blocks upon my knowledge and
experience with analysing the 8726's register 
behaviorhttp://d81.de/shared/8726R1lowmq.jpg
http://d81.de/shared/8726R1lowmq.jpg
from software tests. I don't know, if my identifications
make any sense, in fact it is not much more than an
educated guess.

     http://d81.de/shared/8726R1lowlq_building-blocks.jpg

The abstract layer alone:

     http://d81.de/shared/8726R1low_building-blocks_abstract.png

Your full screen image in lower resolution matching to
the abstract layer above:

     http://d81.de/shared/8726R1lowmq.jpg



I believe that at least one open question from all
my analyses can be answered now. I asked myself, how
Commodore MOS engineer would have implemented the
SWAP function of the REU. Using two 8-bit register
pairs to store the databus direction from C64->REU
and vice versa would have been the straight forward
design.
I was a bit unsure, if they perhaps tried to find a
register optimized variant by using only one 8-bit
register for this operation instead of requiring
two. This could have been achieved by a complex
interleaving strategy, letting the REU doing
internal operations while the C64 is in its VIC bus
access cycle.

But as you can see for the left column of green
boxes denoting somthing like a D-FlipFlop register,
there are two register paired together, one for the
C64->REU direction and one for the REU->C64
direction. 8 of these register pairs build the whole
column.


By the way, the red boxes denote something like a
register combined with a 1-bit counter cell. You can
surely see that one half of that register-counter
cell is identical to the contents of the green box.

Not all red boxes are 100% identical (beside the
mirroring). The cells that build the register pair
0xDF07/0xDF08 contain less "parts" for the counter
cell as all the other register-counters.


Womo

       Message was sent through the cbm-hackers mailing list