Re: MISC-d Digest V99 #91

[theFox@xxxxxxxxxxxxxxxxxxx (Jeff Fox)]

Dear MISC readers:

> I have some information
>about F21d chips and barely nothing about P32.

The two P32 that were done a few years ago were proprietary chips and
most of the information about them was not released to the public.  Most
companies use non-disclosure agreements to protect their property. As
the years have passed I published some of the information on some of
these chips that belonged to other people.  We have published that
there were six five bit instructions and a 32 bit data bus but little
about the details of the coprocessors etc.  Of the fifty or so chips
that Chuck has done only a couple have been exposed to the public.

>P32 will be much more adequate, but I have no enought info about it. what about
>stage of development, specifications chart, expected price, etc.?

The two P32 done so far were not for sale, they were intended for 
specific products.

>From: rodger clampitt <rclampitt@QNET.COM>
>
>	If I am not mistaken the P32 never became silicon. It existed only
>	on paper or as simulation.

Two P32 were laid out in 95 in OKAD.  One was fabed.  The variant for
the PDA that was fabed used BGA packaging.  As as result it was smaller
than the 20 bit chips which have used a larger rectangle of pads for
chip mounting.  That was almost five years ago but it was silicon. 

At the time Chuck was just using P as a prefix for most chips as in
MuP21.  (actually M myu P 21)  The two P32 were for different companies
and could have had some other name.  There are other 32 bit variations
in the works.  

One of the most interesting characteristics of Chuck's layout is the
vertical and horizontal rows and columns.  Registers and instructions
are in rows and the bits for each are in columns.  Chuck has said
that it is about a days work to copy and paste 12 columns to make
a 20 bit chip into a 32 bit chip. There there is about a months
work in changing the decoding and patching the logic gates that
are not just a matter of adding columns.

The main issue has always been development costs.  You cannot
predict how many fab runs will be needed to get completely
working silicon but the fab costs are proportial to die size
and pin count.  The memory bus and pinout of P21 was designed
to squeaze things into a cheap development package.  

Although in volume the difference between a 20 bit and a 32 bit
variation is $1 vs $2 the memory bus, memory chips etc. all are
twice as big.  Not a signifigant issue except for embedded apps
where every cent might count.  What is more signifigant is the
few hundred K difference in development costs.

>	The only 32bit forth like CPU I know of is the PSC1000.

There is also the SC32 available from Silicon Composers.

Jeff Fox