[pdp7-unix] character table mysteries...

Lars Brinkhoff lars at nocrew.org
Tue Oct 22 18:32:47 AEST 2019


Phil Budne wrote:
> Is it possible all of the display character shapes were in one file?

That seems likely.  Now that we see that the pages "04-cas" has been
rearranged a bit, it's not too far fetched to assume they belong with
the "03-scope" pages.


> How did the character tables get into memory?
>
> [...] Nor have I (yet) spotted any software that seems to write to
> location 3072 (tho it wouldn't have to be very large: an executable
> with the table, and a copy routine).  Now it was core, and if
> undisturbed, it would just "be there" (much as I think DDT was on the
> RLE PDP-1), and it could have been loaded from paper tape using a
> "boot tape"

I have been thinking in similar terms.

Some more data:

Presumably there would be a 96 or 128-word table, likely located first
thing at 3072.  The version of the the character shapes we have results
in exactly 867 words.  So we're looking at a total which is slightly
less than 1K words.

The table words have 10 bits to specify the location of the shape data,
so that's nicely consistent with 1K.  Is the memory split 4K/4K between
kernel and user?  So really, 1K is all there is.

Note that cas outputs characters like this:

   x:
   <octal numbers>
   label=x-b+<size << 10>

So if this is fed to the regular as assembler and b is defined to 3072,
we should get a good output binary for writing to 3072.  I imagine a
file chargen.s like this:

   b=3072

   " table
   .=b
   " maybe 32 empty words here
   space
   excla
   quote
   "...
   rbr
   tild
   null

   " shape data
   <insert cas output here>


More information about the pdp7-unix mailing list