<The PDP11 (at least every PDP11 that I've used) addresses memory by
<bytes. OK, a 16 bit word, like an instruction, does have to lie on at an
<even address, but the 16 bit address that a program generates (before
<translation by the MMU) addresses one of 64K _bytes_
True but, if the memory contained nothig but instructions and addresses
32KW is the limit. Only data or IO is addressed as bytes. This would
seem a trivial item it makes a difference in terms of the total number of
instructions in any given amount of ram. This is more important when
applied to machines with I&D space where the byte addressability can be
exploited for buffers and the like.
<The I/O page (certainly on Unibus PDP11's) is 4 K _words_ long. Without
<an MMU you have 28kW memory, 4kW I/O
Some versions and OS vaiations allowed a 2KW IO space. I've run RT-11
that way.
<Wit hthe MMU enabled you can have 64k Bytes (32 k Words) of memory mapped
<in. In fact you generally don't map the I/O page into user processes,
<since you don't want user processes accessing device registers.
True but then people ahve to understand that the MMU design seperates user
space from system space.
<Even if you have the MMU it's a lot simpler if your program fits into 32
<kW (or 28kW if you want access to the I/O space). Otherwise you have to
<do something like using a software interrupt to change the MMU to a
<different mode.
I tried to keep it simple.
<Unless page K1-11 of my PDP11/44 printset is wrong, the PDP11/44 has 22 b
<addressing. The PDP11/45 does, indeed, only have 18 bit, alas...
With the exception of the 11/70 I thought most of the unibus machines were
limited to 18bits, for some devices that can be a limit. My interaction
with U-bus machines is limited to the 11/70 and a 11/34 all my PDP11 time
is with Q-bus (11/03, falcon, 11/23, 11/23A, 11/23b, 11/73) which has
sufficient enough variation to track.
Allison
Received on Sat Apr 11 1998 - 17:26:09 BST