|
Authored by: Anonymous on Sunday, March 31 2013 @ 04:06 PM EDT |
I'd look in the early mathematics and computer science literature for papers on
floating point and it's hickups. I know rounding issues were discussed, and
there may even be a paper or letter in a journal debunking rounding before
operations. I was taught these issues back in the mid '80s. It was old back
then.[ Reply to This | Parent | # ]
|
|
Authored by: jesse on Sunday, March 31 2013 @ 07:17 PM EDT |
Took up an entire 4K segment.
You trapped to it using FENT (which I believe was an IO instruction that caused
a trap), which switched memory pages. Then it interpreted a variation on the
PDP-8 instruction set with floating point, and returned to the original segment
when it interpreted a FEXIT. As I recall it implemented a register machine for
the interpretations.
The PDP-11 version was used on PDP-11/05, originally designed for controllers,
but was heavily used for navigation systems too. I used it starting 1978, but it
was generally available much earlier. This emulator used a software stack, and
threaded code for interpretation.
For those that don't know, threaded code on a PDP-11 was started by a subroutine
call - the return address was stored in a register (R5 I belive), from then on,
the interpreter just did a JMP @(R5)+, which used the next 16 bit word as the
address of the next function to call, and incremented the pseudo pc (register
5). So the last instruction of every function was JMP @(R5)+, making the
interpretation overhead only one instruction. It also allowed you to extend the
interpreter trivially.[ Reply to This | Parent | # ]
|
|
|
|
|