From: minyard@crchh453.bnr.ca (Corey Minyard) Subject: Re: VM386? Possible? Date: 2 Sep 1992 14:05:41 GMT
In article <1992Sep2.083005.6772@ugle.unit.no>, hta@boheme.er.sintef.no (Harald Tveit Alvestrand) writes:
|> Forgive my ignorance....
|> what would happen if one did a preprocessing pass through the EXE files when
|> loading replaced all occurences of POPFD with INT NN (which would trap)?
|> Of course one would not catch self-modifying code and things that swapped in
|> code after starting, but still, it seems easier than emulating a 600-point
|> API....
Well, how would you tell the difference between a section of data that
happenned to have the data that was the same as the POPFD (or PUSHFD, for
that matter), and a real instruction. Also, I don't know if any other
problems exist in the 386 besides that one. Knowing the design of Intel
processors, there probably is. Someone sent me mail saying Intel didn't
know how to design processors; I agree completely. Too bad the world
isn't infested with 680x0 clones that are not Macs, IMHO.
Also, I didn't say the 600 point API would be easy, just possible. Also,
are there really 600 OS trap call, or are there 600 functions that might
boil down to fewer unique OS calls. I don't know much about ms-windows.
Someone mentioned that there would need to be a "vm386" mode. That would
be true for current design, but if they had designed there processor
correctly, it would not be necessary (nor would the vm86 mode, either).
The 680x0 processors (x > 0) can do virtual machines of themselves.
So can IBM 370 series processors. (You know, that famous OS called VM.
You can, run MVS or another VM underneath it!).
======================================================================
Corey Minyard