🚀 go-pugleaf

I'm writing a DOS emulator.  It's very very close.   Just got 
TASM running last night on it.  PK*ZIP too.  

In running some msdos executables on this beast, I find that sometimes
I get slightly different results from the same codes run on PC's.
This is not a good thing, for the most part. :-).  The differences are
from TC's runtime floating emulation (and at the 5th digit or so).
Many other programs run exactly the same, including a small suite of
programs which are designed to test the opcodes which affect flags
(add, sub, or, rcl, etc).

I recall reading somewhere (net.rumor, maybe) that the TC floating
runtime (or in some libraries, or fractint, or somewhere) that someone
discovered how to speed up floating point speed by utilizing 
some opcode (I think it was either div or mul, but perhaps it was
something else) which consistently set the flags in some usable 
way (sign or zero, perhaps) but also that those particular flags were
offically "undefined" for this opcode.  I tend to believe a rumor
such as this based on past experience debugging this beast.  

I realize that the above paragraph is somewhat vague, but if anyone 
out there knows what I'm talking about or has any similar war stories
I'd love to hear them.  Keep in mind that what I'm trying to avoid
is single-stepping through 100K of instructions without really being
sure of what I'm looking for.

Thanks in advance...

JHH





--
Jim Hudgens 			Supercomputer Computations Research Institute
hudgens@sun13.scri.fsu.edu