Jan writes:
>
> In practise your CM program will run faster after it is relinked with
> OCTCOMP, but if you recompile your source code to a real NM program you
> will get a much better performance.
>
> OCTCOMP does not translate your code to NM, it relinks your code so only
> calls to CM libraries will be switched to NM libraries.

Actually, it does.  It translates ("compiles") each CM instruction
(or block of CM instructions that no one branches into or out of)
into zero or more NM instructions ... NM instructions that accomplish
the same goal as the CM instructions (but more efficiently than
what the emulator does).  The resulting NM code is appended to the
end of a copy of the original CM code, because in some cases the
emulator may still need to assist in the execution.

The resulting NM code is still subject to exactly the same
limitations as the original CM code: 64 KB CM stack, 16-bit
addressing, etc.

The primary result is a file that's still called a "CM PROG" file,
but it runs about 3 times faster, and is about 8 times bigger.

And, the calls the CM program made to intrinsics?  The OCT'd
version still calls the same CM intrinsics, *not* the NM
counterparts.  Why?  *because the program is still CM* ... it's
just faster than before because the emulator is bypassed most of the
time.

--
Stan Sieler                                          [log in to unmask]
                                     http://www.allegro.com/sieler.html