+100 - I was just writing the same thing.  DOSBox is not cycle accurate.

I disassembled the code generated by Open Watcom 1.9 and it is very memory
intensive.  Memory accesses on real hardware are fairly expensive, and
there are a lot of them.  A small difference in the instruction mix and how
DOSBox handles those instructions would make a huge difference in the
timings under DOSBox that don't reflect what happens on real hardware.


-Mike


On Wed, Jul 22, 2026 at 1:24 PM Eric Auer via Freedos-devel <
[email protected]> wrote:

>
> Hi Jim
>
> According to some people writing retro-compatible games,
> dosbox is probably NOT very realistic in speed comparisons.
>
> So it would be interesting to compare on real hardware :-)
>
> Regarding size, a good compiler for "small" hardware
> should be able to omit unused runtime parts if possible.
>
> Cheers, Eric
>
> > ...this is emulated in DOSBox (not real CPU) but it's
> > still a good comparison.
> >
> > It's also interesting to see the EXE size differences:
> >
> > IA16-GCC: 19728 Bytes (19 kiB)
> > Watcom: 20544 Bytes (20 kiB)
> > Turbo C++: 17510 Bytes (17 kiB)
>
> How about the UPX-UCL compressed size differences, by the way?
>
> > Turbo C/C++ was very good at keeping the EXE size small, but the
> > counter is that Open Watcom meets a later C standard so probably has
> > more in the runtime...
>
>
>
>
> _______________________________________________
> Freedos-devel mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/freedos-devel
>
_______________________________________________
Freedos-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/freedos-devel

Reply via email to