[Coco] OVCC development decisions

Joel Rees joel.rees at gmail.com
Fri Nov 22 08:14:23 EST 2019


なんや

2019年11月15日(金) 15:05 Walter Zambotti <zambotti at iinet.net.au>:

> Hi people.
>
> I need to make a decision as to a feature I am adding to OVCC.
>
> I have finished testing the MMU hardware emulation and it considerably
> speeds up the CoCo CPU performance.


My Intel i7-3740QM which previously topped out at 160mhz now reaches 225mhz.
> Which is about 40% faster.
>

Say what?

I have now set up a mechanism to switch between hardware MMU and software
> MMU from the configuration
> panel (just like with the CPU).
>

Sounds good.

However this slows down the max speed to 210mhz.  So a little is lost when
> it is made configurable.
>

Wait, is the effective clock rate of the 6809?

That's way fast!

Now the hardware MMU version will only work on Linux (only the software
> version will work on Windows).
>

Okay, it sounds like it's working well enough on Linux that I need to get
it.

So instead of making it selectable which could only happen on Linux (as
> there is only one choice on windows)
> I though rather than making it configurable I could make it a separate
> executable.
>

How big is the effected part? Could you build it so both exist in the
object?

May take some munging or playing with namespaces.

The Linux would have two builds one with hardware MMU and one without
> (Windows would just have one
> without) and for us speed freaks we would run with hardware MMU unless we
> found some incompatibility and
> had to run the other.
>

On the other hand, if that's the effective speed of the emulated 6809, slow
is fast enough for me.

My thinking is if the hardware MMU works compatibly 100% of the time on
> Linux why else would you select the
> slower model.
>

There is a point to that.

Oh and my new land speed record for OVCC is 566mhz (hardware MMU) on an
> i7-8700.
>

At  those speeds, maybe we should see if we can give it 32 bit index
registers and run Debia on it.

I have also decided to remove the assembly version of the 6309.  The code
> will remain in the git repo for
> interest only.
>

I'm not seeing what you mean about assembly version.

I have also managed to get the Linux version of OVCC running on WSL on
> Windows.  It needs an Xserver and a pulse
> audio server installed but it does work.
>

However it is not very stable.  The GL support in the Xserver is not perfect
> and it is not possible open up any secondary
> window (like the config or about or file dialog) without it crashing when
> the secondary window closes.
>

I can see there are some unimplemented X-GL protocols.  I might need to try
> a different Xserver!
>
> My intention here is if OVCC can run well in WSL then there is no longer a
> need to maintain two builds!


It does however mean  a more difficult setup on Windows for OVCC.
>
> Thoughts guys/girls
>

Did you say you were building it on MSWindows using Microsoft's compiler or
Cygwin?


More information about the Coco mailing list