Re: [hatari-devel] No address error when accessing odd addresses?

[Thomas Huth <th.huth@xxxxxx>]

Am Wed, 11 Sep 2013 08:41:35 +0200
schrieb Uwe Seimet <Uwe.Seimet@xxxxxxxxx>:

> Hi Thomas,
>  
> > > When the Hatari CPU type is 68000 and I run TOS 2.06 I don't get
> > > an address error when trying to read longs from an odd address,
> > > e. g. with
> > > 
> > > lea ODD_ADDRESS,a0
> > > move.l (a0),d0
> > > 
> > > This works fine, but I would have expected an address error as the
> > > 68000 does not support reading words or longs from odd addresses.
> > > Can it be that something does not work as it should?
> > 
> > I guess you've disabled the "Slower but more compatible CPU" option
> > in the System settings. In that case, the CPU core does not check
> > for odd addresses to be slightly faster. Could you try to enable
> > that option to see whether it makes a difference?
> 
> This option does not make a difference. I started Hatari with TT
> compatible settings, then I loaded a new configuration, which is ST
> compatible. Next I enabled "Slower but more compatible CPU" and reset
> Hatari. After that I can still read longs from odd addresses.
> By the way, when enabling "Cycle exact" and rebooting the IDE
> emulation does not seem to work anymore. Instead of the devices names
> for master and slave only "-" characters are returned as devices
> names.

Ah, wait, you're writing about the "Cycle exact" option... are you
using the WinUAE cpu core? ... that's slightly different and there might
be some bugs left in its option management (that's why it is marked as
experimental ;-)).

But first ... what did you exactly change in the GUI? The "Slower but
more compatible CPU" option is named there "Prefetch mode, slower"
instead, if I remember correctly.

Also, do you see a line starting with "Building CPU" on the console
when you change that option and return the emulation? If yes, what does
it exactly print?

 Thomas