Re: [hatari-devel] 68030 MMU work

[Andreas Grabher <andreas.grabher@xxxxxxxxxxxx>]

Thanks for testing and finding the problem.
After fixing transparent translation and function code handling, the MMU should now work properly.
I think it would make sense, to re-test other software that uses the MMU. I think there is a good chance that previously occurring problems are now gone.
The now fixed problem with transparent translation in fact disabled address translation in many cases.

The last source of problems might be exceptions. Some more work needs to be done there. This might this time also affect 68040 MMU, but i'm not sure yet.

Regards,

Andreas


Am 03.10.2012 um 23:52 schrieb Thomas Huth:

> 
> Hi Andreas,
> 
> thanks again for the patch. I've committed it now to the Hatari
> repository, together with a patch/hack needed for the MMU to access our
> SysMem in Hatari.
> 
> However, this needs some investigation on a real Falcon, I think. If
> somebody has a Falcon up-and-running, could you please try whether it
> is still required there to switch to supervisor mode when accessing
> addresses below 0x800 ? I mean, just do something like this in
> user-mode:
> 
>  move.l  $700,d0
> 
> Does this cause a bus error on the Falcon or does it work?
> 
> Regards,
>  Thomas
> 
> 
> Am Sat, 29 Sep 2012 10:34:38 +0200
> schrieb Andreas Grabher <andreas.grabher@xxxxxxxxxxxx>:
> 
>> Dear Harati community!
>> 
>> I've improved my 68030 MMU emulation for Previous. Now function codes
>> should be handled correctly in all cases. Also a bug in transparent
>> translation is fixed. Functions for MOVES (memory access using SFC
>> and DFC registers) are now implemented.
>> 
>> I think this patch is worth merging into Hatari.
>> It's patch 215 in Previous' SVN repository.
>> 
>> Regards,
>> 
>> Andreas
>> 
> 
>