Re: [hatari-devel] Suspected Hatari 2.3.0 regression (prefetch?)

[Nicolas Pomarède <npomarede@xxxxxxxxxxxx>]

Le 16/12/2020 à 01:13, Eero Tamminen a écrit :

> According to real device test results done with byte and word access:
> https://git.tuxfamily.org/hatari/hatari.git/tree/tests/buserror/results
>
> Only palette regs are readable, not the the area
> after them.
>
> If prefetcher triggers now (2.2 -> 2.3) access
> errors when reading something that program didn't
> actually read, that does sound like a regression
> bug.

Hi

well toni improved cpu prefetch/bus error a lot and this new version is 
now in hatari 2.3, so this might explain differences (although I have 
the feeling cpu core is now more correct than before, so maybe it's 
hatari 2.2 that did not have the correct behaviou ?)à

> Nicolas, any idea why MMU would prevent those?
>
>
> (I've been wondering about why some programs now
> need MMU, and bug like this sounds a plausible
> explanation.  Toni is less likely to run across
> issues of bus errors happening at wrong place
> with Amiga, or it could be Hatari integration
> issue.)

MMU emulation for 68030 is complex, it's possible that at this point not 
all HW accesses that trigger a bus error in non MMU mode are correctly 
emulated in MMU mode and the bus error is not correctly triggered, which 
would allow buggy programs to work only in MMU mode instead of crashing.

Now that we have several examples in compat list, I will check them 
later in more detail to really see what is happening at the cpu core level.
But in all cases it would be good to have confirmation on a real falcon 
if these programs are working or not.

Note that as real falcon have a MMU enabled, it's possible that those 
programs are buggy and only work emulated when using the MMU because it 
catches bus error.
Anyway, can't be sure at the moment, need to look at precise examples.

Nicolas