Re: [hatari-devel] TT palette issue

["Roger Burrows" <anodyne@xxxxxxxxxxxx>]

On 14 Feb 2016 at 21:04, Eero Tamminen wrote:

> Hi,
>
> On 02/12/2016 12:42 AM, Nicolas Pomarède wrote:
> > Le 11/02/2016 23:08, Eero Tamminen a écrit :
> >
> >> Nicolas, should shifter accesses on TT (and Falcon) be on
> >> 4 cycle boundary like they're on ST/STE?
> >
> > Yes, I think the shifter accesses are also limited due to sharing memory
> > between shifter and cpu, but I have not measured it.
> > I think you should leave the call to M68000_SyncCpuBus_OnWriteAccess(),
> > it's clearer to have it at right at the start of the function and common
> > to all cases (and since there's no "color plasma" or similar demos on TT
> > or Falcon, it should not hurt in case where palette is changed only once
> > par line)
>
> Ok, I commited the fix with that change.
>
> I'm assuming that TT handles byte accesses to ST palette registers
> similarly to ST/STE, but it would be nice if somebody could verify
> that on real TT.
>
OK, I've done the tests on my TT.  The TT *doesn't* do it the same way as the
ST/STe.  Byte writes (even or odd address) to either the STe palette registers
or the TT palette registers are silently ignored.  Word accesses on word
boundaries are acted on (just verifying my code).  I didn't try unaligned or
long accesses.  If you need the source/binary, let me know.

> (And same info for Falcon would also be appreciated.)
>
I'll do that next.

Roger