Re: [hatari-devel] Release 1.6.2 in mid May

[Nicolas Pomarède <npomarede@xxxxxxxxxxxx>]

Le 03/05/2012 01:50, Roger Burrows a écrit :
> Hi,
> On 2 May 2012 at 12:02, Laurent Sallafranque wrote:
> > >   Maybe we can prepare a synthetic test for the real TT and measure div's
> > cycles against video counter?
> > >   As soon as I receive my TT, I'm able to run such a test e.g. for five
> > days.
> >
> >
> > That would be great, we could have a better accuracy for 68030 divs.
> > But it means we'll have to guess the div algo of the 68030 by ourselves.
> >
> > Nice challenge to come :)
> If someone has such a test already written, I volunteer to run it on my TT.
>
> Regards,
> Roger Burrows

Honnestly, I don't want to be too pessimistic, but I think we will never 
"guess" the algo by running tests and taking averages.

For example, just have a look at the 68000 divs algo, it's really 
complicated, it depends on the number of 0 and 1 in each operand. Now if 
we consider the 68020+ algo is faster, this means it's even more 
complicated.

I'm afraid it will be hard to come to something close, only motorola 
knows the inner work.

Nicolas