Re: [hatari-devel] EXA demo Entracte : Improvement in search

[Laurent Sallafranque <laurent.sallafranque@xxxxxxx>]

I've got the same value for D0, ie $3e65.

What I find strange is that D0 increase in a loop that waits the time of 
a VBL, and it only increase 15 973 times during a VBL ($3e65) ?
I would have expected a much higher value.
At least, the add is a word add, so it may overflow many times during 
the test, which also means that the accuracy will have to be very 
precise if we must reach this $4100 to $4200 area after many "overflow 
loops"





Le 05/02/2017 à 17:46, Nicolas Pomarède a écrit :
> Le 05/02/2017 à 11:48, Christian Zietz a écrit :
> > Laurent Sallafranque schrieb:
> >
> > > The problem is in the 5 last lines : the code test for a min and max
> > > value for register DO.
> > > if D0 is not between $4100 and $4200, the code goes to $4eb1c and this
> > > quit the demo immediatly.
> >
> > To me this simply looks like a CPU speed test, counting the number of
> > loop iterations between two VBLs. I'd assume that the demo requires a
> > certain CPU speed to run correctly and so it wants to quit on (for
> > example) over-clocked CPUs.
> >
> > I think the 68030 cycle accuracy of Hatari is not good enough here for
> > this test to pass.
>
> You're right, this compute cpu speed.
>
> At the moment with my dev version I get D0=$3e65, which is ~5% too 
> slow compared to real HW $4100-$4200 for this code loop :
>
> $0001cd1e : 4a78 0468                          tst.w     $0468.w
> $0001cd22 : 6600 0006                          bne       $1cd2a
> $0001cd26 : 5240                               addq.w    #1,d0
> $0001cd28 : 60f4                               bra.s     $1cd1e
>
> Unfortunately, 100 % accuracy will not be reached immediately in 
> Hatari, some complex parts of the 68020/68030 pipeline and bus 
> sequencer are not known yet to ensure exact similar behaviour 
> regarding how the cpu order the memory accesses to be optimum.
>
> Nicolas