Re: [hatari-devel] UAE 68030 MMU + prefetch + instruction and data cache emulation

[Laurent Sallafranque <laurent.sallafranque@xxxxxxx>]

I've tested a few more demos.

The cycle exact seems to run well for demos that are not MMU related 
when I test them with the MMU.
The cycles bahave the same than when I test without MMU.

I'll do more tests.

Cycle accuracy still need some work anyway.
Regards
Laurent



Le 03/08/2017 à 18:47, Nicolas Pomarède a écrit :
> Le 28/07/2017 à 15:25, Nicolas Pomarède a écrit :
>
> > MMU not enabled :
> >
> > - Cycle Exact    68030 CE + caches (internal mode=4 op_smalltbl_24_ff)
> > - Prefetch    68030 prefetch, no cache (internal mode=3 
> > op_smalltbl_22_ff)
> > - none        68030 basic, no cache (internal mode=0 op_smalltbl_2_ff)
> >
> >
> > MMU enabled :
> >
> > - Cycle Exact    68030 MMU + prefetch + cache (internal mode=6 
> > op_smalltbl_34_ff)
> > - Prefetch    68030 MMU + prefetch + cache (internal mode=6 
> > op_smalltbl_34_ff)
> > - none        68030 MMU, no cache (internal mode=5 op_smalltbl_32_ff)
> >
> >
> > NOTE :
> >
> >   - there's no "cycle exact" mode for now in 68030 MMU, it's the same 
> > emulation as "prefetch". For "cycle exact" 68030 (instructions 
> > tail/head and more accurate cycles) don't use MMU mode.
> >
> >   - even in "cycle exact" mode without MMU, cycles are not always 
> > correct yet, this is still work in progress.
>
> As a followup to this description, a specific opcodes table was added 
> for 68030 + MMU in cycle exact mode.
>
> So now, it uses internal mode=7 and table op_smalltbl_35_ff, which 
> should basically gives the same cycles results as already available 
> for "68030 without MMU + CE" (ie internal mode=4 op_smalltbl_24_ff), 
> but with additional MMU support.
>
> Eventually, this mode will be the prefered one to emulate a Falcon 
> (when cycle accuracy will be complete for all cases)
>
> Nicolas