Re: [hatari-devel] WinUAE and 030 cache hits/misses?

[Eero Tamminen <oak@xxxxxxxxxxxxxx>]

Hi,

On sunnuntai 27 tammikuu 2013, Laurent Sallafranque wrote:
>  From my point of view, it's inconsistant to have so many different
> cores for the Falcon.
> The Falcon is already quite slow to emulate (it needs a good hardware).
> 
> So, choosing cycles exact instead of no cycle exact should not take that
> much extra speed comparing to the DSP calls.

Ok, I changed to using Cycles_GetCounter(CYCLES_COUNTER_CPU) [1] in
profiler.  As it works for old UAE core and the default WinUAE CPU core
function, I think it's the best option.    Now one can get both
cycle counts and i-cache misses with the profiler.

[1] It returning bogus values instead of zeros on "non-supported"
    core variants is a bit annoying though.


> I think we should have only 1 CPU (2 if we want to keep with/without
> MMU), but this have no sence in a Falcon, and focuse on this(these) CPUs
> instead of having 3/4 different CPUs with different behaviour.

Agree.  Cycles_GetCounter() doesn't seem to work for MMU version
either, could that be fixed at some point?


Btw. What about the DSP_Run() call which didn't have "*2" in it,
should that be fixed?  It was in the old UAE core non-prefetch
variant (m68k_run_2()).


	- Eero