Re: [hatari-devel] Issues with cache hits/misses?

[ Thread Index | Date Index | More lists.tuxfamily.org/hatari-devel Archives ]


Le 01/06/2015 23:04, Nicolas Pomarède a écrit :

I don't see either :) But just try the lines I show in previous mail,
and you will see that speed gets higher.

No, this just returns the same value, so I don't see how it could change
the emulated ST.

"mystery" solved, altough cache and direct memory read should return the 
same value, the TOS was slow during blitter access, and it turned out 
that when I factorized some code earlier, I forgot to remove an old call 
that wrongly marked the IO region as cachable :( So, when TOS waited for 
blitter to end, it never read the real time status, but an old value and 
waited much longer than expected...
That's better, because I didn't see how a shift in the code could have 
made the emulation slower ; in fact it was the shifted value that was 
wrong in the particular case of the blitter (which explains why TOS 3 
for TT was not slowed down, it doesn't have a blitter :) )
Nicolas



Mail converted by MHonArc 2.6.19+ http://listengine.tuxfamily.org/