Re: [hatari-devel] Bootup differences

[Nicolas Pomarède <npomarede@xxxxxxxxxxxx>]

Le 16/06/2013 22:03, Eero Tamminen a écrit :
> Hi,
>
> I'm seeing differences in profiles of successive, identical
> EmuTOS boots with Hatari.  Profile is from Hatari start until
> EmuTOS desktop first calls evnt_multi().
....
                                                 ---^   ^----
> I was assuming that host clock could through RTC affect when interrupts
> start ticking, but disabling that didn't help.
>
> Nicolas, what else could be causing differences between successive
> TOS bootups?
>
> (This isn't a new problem, it happens also with e.g. Hatari v1.6.2.)
>
>
> 	- Eero



well, those differences are rather small in the end, and are certainly 
higher on a real STF where the boot state is not necessarily known, 
because some components like MFP or ACIA use their own clock, and this 
clock is not synchronised to the 68000's one.

That's why when the CPU needs to "talk" to the ACIA for example, a use 
of the E clock is required to synchronise both sides. This E clock has a 
fixed frequency (1/10 th of the CPU freq), but its phase is not known at 
start, this means that the same instruction at the same time will 
sometimes wait for 0 cycle, 2 or 4, etc.

So, you can consider that some components in a ST create some kind of 
"fussyness", even if there's no real randomness in it.

Nicolas