Re: [hatari-devel] Hatari data cache tests

[Nicolas Pomarède <npomarede@xxxxxxxxxxxx>]

Le 20/06/2015 00:24, Nicolas Pomarède a écrit :
> I need to write some simple cache clear / fill cache / read data tests
> to simulate all this.

as a quick check, I wrote a few line under devpac :

	move.l	$40000,d0
	nop
	move.l	$40000,d0
	nop
	move.l	$40000,d0
	nop

this is the result with cycles when tracing cpu_all :

cpu video_cyc=127276 300@248 : 000C4EE8 2039 0004 0000           MOVE.L 
$00040000,D0
cpu video_cyc=127284 308@248 : 000C4EEE 4e71                     NOP
cpu video_cyc=127286 310@248 : 000C4EF0 2039 0004 0000           MOVE.L 
$00040000,D0
cpu video_cyc=127292 316@248 : 000C4EF6 4e71                     NOP
cpu video_cyc=127294 318@248 : 000C4EF8 2039 0004 0000           MOVE.L 
$00040000,D0
cpu video_cyc=127300 324@248 : 000C4EFE 4e71                     NOP

So, the first move takes 8 cycles, it's a miss since no one accessed 
$40000 before. Next moves will be some hits and we see they now take 6 
cycles.

At least, we see that cache has an effect on number of cycles (but I 
didn't check 68030 doc, are 8 cycles for miss and 6 cycles for hit the 
expected values when accessing 16 bit ram ?)

Nicolas