[Coco] real-time mouse cursor

[Robert Gault]

Roger Taylor wrote:


> At 09:54 PM 1/26/2008, you wrote:

> 

>> Or better yet, replace the TST $FF03, BPL a@ with a SYNC instruction.

>> (Of course, time isn't really crucial if you're spending the bulk of 

>> it doing

>> nothing while waiting for the next interrupt.)

> 

> 

> 

> Doesn't this SYNC to any interrupt, though?

> 

> Btw, it just dawned on me after all the years... why do sprite calcs 

> like this:

> 

> LDD     ,X      screen data

> ANDA    ,U      mask data

> ANDB    1,U

> ORA     128,U   sprite data

> ORB     129,U

> STD     ,X

> 

> when this is the same:

> 

> LDD     ,X      screen data

> ANDA    ,U      mask data

> ANDB    1,U

> ADDD    128,U   sprite data

> STD     ,X

> 

> No reg.a/b OR'ing is needed. You can do a 16-bit simulated OR with the 

> ADDD instruction.

> And the speed increase is very noticeable over 64 sprites!  Actually, 65 

> now... for the mouse cursor.

> 

> I'm using direct offsets into the sprite and screen data instead of auto 

> incs ( ,r++) because an entire sprite is plotted with precalced byte 

> offsets like the above.  Spaghetti and meatballs, anybody?

> 

> With the 6309, "ANDD" is even possible for the masking.

> 

> LDD     ,X      screen data

> ANDD    ,U      mask data

> ADDD    128,U   sprite data

> STD     ,X

> 


It might work for you but I can see one cause for failure trying to 
replace ORA ORB with ADDD.

Let's take and example.
start ldd #$1234
       ora T1
       orb T2
       swi
T1    fcb $BB
T2    fcb $FF

regD will = $BBFF

start ldd #$1234
       addd T1
       swi
T1    fcb $BB
T2    fcb $FF

regD will = $CE33

I think you will find you can only use ADDD if your mask is $0000. OR 
may look like addition but it is not equivalent.