Busy wait for screen lines

[Quagliarulo]

Hello, consider this code (adapted from an exercise I'm doing - I can provide the real source if needed):

; Wait for line 100
loop1:
	move.l	$dff004,D0
	lsr.l	#8,D0
	and.w	#%111111111,D0
	cmp.w	#100,D0
	bne.s	loop1

; Wait for line 310
loop2:
	move.l	$dff004,D0
	lsr.l	#8,D0
	and.w	#%111111111,D0
	cmp.w	#310,D0
	bne.s	loop2

; Do something and repeat forever
	bsr do_something
	bra loop1

Assume we are in lores 320x256. We wait for line 100, then we wait for line 310, which is much further down the screen. Removing the first loop, I would expect "do_something" to be executed at the same pace as before; instead, it's called twice as fast.
Why is this?

[ross]

Originally Posted by Quagliarulo

Hello, consider this code (adapted from an exercise I'm doing - I can provide the real source if needed):

; Wait for line 100
loop1:
	move.l	$dff004,D0
	lsr.l	#8,D0
	and.w	#%111111111,D0
	cmp.w	#100,D0
	bne.s	loop1

; Wait for line 310
loop2:
	move.l	$dff004,D0
	lsr.l	#8,D0
	and.w	#%111111111,D0
	cmp.w	#310,D0
	bne.s	loop2

; Do something and repeat forever
	bsr do_something
	bra loop1

Assume we are in lores 320x256. We wait for line 100, then we wait for line 310, which is much further down the screen. Removing the first loop, I would expect "do_something" to be executed at the same pace as before; instead, it's called twice as fast.
Why is this?

It depends on what 'do_something' does

If it is a very fast routine it can also be performed twice in the same video line.

[mcgeezer]

Originally Posted by Quagliarulo

Hello, consider this code (adapted from an exercise I'm doing - I can provide the real source if needed):

; Wait for line 100
loop1:
	move.l	$dff004,D0
	lsr.l	#8,D0
	and.w	#%111111111,D0
	cmp.w	#100,D0
	bne.s	loop1

; Wait for line 310
loop2:
	move.l	$dff004,D0
	lsr.l	#8,D0
	and.w	#%111111111,D0
	cmp.w	#310,D0
	bne.s	loop2

; Do something and repeat forever
	bsr do_something
	bra loop1

Assume we are in lores 320x256. We wait for line 100, then we wait for line 310, which is much further down the screen. Removing the first loop, I would expect "do_something" to be executed at the same pace as before; instead, it's called twice as fast.
Why is this?

This is something that can trip you up if you don't understand the video beam properly.

I would suggest you watch the first 4 minutes of the video below:
[ Show youtube player ]

The Amiga can track the 312 vertical PAL lines but you also have the horizontal time it takes to get from one line to the next. As Ross says your routine is probably executing twice on the same line because your loop is checking if it present on the same line.

If you want to fix it, then check you have reached a line (which you have), then you can either wait for the beam to reach a horizontal position on the line or wait until you reach a VPOS that is different to the one you are currently on.

This tripped me up on one Bomb Jack Beer Edition where the round start sequence moved faster on AGA machines.

Geezer

[ross]

There is also another reason why I ALWAYS advise against doing such a single check.

Nobody assures you that your code can be executed on that line!
Okay, no problem in a test program, but when does the program get complex?
When do you have IRQs arriving at unpredictable moments (like a CIA timed IRQ) and it use all the line time?
You risk failing because, for your check code, this line is never existed and, for the frame, code is skipped...

So always do a cross check and not for Z but for different conditions.

[meynaf]

I would personnally never do an equality check for a scan line, but a check for greater than. This alone can eliminate a lot of troubles.

[Quagliarulo]

Thanks for all your advice. I know that the code I posted is blasphemy, but I had to understand.
The BNE was the problem!