Interrupt and VDP processing about devkitsms HOT 8 CLOSED

I see your point. It's the way the SMS VDP works, unfortunately even setting a register (in your case, the horizontal scroll register) affects the destination address of VRAM operations (in your case, the address you write using the setTile call).
SMSlib doesn't address this problem at all, as any possible solution would be cumbersome so I suggest you some possible workaround scenarios:

• make sure you do all your setTile calls before (or after) the horizontal interrupt happens
• split the setTile calls in two parts so that you can perform the first part before the horizontal interrupt happens and the second part after that, making sure you set again the destination address before the second part of setTile calls
• create your own atomic and uninterruptible setTile API that sets address and writes the data all before re-enabling interrupts (this is not the suggested way as it would slow down quite a bit and also keep your horizontal interrupt to be fired at the right line)

from devkitsms.

Thank you.

I can agree your opinions.

from devkitsms.

Can you elaborate on this? What are the glitches you're getting?

from devkitsms.

(1) main loop invokes setTile api repeatedly
(2) interrupt routine invokes SMS_setBGScrollX

→ setTile api failed. Because VRAM address is changed by SMS_setBGScrollX().

I could figure out the reason.

My application invoke DI before setTile to avoid interrupt.
But, the crt0 invokes EI instruction in the VDP APIs(see the below). So, I can't disable interrupt... ;-<

_SMS_crt0_RST08::
ld c, #0xBF ; set VDP Control Port
di ; make it interrupt SAFE
out (c),l
out (c),h
ei
ret

from devkitsms.

It looks like we need an API like SMS_copySpritestoSAT() for all VDP instructions, just like sprite APIs.

from devkitsms.

Strategy:

(1) VDP APIs just queue VDP Address and DATA.
(2) Commit all queue data in VBlanking by like CommitVDPData() API and so on.

from devkitsms.

(2) Commit all queue data in VBlanking by like CommitVDPData() API and so on.

I've created "Queue" mechanism. It seems okay for my program. Thanks.

from devkitsms.

That's a good approach if you don't have to update too much data, as the vblank phase is anyway pretty short.
Good you found an appropriate workaround! 😄

from devkitsms.