You may want to try enabling the timer interrupt, and put your
countdown-til-buffer-swap in the timer interrupt handler, that way
you don't have to worry about synchronization. If you can eliminate
the need for timer 0, you can cascade 1 into 0, put the countdown in
timer 0, then generate a single interrupt when the buffer swap is
needed. If you're still getting the click, you may need to issue the
swap earlier (maybe the FIFO is starved)

John Marco Panettiere

--- In gbadev@y..., "Tristan Rybak" <tristan.rybak@w...> wrote:
> I think, that's not the problem. I isolated my code so there is no
other DMA transfer then that feeding SOUND FIFO's.
> And I don't use DMA interrupt when sound completes (I use VBLANK).
> I think the problem is in synchronizing DMA restart with VBLANK.
But I can not find working values of frequency/buffer size.
> When I use emulator and read DMA SOURCE ADDRESS at VBLANK IRQ (it's
possible in emulator) the address is pointing at the first byte after
my buffer and sometimes (lets say every 4th sound click:-) 16 bytes
after. I guess DMA is feeding sound fifo 16 bytes a time.
> But I know emulator is not enough reliable for such measurements.
> I have no idea what to do, If someone share with me working buffer
switching routine (at vblank) that would be great.
> Thanks very much
> Tristan
> ----- Original Message -----
> From: rogermilne100
> To: gbadev@y...
> Sent: Friday, January 04, 2002 6:17 AM
> Subject: [gbadev] Re: sound loop problem, please help!!!
>
>
> If you are performing DMA's *anywhere* in your code, while the
sound
> is playing, you risk pre-empting the CPU's interrupt handler from
> being able to feed more sound when the existing sample
completes.
> That situation caused random, intermittant clicking in my code.
> Basically, you really have to make sure that when the sound DMA
> completes its' last transfer, and generates an interrupt so that
it
> can be given more sample-data, you really need to service it
ASAP.
> DMA's take higher precidence than CPU, and if active when the
> interrupt is triggered, can cause a delay before the interrupt is
> serviced, and thus, a click is heard once in a while. There may
be
> more reasons for these clicks, but that's the only problem I ran
into
> which caused them. I hope that helps lead to a solution...
>
> Roger