Am 02.05.2010 20:11, schrieb Jeff Goode:
On 5/2/2010 14:03, Jeff Goode wrote:
On 5/2/2010 04:44, Jens Arnold wrote:
On 01.05.2010, Jeff Goode wrote:
I'd like to know if there's some way to force SDL to behave
like a target, with a very small FIFO buffer rather than
its own large playback buffer, or at least not always
insist on its being full. Ideas?
Perhaps just make the mix buffer larger for the sim?
I have no real idea how to get around this while keeping SDL
(and hence portability).
I think you're right about this and I'm going to give that a try
next. The sim audio driver still won't "DMA" like a target but at
least it ought to fill the buffer as SDL expects. Latency will
skyrocket though. Instead of 23 ms tops it'll be more like 250 ms.
I guess this isn't a problem as long as we're just using the sim to
test code, but it might if we're going to use it as a basis for
Rockbox-as-app. In that instance, the Rockbox mixer code would have
to be thrown out in favor of SDL mixer or some other library function.
I just thought of another implication. If the Rockbox mixer has to be
disabled later for the sim or Rockbox-as-app in favor of a library
function, that would imply that some of the mixer code (such as the
mixer ISR callback) doesn't belong in pcmbuf, but lower down toward
the driver level, perhaps in pcm.c. Code in pcmbuf really ought not
be target specific. Actually, neither should pcm.c but it's closer to
the hardware level. I'm a bit conflicted here. Either way it's
bloody mess.
Jeff
Don't think about the future too much. You could invent some #defines,
but just don't worry about RaaA too much :)
Best regards.
