> At Wed, 22 Nov 2006 10:44:55 +0100, > Jonas Eymann wrote: > > > > Hi everyone, > > > > I'm currently developing/ enhancing an ALSA driver for a Silicon > > Motion SM501 compagnion chip which contains also an 8051 µC acting > > as AC97 controller. It doesn't support DMA and so the copy of audio > > data is done in the driver. There is a small hardware buffer of 1920 > > bytes which is devided into two halves and used as ring buffer. When > > the hardware has finished playing one half, it sends an interrupt to > > the driver so it can fill this half with new audio data while the > > hardware plays from the other one and so on. In the driver, I use an > > intermediate buffer with vmalloc(). > > > > Actually, I have a working driver now, but one small problem remains > > and I'm absolutely running out of ideas how to fix it: when the > > playback starts, the hardware expects the buffer to be filled > > completely (both halves), but the copy routine which does this first > > filling (snd_sm501_pfill()) only gets one period, not the two which > > are needed to fill the buffer completely. The second half it just > > reads zeros. It seems that the intermediate buffer is not yet filled > > with more data. This results in a gap of one period at the beginning > > of the playback (it's the second period that doesn't get copied). > > Hm, this kind of implementation gets more common nowadays for small > devices, so we may need a better/simpler framework in furture. > > Usually, the filler routine should fill the data as much as possible > at the time it's called. The size of available data on the > itermediate buffer can be calculated via PCM appl_ptr. > > Possibly, helper functions defined in pcm-indirect.h could be used in > your case. Check the alsa-devel list archive in the last months, > threads "Driver design question". The discussions went on there > regarding a driver implementation without DMA. (The pcm-indirect.h > routines didn't work for his case well, unforutnately, though...) > > > Takashi Thanks for your answer, Takashi! when I check the appl_ptr in the trigger callback, it gives me exactly one period of available data. As soon as we are in the copy function called from the isr (snd_sm501_copy_pdata), the appl_ptr is a few thousand periods further. Is there a simple way to change the default value for this first portion of available value? I'm also trying to implement the driver using the pcm-indirect.h. I just thought maybe there is a simpler solution. Thx for your answers! Jonas ------------------------------------------------------------------------- Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT & business topics through brief surveys - and earn cash http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV _______________________________________________ Alsa-devel mailing list Alsa-devel@xxxxxxxxxxxxxxxxxxxxx https://lists.sourceforge.net/lists/listinfo/alsa-devel
