Something noone has answered with "yes" or "no" yet:

To avoid channel swapping in all situations and leave the CPU load lighter, should we not just lose the two least significant bits and simply use 32-bit L-R pairs as is done in my patch? As soon as the reason for the two bit delay is discovered, remove the shifting. That still leaves a dynamic range of ~80db which is still larger than the new volume range and sounds just fine really. I'm not at all for increasing the number of FIQs.

I'll emphasize again that PP502x (maybe 5002 as well) based iPods, since they need no shift hack, will also be able to send data directly out of the PCM buffer, lighten their FIQ load and become candidates for DMA audio transfer. Those with iPods can take about 5min to verify that it also works for them. Time to dump the excess of 32-bit samples imho since it is now known how to use 16-bit.
  ----- Original Message -----
  From: Antonius Hellmann
  To: Rockbox development
  Sent: Saturday, May 19, 2007 4:14 AM
  Subject: Sansa: avoid channel swapping issues

  Based on an idea of jhMikeS I modified svn code (i2s-pp.c) in the following way:

  // IISFIFO_CFG |= 0x33; /* 12 slots full/empty config */
      /* Bit2-3, Bit6-7 seem to have no meaning */
      IISFIFO_CFG |= 0x22; /* 8 slots full/empty config */
  which solves the channel swapping issues. The modification triggers the interrupt already when 8 slots are empty, giving the system additional 45.3usec to execute the fiq without the FIFO running empty. With current frequency setting the cache releases the bus in time to execute the next fiq. The drawback is, that 50% more fiq calls are issued. But the fiq handler itself has some performance tuning potential.
Received on 2007-05-22