On 2018-10-26 09:00, Jarkko Nikula wrote: > On Thu, Oct 25, 2018 at 04:48:24PM +0300, Peter Ujfalusi wrote: >> For samples sizes of 20/24/32 bit the DRR2/DXR2 registers must be used to >> receive or transmit the most significant part of the 32bit sample. >> We can not trick the system by using the DRR2/DXR2 register offset and use >> 32bit element size since these (and other McBSP) registers are 16bit ones. >> >> We would need support for port_window in cyclic mode from the sDMA driver, >> but it is not straight forward as we would need to have the sDMA frame to >> cover the 32bit (DRR2+DRR1/DXR2+DXR1), but the frames must cover the ALSA >> period to be able to receive periodic interrupts. >> >> Since the 32bit samples are not working (DMA timeout), just remove it in >> case we have McBSP with reg_size of 2. >> >> Signed-off-by: Peter Ujfalusi <peter.ujfalusi@xxxxxx> >> --- >> Hi, >> >> I have noticed that on Nokia n810 32bit samples just don't work at all for audio >> >> Regards, >> Peter >> >> sound/soc/omap/omap-mcbsp.c | 5 +++++ >> 1 file changed, 5 insertions(+) >> > I guess this same holds for OMAP1 too as it has 16-bit McBSP registers > too. Yes, it does. That's why I check for the reg_size. > Acked-by: Jarkko Nikula <jarkko.nikula@xxxxxxxxxx> Thanks. There is a slight chance that I can make the 32bit samples to work on the sDMA side, but it is a long shot: Add support for port_window to cyclic Set the dev_width for omap1/omap2420 to 16bit (as this is the case) If port_windows is set and it is multiple of dev_width use single indexed addressing on the McBSP side instead of constant set the csdp.xxx_packed set the csdp.data_type to 32bit set the ES (element size) to 16bit set EI to -4 Aaargh, no, it can not be done... The TRM states that Packed access only works if EI=1 in single/double indexed modes. I don't see how this can be done :( - Péter Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki _______________________________________________ Alsa-devel mailing list Alsa-devel@xxxxxxxxxxxxxxxx http://mailman.alsa-project.org/mailman/listinfo/alsa-devel
