These formats are sometimes advertised by drivers, e.g. SOF. The format is 3 bytes packed in 32-bit container, with the MSB zeroed out.

sample: 0x00123456

S24_LE format: b0 56 b1 34 b2 12 b3 00

S24_BE format: b0 00 b1 12 b2 34 b3 56

I only tested the S24_LE format with the SOF driver, S24_BE was added for symmetry only.

Signed-off-by: Pierre-Louis Bossart pierre-louis.bossart@linux.intel.com --- speaker-test/speaker-test.c | 16 ++++++++++++++++ 1 file changed, 16 insertions(+)

diff --git a/speaker-test/speaker-test.c b/speaker-test/speaker-test.c index 773af0a..fd13d88 100644 --- a/speaker-test/speaker-test.c +++ b/speaker-test/speaker-test.c @@ -285,6 +285,8 @@ static const int supported_formats[] = { SND_PCM_FORMAT_FLOAT_LE, SND_PCM_FORMAT_S24_3LE, SND_PCM_FORMAT_S24_3BE, + SND_PCM_FORMAT_S24_LE, + SND_PCM_FORMAT_S24_BE, SND_PCM_FORMAT_S32_LE, SND_PCM_FORMAT_S32_BE, -1 @@ -338,6 +340,20 @@ static void do_generate(uint8_t *frames, int channel, int count, *samp8++ = BE_INT(res.i) >> 8; *samp8++ = BE_INT(res.i) >> 16; break; + case SND_PCM_FORMAT_S24_LE: + res.i >>= 8; + *samp8++ = LE_INT(res.i); + *samp8++ = LE_INT(res.i) >> 8; + *samp8++ = LE_INT(res.i) >> 16; + *samp8++ = 0; + break; + case SND_PCM_FORMAT_S24_BE: + res.i >>= 8; + *samp8++ = 0; + *samp8++ = BE_INT(res.i); + *samp8++ = BE_INT(res.i) >> 8; + *samp8++ = BE_INT(res.i) >> 16; + break; case SND_PCM_FORMAT_S32_LE: *samp32++ = LE_INT(res.i); break;