On Thu, 21 Apr 2016, Reimar Döffinger wrote:
On 20.04.2016, at 23:59, Marton Balint <c...@passwd.hu> wrote:Signed-off-by: Marton Balint <c...@passwd.hu> --- ffplay.c | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/ffplay.c b/ffplay.c index 804bcbc..89a34d2 100644 --- a/ffplay.c +++ b/ffplay.c @@ -2578,12 +2578,17 @@ static int audio_open(void *opaque, int64_t wanted_channel_layout, int wanted_nb static const int next_nb_channels[] = {0, 0, 1, 6, 2, 6, 4, 6}; static const int next_sample_rates[] = {0, 44100, 48000, 96000, 192000}; int next_sample_rate_idx = FF_ARRAY_ELEMS(next_sample_rates) - 1; + char driver_name[32]; env = SDL_getenv("SDL_AUDIO_CHANNELS"); if (env) { wanted_nb_channels = atoi(env); wanted_channel_layout = av_get_default_channel_layout(wanted_nb_channels); } + /* By using the default dmix sample rate we should be able to avoid the + * ALSA resampler because using it causes small buffers and underruns. */That only makes sense if the alsa device used actually uses dmix. Plus I am generally sceptical about such a hackish workaround (in particular making assumptions on internals like default sample rate) for what sounds like an ALSA issue (assuming we can't/shouldn't just request a larger buffer or query the "native" sample rate).
You're right, this is a hack, and probably alsa is buggy, or SDL could create the player differently. I tried to reproduce the issue with pure alsa, unfotunately without success.
In the meantime I also found another workaround, setting the SDL_AUDIO_ALSA_SET_BUFFER_SIZE environment variable, as this causes the initialized alsa buffer to have 4 periods instead of 2, which also makes the issue disappear. Setting this (if the variable is not already defined) is an acceptable workaround?
I see little chance of me digging into the deep of ALSA, I have no better idea which benefits everybody other than setting the SDL_AUDIO_ALSA_SET_BUFFER_SIZE environment variable.
Regards, Marton _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
