Re: [mpg123-users] Increasing CPU usage in mpg123_decode with rate conversion.
Brought to you by:
sobukus
From: Rune T. <ru...@in...> - 2009-09-15 16:24:31
|
Think I figured out (at least part of) the problem. In the function ntom_frmouts() in ntom.c, it uses a for loop over 0 to framecount to do sample conversion. The problem on streaming audio is that the framecount is an ever increasing number, so that for loop takes longer and longer to complete. Not sure how to fix. For a stream, the framecount should probably be reset everytime mpg123_decode() is called with a new input buffer, and there are no leftover samples. > -----Original Message----- > From: Thomas Orgis [mailto:tho...@or...] > Sent: Monday, September 14, 2009 4:09 PM > To: mpg...@li... > Subject: Re: [mpg123-users] Increasing CPU usage in > mpg123_decode with rate conversion. > > Am Mon, 14 Sep 2009 12:07:00 -0500 > schrieb "Rune Torgersen" <ru...@in...>: > > > With rate conversion, my testapp starts to spend > progressively more time > > in the mpg123_decode function. > > Please describe "progressively". That the resampling code can > take more CPU time, generally, is not surprise since only the > native rate decoding is accerelated using assembly optimizations. > > > Alrighty then, > > Thomas. > |