Subject: Re: [linux-audio-dev] more numbers for multichannel playback
From: Bill Gribble (grib_AT_billgribble.com)
Date: Tue Apr 04 2000 - 19:01:07 EEST
Paul Barton-Davis <pbd_AT_Op.Net> writes:
> This is output from a run with a 4 second track buffer size.
I'm curious about how much work the butler was supposed to be doing on
each request. You show how long it took, but not how full/empty the
buffer queue was at the time of the request. If the "minimum work
threshold" is reasonably small and the butler is keeping the queue
mostly full, it looks like a minimum-work-quantum is probably the
amount that caused the excessive lag.
If everything's working properly, you shouldn't have an increasing
memory consumption during the run of the recorder. Have you watched a
memory-usage trace while you're running? The only thing I can think
of that would cause a 7 second (!) latency for a medium-small IO
request is the machine swapping its brains out. Is there a memory
leak somewhere that could cause this? Does C++ delete actually return
resources to the system, or are you sucking up a monotonically
increasing amount of memory?
Bill Gribble
This archive was generated by hypermail 2b28 : Tue Apr 04 2000 - 20:11:25 EEST