On Fri, Sep 10, 2004 at 12:42:24AM -0400, Lee Revell wrote:
On Thu, 2004-09-09 at 23:29, Eric Dantan Rzewnicki
wrote:
Eh, this is how free software works, the concept of wasting people's
time doesn't mean a lot. I spent a month banging my head against the
emu10k1 ALSA driver trying to figure out why the latency was 10 times
worse than in windows. Over a period of weeks I posted 5 or 6 crazy
theories to alsa-devel, all wrong. Turns out it was just a stupid bug
where a bunch of instances of SND_PCM_PERIOD_SIZE needed to be changed
to SND_PCM_PERIOD_BYTES. I submitted a patch and now the latency is
better than on Windows.
Moral of the story: in free software, the only person whose time you can
waste is your own, and it's only wasted time if you don't learn
anything.
Thanks for the pep talk. :) I did learn a good bit over the past week.
And I'm in a good place to build a lot of different sound tools for
myself now.
This is very, very important, and also very good news!
There have not
been a lot of reports yet from real linux audio users. So far UP users
seem to be reporting very good results, while SMP/HT users are still
seeing weird behavior.
Can you please try the following so I can report the results to Ingo:
Run "echo 0 > /proc/sys/kernel/preempt_max_latency"
Do some audio work for while
Copy /proc/latency_trace to a file and send it to me via private mail
Also, once your system seems to be running well, you should disable
latency tracing with "echo 0 > /proc/sys/kernel/trace_enabled". The
latency tracer itself can generates latencies of a few ms when it
updates /proc/latency_trace; this is not reported in the traces.
Thanks for these notes. I'll get the latency trace to you this weekend.
Thanks,
Eric Rz.