[linux-audio-dev] Latency test (RTcmix/Linux)

New Message Reply About this list Date view Thread view Subject view Author view Other groups

Subject: [linux-audio-dev] Latency test (RTcmix/Linux)
From: David J. Topper (topper_AT_virginia.edu)
Date: Mon Jul 17 2000 - 02:55:37 EEST


Hey folks,

Yeah ... I shouldn't have gone here (I've got plenty else to do) ... but
I couldn't resist. I'll upload soundfiles and corresponding tabulations
tomorrow in case anyone cares. But here's the final score. I'll
revisit later on I think. The format of the table at the end might get
garbled.

I'd love to find out ways to reduce latency as these results are pretty
bad.

DT

---- <snip> RESULTS.txt ----

This was a pretty simple latency test designed to look at the following:

     general Linux latency issues
     RTcmix latency issues
     latency in GTK when gdk_input_add() used to monitor MIDI device

The test involved recording a stereo audio stream. Channel 1: the
output
of a Korg X5. Channel 2: the output of RTcmix, triggered by that same
device. I tried using rtstools-1.0 in order to change priority.
Priorities
were set to 99.

NAMING CONVENTION: rtcmix_midi_test: simple application (1 process)
that fires off RTcmix commands. pslider: more elaborate GTK mechanism
for
triggering RTcmix. File names are designated by one of these two types,
then:

___frags_buf_sched

where frags are the number of OSS fragments used by RTcmix, buff is the
audio buffer size (in samples) and sched is one of the rtstools types
(FIFO,
RoundRobin, and default ... meaning no manual priority change).

PROBLEMS: RTcmix fires off 4 processes it seems. I'm guessing this is
2
per thread. One for thread control? Another being the actual process?
I
unfortunately only set the priority of 1 thread which showed up under
top as
using noticable CPU cycles. So it could be that all threads should have

scheduling priority changed.

There did not seemt to be any measurable difference between events
triggered
directly (eg., rtcmix_midi_test) or via pslider. So cheers to the
GTK folks.

Overall latency was around 0.04 sec, which is not so good. More tests
will
have to be done.

RESULTS:

Test Frags Buff Sched_type Latency (sec)
---------------------------------------------------------------------
rtcmix_midi_test 10 8 d 0.0583
" " " f 0.0364
" " " r 0.0484
" 3 " d 0.0522
" 3 " f 0.0518
" 3 " r 0.0435
pslider 10 8 d 0.0476
" " " f 0.0508
pslider 10 16 d 0.0376
" " " f 0.0566
rtcmix_midi_test 10 16 d 0.0456
" " " f 0.0365

----- <end snip> RESULTS.txt -----


New Message Reply About this list Date view Thread view Subject view Author view Other groups

This archive was generated by hypermail 2b28 : Mon Jul 17 2000 - 03:22:13 EEST