No , I don't mean playback from disk, only from RAM,
example: opening the /dev/dsp with a buffer of 4 fragments of 1024 bytes = 23ms
and then only do the following (scheduled RT FIFO max priority)
while(1)
{
gettimeofday(&time1);
write(audio_fd, my_audiobuffer,1024);
gettimeofday(&time2);
if( time2 - time1 > 23ms) then count and report buffer overrun
}
and in background there are scripts which perform heavy disk I/O.
again here is the URL for the bench:
http://www.gardena.net:80/benno/linux/latencytest0.3.tgz
when I get some time (in the next days) I will perform similar tests using a
pure CPU bound task, (this time scheduled RR), and record the timedifferences
between 2 for(i=0;i< bignumber ; i++) so that the for loop takes at least 1ms
, then i will measure the time diffs with the Pentium RDTSC instruction to get
more accurate results.
I think the result will be that this bench does'nt suffer of 130ms latencies,
but I must get the proof for that.
if the new bench will stay into the 20ms limits, then it is a pure locking
problem, if not, then it's the disk which blocks the system.
But my windoze test was the proof that the limit is not the disk.
I will let you know soon !
ciao
Benno.
sbenno@gardena.net
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/