Hi Ingo, I finally have a rebuilt kernel with latency tracing enabled, a jackd with the proper prctl incantations built in and I'm getting some (hopefully) meaningful data.
I'm running a script that samples /etc/latency_trace every second and logs the maximums to a file. First attachment: jackd running 2x128 @48KHz, idle (no clients). Ignore the first trace (leftover from a previous run). After the first three triggered latencies I added a task doing a "tar cf usr.tar /usr" to see what disk i/o load would do. Second attachment: jackd + ams + japa + disk i/o Are these good for anything? [Maybe I need to just enable tracing all the time instead of just tracing inside the jack client loop?] -- Fernando
latencytraces-jack-idle.gz
Description: GNU Zip compressed data
latencytraces-jack-ams-japa.gz
Description: GNU Zip compressed data