On Sun, 2 Jan 2005 12:46:38 -0000 (WET), Rui Nuno Capela <rncbc@xxxxxxxxx> wrote: > Hi, > Matthew Allen wrote: > > using 2.6.10-rc3-mm with Ingo's latest Real-Time Preempt on a pentium > > M 1.6ghz laptop with a gig of ram. The test was done as a normal user > > using the realtime-lsm patch and I have Rui's rtirq script up and > > running. One of the three tests was completed while compiling gnuplot > > in the background :). > > > > Please always try with latest JACK CVS, where some additional statistics > are available [0.99.41+]. Read on. Ok so I now have the latest jack: $ jackd -V jackd version 0.99.43 tmpdir /tmp protocol 15 $ uname -r 2.6.10-rc3-mm1-V0.7.33-04 patched with Ingo's Real-time preempt. CONFIG_PREEMPT_RT=y CONFIG_PREEMPT=y CONFIG_PREEMPT_SOFTIRQS=y CONFIG_PREEMPT_HARDIRQS=y CONFIG_PREEMPT_BKL=y and as far as I know its on: $ cat /proc/sys/kernel/kernel_preemption 1 The machine definetly hangs after about 18 clients (a hard lock that requires cycling power). With 14 clients and everything else the same (actually I set my playbackrate to 48k) I get this. Which looks pretty poor. I am testing with an i8x0 in a 1.6ghz centrino. The alsa-driver is 1.0.8_rc1 *********** CONSOLIDATED RESULTS ************ Total seconds ran . . . . . . : 600 Number of clients . . . . . . : 14 Ports per client . . . . . . : 4 Frames per buffer . . . . . . : 64 ********************************************* Summary Result Count . . . . : 2 / 2 ********************************************* Timeout Rate . . . . . . . . :( 0.0)/hour XRUN Rate . . . . . . . . . . : 138.0 /hour Delay Rate (>spare time) . . : 1515072.0 /hour Delay Rate (>1000 usecs) . . : 0.0 /hour Delay Maximum . . . . . . . . : 47335 usecs Cycle Maximum . . . . . . . . : 966 usecs Average DSP Load. . . . . . . : 60.0 % Average CPU System Load . . . : 39.5 % Average CPU User Load . . . . : 15.6 % Average CPU Nice Load . . . . : 0.0 % Average CPU I/O Wait Load . . : 0.5 % Average CPU IRQ Load . . . . : 0.0 % Average CPU Soft-IRQ Load . . : 0.0 % Average Interrupt Rate . . . : 1819.7 /sec Average Context-Switch Rate . : 92310.4 /sec ********************************************* jacking up the frames to 128 gives me a little better performance, but I really must be missing something. *********** CONSOLIDATED RESULTS ************ Total seconds ran . . . . . . : 600 Number of clients . . . . . . : 14 Ports per client . . . . . . : 4 Frames per buffer . . . . . . : 128 ********************************************* Summary Result Count . . . . : 2 / 2 ********************************************* Timeout Rate . . . . . . . . :( 0.0)/hour XRUN Rate . . . . . . . . . . : 24.0 /hour Delay Rate (>spare time) . . : 761487.0 /hour Delay Rate (>1000 usecs) . . : 0.0 /hour Delay Maximum . . . . . . . . : 45056 usecs Cycle Maximum . . . . . . . . : 1888 usecs Average DSP Load. . . . . . . : 73.8 % Average CPU System Load . . . : 40.5 % Average CPU User Load . . . . : 22.9 % Average CPU Nice Load . . . . : 0.0 % Average CPU I/O Wait Load . . : 1.5 % Average CPU IRQ Load . . . . : 0.0 % Average CPU Soft-IRQ Load . . : 0.0 % Average Interrupt Rate . . . : 2416.9 /sec Average Context-Switch Rate . : 94092.5 /sec *********************************************