I wrote:
> I have high hopes! I'm compiling 2.4.0-test8 right now... will
> report back on how that checks out.
Wow. *No* overruns on any of the tests. A drastic improvement of the
2.4.0-test6 patch! The X11 stress test looks a bit worse, but still
no overruns; meanwhile the disk tests have improved dramatically,
especially the copy and read tests. This is tested with
latencytest's CPU usage set to 76% on a Celeron 333MHz. I've got the
IDE drives set to UltraDMA (hdparm -d1 -X66) which seems to improve
things a bit over plain old DMA - but on 2.4.0-test6 I was still
getting bad overruns.
X11 stress:
fragment latency = 1.451247 ms
cpu latency = 1.102948 ms
4.0ms ( 0)|
1MS num_time_samples=24820 num_times_within_1ms=24813
factor=99.971797
2MS num_time_samples=24820 num_times_within_2ms=24815
factor=99.979855
Proc stress:
3.1ms ( 0)|
1MS num_time_samples=20925 num_times_within_1ms=20821
factor=99.502987
2MS num_time_samples=20925 num_times_within_2ms=20925
factor=100.000000
Disk write stress:
3.8ms ( 0)|
1MS num_time_samples=31573 num_times_within_1ms=31488
factor=99.730783
2MS num_time_samples=31573 num_times_within_2ms=31572
factor=99.996833
Disk copy stress:
3.4ms ( 0)|
1MS num_time_samples=69305 num_times_within_1ms=69126
factor=99.741721
2MS num_time_samples=69305 num_times_within_2ms=69305
factor=100.000000
Disk read stress:
3.1ms ( 0)|
1MS num_time_samples=34419 num_times_within_1ms=34348
factor=99.793719
2MS num_time_samples=34419 num_times_within_2ms=34419
factor=100.000000
--
................. paul winkler ..................
slinkP arts: music, sound, illustration, design, etc.
web page: http://www.slinkp.com
A member of ARMS: http://www.reacharms.com
