Performance issue in 2.5.32+
Quoted 2 lines
lmbench shows major changes in context switching latencies on a dual celeron system
Some lmbench tests vary a lot so I run it 25 times and average the results. Some variations are only apparent on particular cpus (cache size differences, smp, etc).
The lmbench averaging script I use is:
http://home.earthlink.net/~rwhron/kernel/lmbsum
lmbench on a quad xeon and a bit on how to use lmbsum:
http://home.earthlink.net/~rwhron/kernel/bigbox.html#lmbench
Recent uniprocessor results:
http://home.earthlink.net/~rwhron/kernel/latest.html#lmbench
Signature · 2 lines
Randy Hron http://home.earthlink.net/~rwhron/kernel/bigbox.html