I put together these pages to help others understand
what is in Andrea's kernels a little better.
http://home.earthlink.net/~rwhron/kernel/2.4.17rc2aa2.html
http://home.earthlink.net/~rwhron/kernel/2.2.20aa1.html
The pages are created from his patch diff logs.
I've stress tested and benchmarked 2.4.17rc2aa2 a lot and
it's been very solid. For a workload that creates a lot
of processes 2.4.17rc2aa2 has a definite edge. This is
easiest to see in a couple unixbench tests:
2.4.17-mjc1 2.4.17rc2aa2 2.5.1-dj10
System Call Overhead 352361.7 362120.5 255809.4 lps
Process Creation 817.5 2037.7 1212.0 lps
Execl Throughput 265.8 458.4 316.5 lps
And lmbench; the highest and lowest results of 3 runs were dropped.
Processor, Processes - times in microseconds - smaller is better
----------------------------------------------------------------
OS fork exec sh
proc proc proc
------------ ---- ---- ----
2.4.17rc2aa2 745 2769 9583
2.5.1-dj10 810 3504 11.K
2.4.17-mjc1 1128 4244 12.K
Good stuff comes from a lot of sources, and I'm hoping some of
the tree maintainers will start cherry picking from Andrea's
tree too. :)
Note: mjc1 above was configured without preempt, rtsched
or lockbreak.
2.4.17rc2aa2 consistently does better at dbench too.
More results are at:
http://home.earthlink.net/~rwhron/kernel/repo.html
-- Randy Hron- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Mon Jan 07 2002 - 21:00:17 EST