On Sun, 2011-02-27 at 05:09 +0000, Ben Hutchings wrote: [...] > For this limited test, the 486 kernel actually seems to be slightly > faster. Note that this was *not* run on an idle system, so other > activity could affect the measurements a little. > > The Pentium M processors are likely to have different performance > characteristics so I would like to see someone test them as well. > > It might be worth doing some kind of networking benchmark too. I ran some basic tests with netperf, connecting my reasonably fast laptop to the system under test with 1000BASE-T and stopping all other network traffic. I left the firewall rules in place. 3 iterations each of TCP_STREAM and UDP_RR on the 686 flavour: TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 192.168.2.1 (192.168.2.1) port 0 AF_INET : demo Recv Send Send Socket Socket Message Elapsed Size Size Size Time Throughput bytes bytes bytes secs. 10^6bits/sec 87380 16384 16384 60.03 325.73 87380 16384 16384 60.03 326.83 87380 16384 16384 60.02 323.85 UDP REQUEST/RESPONSE TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 192.168.2.1 (192.168.2.1) port 0 AF_INET : demo Local /Remote Socket Size Request Resp. Elapsed Trans. Send Recv Size Size Time Rate bytes Bytes bytes bytes secs. per sec 126976 126976 1 1 60.00 5002.73 114688 114688 126976 126976 1 1 60.00 5024.31 114688 114688 126976 126976 1 1 60.00 5016.31 114688 114688 and again with the 486 flavour: TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 192.168.2.1 (192.168.2.1) port 0 AF_INET : demo Recv Send Send Socket Socket Message Elapsed Size Size Size Time Throughput bytes bytes bytes secs. 10^6bits/sec 87380 16384 16384 60.02 350.12 87380 16384 16384 60.02 349.91 87380 16384 16384 60.02 350.70 UDP REQUEST/RESPONSE TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 192.168.2.1 (192.168.2.1) port 0 AF_INET : demo Local /Remote Socket Size Request Resp. Elapsed Trans. Send Recv Size Size Time Rate bytes Bytes bytes bytes secs. per sec 126976 126976 1 1 60.00 5339.43 114688 114688 126976 126976 1 1 60.00 5389.63 114688 114688 126976 126976 1 1 60.00 5417.86 114688 114688 Again, we see a performance benefit from the 486 flavour. My guess is that the loss of 686 optimisations using e.g. the 'cmov' instruction is outweighed by the removal of SMP locking overhead. SMP-alternatives don't remove all the overhead on UP systems, and in particular the code size will be larger with SMP-alternatives than with SMP disabled altogether. This really ought to be checked on a Pentium M as well, though. Ben. -- Ben Hutchings Once a job is fouled up, anything done to improve it makes it worse.
Attachment:
signature.asc
Description: This is a digitally signed message part