I've got a graph of worker polution on 2 cpus now too (this one was done
with the correct priorities too).

http://plugin.org.uk/spmp/worker-2p.png

I think we need to see larger worker sizes but with smaller amounts
polluted, I think that this one is pretty meaningless with large ammounts
of pollution, as no app could mangle that much data in realtime even if it
wanted to.

So, in this one

http://plugin.org.uk/spmp/worker-2p-big.png

each processing thing polutes the ammount of ram shown on the x axis, but
allocates (and pollutes before the counting is started) 100 times that
much. At 15k polluted data per cycle with 15meg of total data its still
pretty reasonable.

- Steve

Reply via email to