2.6.0-test6 scheduler goodness
A message that is not in the archive (sent off-list or not yet received): [email protected]
Quoted 4 lines · David B Harris
4) In either 2.6.0-test5, or 2.6.0-test6 (I'm using 2.6.0-test6, I skipped test5), responsiveness was magically fixed for my workload case. I still have lower throughput, apparently (big compiles and whatnot take about 20% longer), but I recently got a CPU upgrade so I don't
Everything CPU bound should run a few percent slower on 2.6 because it uses HZ=1000. You could recompile with HZ=100 and see if that fixes that (just change HZ to 100 in asm/param.h)
20% sounds a big high just for HZ degradation though.
-Andi