Hallo, guenter geiger hat gesagt: // guenter geiger wrote:
On Tue, 24 Feb 2004, Josh Steiner wrote:
there are reports the the 2.6.x kernels are significantly (like 40% sometimes) faster than the 2.4.x kernels.
I would really like to know when this "sometimes" happens :) Surely you can not map the kernel performance to pd performance, if lets say pd spends 1% of its execution time in kernel, 40% improvement would not make pd noticeable faster.
I have heard that a 2.6 kernel has better realtime behaviour than an unpatched 2.4 kernel, which may help in some cases to lower latency.
I've switched to 2.6 now, and must say 2.6 is very okay. It had problems compared to a LL-patched 2.4 kernel, but a) those problems did'nt affect me and the way I work (I don't do 16-channel, 4 ms latency recordings) and b) from what I've heard, 2.6.3 fixed a lot of these latency problems. Still, for an out-of-the box kernel, 2.6 is impressive and I'd say: totally useable. (Except: I cannot print anymore.)
The new scheduler is wonderful. The system feels faster everywhere, alhough it probably isn't.
Frank Barknecht _ ______footils.org__