But on many OSs, you can lock a process to a processor. I know this is true
under HPUX and AIX at least under certain conditions it does become useful.
Say you have at least 2 processes. One of them is doing work to feed another
and then another. If they are each tied to a different processor, they act
better in a pipeline. They can each do their little bit in the pipeline of
work.
Example:
find /usr/local -xdev | cpio -oc | compress -9c | dd of=/dev/rmt0 bs=32768
For small amounts data it won't matter. But I have seen this go much quicker
when the processes were tied to different processors. It shaved time from 2
hours to 1 hour 45 minutes on AIX in one case. Of course this system was doing many
other things at time.
It is off by default on AIX and HPUX. It would be interesting to see if
similiar things are ever true under Linux versus saying that the kernel will
always know better. Of course I say this and I run mostly UP on linux.
Aaron
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/