|
fgm, I tested your scenario on an Ubuntu 10.04 LTS VM. Here is the base NUMA config:
kangas@orion:~$ numactl --show
|
policy: default
|
preferred node: current
|
physcpubind: 0 1 2 3 4 5 6 7
|
cpubind: 0
|
nodebind: 0
|
membind: 0
|
Single NUMA node, 8 cores.
Now let's run "numactl --interleave=all" to see it has any effect. With exactly one NUMA node to bind against, it should have no material effect.
kangas@orion:~$ numactl --interleave=all -- numactl --show
|
policy: interleave
|
preferred node: 0 (interleave next)
|
interleavemask: 0
|
interleavenode: 0
|
physcpubind: 0 1 2 3 4 5 6 7
|
cpubind: 0
|
nodebind: 0
|
membind: 0
|
"policy" has changed but all other values are precisely the same. "numactl --interleave=all" effectively no-ops in this scenario. So there is no need to complicate the startup script with a different code path that will require separate testing.
|