On Sat, Nov 17, 2012 at 4:35 PM, Alex Shi [off-list ref] wrote:

Just find imbalance issue on the patchset.

I write a one line program:
int main ()
{
        int i;
        for (i=0; i< 1; )
                __asm__ __volatile__ ("nop");
}
it was compiled with name pl and start it on my 2 socket * 4 cores *
HT NUMA machine:
the cpu domain top like this:
domain 0: span 4,12 level SIBLING
  groups: 4 (cpu_power = 589) 12 (cpu_power = 589)
  domain 1: span 0,2,4,6,8,10,12,14 level MC
   groups: 4,12 (cpu_power = 1178) 6,14 (cpu_power = 1178) 0,8
(cpu_power = 1178) 2,10 (cpu_power = 1178)
   domain 2: span 0,2,4,6,8,10,12,14 level CPU
    groups: 0,2,4,6,8,10,12,14 (cpu_power = 4712)
    domain 3: span 0-15 level NUMA
     groups: 0,2,4,6,8,10,12,14 (cpu_power = 4712) 1,3,5,7,9,11,13,15
(cpu_power = 4712)

$for ((i=0; i< I; i++)); do ./pl & done
when I = 2, they are running on cpu 0,12
I = 4, they are running on cpu 0,9,12,14
I = 8, they are running on cpu 0,4,9,10,11,12,13,14

Ops, it was tested on latest V15 tip/master tree, head is
a7b7a8ad4476bb641c8455a4e0d7d0fd3eb86f90
not on this series.
Sorry.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>