Messages in this thread | | | Date | Fri, 25 Apr 2014 18:05:53 +0200 | From | Peter Zijlstra <> | Subject | Re: [PATCH v4 5/5] sched: ARM: create a dedicated scheduler topology table |
| |
On Fri, Apr 25, 2014 at 06:04:19PM +0200, Peter Zijlstra wrote: > > The example above is consistent because CPU2 mask and CPU0 mask are > > fully exclusive > > > > so > > CPU0: cpu_corepower_mask=0-1 > > CPU2: cpu_corepower_mask=2 > > are consistent > > > > CPU0: cpu_corepower_mask=0-2 > > CPU2: cpu_corepower_mask=0-2 > > are also consistent > > > > but > > > > CPU0: cpu_corepower_mask=0-1 > > CPU2: cpu_corepower_mask=0-2 > > are not consistent > > > > and your example uses the last configuration > > > > To be more precise, the rule above applies on default SDT definition > > but the flag SD_OVERLAP enables such kind of overlap between group. > > Have you tried it ? > > I've never tried degenerate stuff with SD_OVERLAP, it might horribly > explode -- its not actually meant to work. > > The SD_OVERLAP comes from not fully connected NUMA topologies; suppose > something like: > > 0------1 > | | > | | > 2------3 > > or: > > ( 10 20 20 0 ) > ( 20 10 0 20 ) > ( 20 0 10 20 ) > ( 0 20 20 10 )
d'0h: s/0/30/
0 <-> 3 is 2 hops, too focused on the single hop case
> Your domain level that models the single-hop/20 distance has overlapping > masks: > > N0: 0-2 > N1: 0,1,3 > N2: 0,2,3 > N3: 1-3 > > I've never tried to construct a NUMA topology that would be overlapping > and have redundant bits in.
| |