On Sep 30, 2021, at 5:44 AM, Barry Song [off-list ref] wrote:

quoted

We have done some performance evaluation with the locktorture module
as well as with several benchmarks from the will-it-scale repo.
The following locktorture results are from an Oracle X5-4 server
(four Intel Xeon E7-8895 v3 @ 2.60GHz sockets with 18 hyperthreaded
cores each). Each number represents an average (over 25 runs) of the
total number of ops (x10^7) reported at the end of each run. The 
standard deviation is also reported in (), and in general is about 3%
from the average. The 'stock' kernel is v5.12.0,

I assume x5-4 server has the crossbar topology and its numa diameter is
1hop, and all tests were done on this kind of symmetrical topology. Am
I right? 

   ┌─┐                 ┌─┐
   │ ├─────────────────┤ │
   └─┤1               1└┬┘
     │  1           1   │
     │    1       1     │
     │      1   1       │
     │        1         │
     │      1   1       │
     │     1      1     │
     │   1         1    │
    ┌┼┐1             1  ├─┐
    │┼┼─────────────────┤ │
    └─┘                 └─┘


what if the hardware is using the ring topology and other topologies

For better or worse, CNA is pretty much agnostic to the physical topology.
So it is true that we might do better by transferring NUMA node preference
to a “closer” node, but it would probably require some sophisticated logic 
to identify such a node. As discussed on another thread, we are opting
for a simple solution that might be refined later if needed.

Best regards,
— Alex


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel