Re: [PATCH 2/3] x86/topology: Fix AMD core count

From: Sherry Hurwitz
Date: Tue Mar 22 2016 - 04:11:46 EST

On 03/21/2016 08:57 AM, Borislav Petkov wrote:
+- AMD: the number of cores in a processor. On a system where cores are
+clustered in groups of 2, 4 or more in compute units, this variable
+denotes the number of*compute units* on the node.
+In both cases, the number of scheduling threads is computed by doing:
+ x86_max_cores * smp_num_siblings
+This means:
+* smp_num_siblings: the number of siblings in a core. On AMD with
+compute units, this number is the number of compute unit siblings,
+i.e., compute unit cores in a single compute unit, according to their
-- hr
Boris, this documentation will help tremendously. In just this line of code:

nr_local_cpus = nr_cores * nr_siblings

we have an Intel HT = AMD core = logical_cpu
and core = AMD compute unit.

Anybody surprised there was a bug?

The surprising thing is that running lscpu on a 32 core 2 socket 6300 Opteron system without
the patches I get the same output as with the patches and it matches Ray's. I don't see an impact.