On Fri, Mar 2, 2018 at 2:35 AM, Peter Maydell <peter.mayd...@linaro.org> wrote: > On 2 March 2018 at 10:29, Peter Maydell <peter.mayd...@linaro.org> wrote: >> On 2 March 2018 at 04:34, Alistair Francis <alistai...@gmail.com> wrote: >>> target/arm: Report the number of cores in the cluster >>> >>> Previously we assumed that we only has a single cluster, which meant we >>> could get away with reporting smp_cpus to the guest. There are cases >>> where we have two clusters (Xilinx's ZynqMP is a good example) so >>> reporting the number of smp_cpus is incorrect. Instead count the cores >>> in the cluster. > >> This seems a bit ad-hoc (it will give the wrong answer if we have >> two clusters both of the same kind of CPU, for instance). >> Maybe it would be better to have a "cluster-size" property on >> the CPU (with the default being number of cores in whole system) ? > > ...though I think "cluster size" isn't quite the right name -- > if you have a big.LITTLE cluster with 2xA57 and 2xA53, there are > 4 cores in the cluster but I think reading the TRM that the cores > will report 2 in the L2CTLR.
Hmmm... I would have thought cluster size is correct enough. The only other option I can think of is "processor size" or "core size"? Alistair > > thanks > -- PMM
