Hi Vladimir, On Thu, Jul 22, 2021 at 04:22:26PM +0100, Vladimir Murzin wrote: > On 7/22/21 10:50 AM, Will Deacon wrote: > > As an aside: I'm more and more inclined to rip out the CnP stuff given > > that it doesn't appear to being any benefits, but does have some clear > > downsides. Perhaps something for next week. > > Can you please clarify what do you mean by "it doesn't appear to being any > benefits"? IIRC, Cortex-A65 implements CnP hint and I've heard that some > payloads seen improvement... Has anybody taped that out? I'd have thought building an SMT design in 2021 is a reasonably courageous thing to do. The issue I'm getting at is that modern cores seem to advertise CnP even if they ignore it internally, maybe because of some big/little worries? That would be fine if it didn't introduce complexity and overhead to the kernel, but it does and therefore I think we should rip it out (or at least stick it behind a "default n" config option if there are some niche users). There are also open questions as to exactly what CnP does because the architecture is not clear at all (for example TTBRx_EL1.CnP is permitted to be cached in a TLB). CHeers, Will _______________________________________________ kvmarm mailing list kvmarm@xxxxxxxxxxxxxxxxxxxxx https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
