----- On Jul 28, 2021, at 2:32 PM, Linus Torvalds torvalds@xxxxxxxxxxxxxxxxxxxx wrote: > On Wed, Jul 28, 2021 at 11:12 AM Mathieu Desnoyers > <mathieu.desnoyers@xxxxxxxxxxxx> wrote: >> >> Are the percpu atomics you have in mind different from what is found in >> Documentation/core-api/this_cpu_ops.rst ? >> >> Namely this_cpu_add_return(pcp, val) in this case. > > Nope. > > Those are only "CPU-atomic", ie atomic wrt interrupts etc. > > The RCU code wants SMP-atomic, and it's mainly that we *could* do the > addressing more efficiently. OK, so combining the addressing tricks of this_cpu operations with smp-atomic operations (e.g. LOCK prefix on x86). It may indeed become worthwhile given enough users, and fast enough atomic operations. Thanks, Mathieu -- Mathieu Desnoyers EfficiOS Inc. http://www.efficios.com
