On Wed, 14 Mar 2018, Vitaly Kuznetsov wrote: > Thomas Gleixner <tglx@xxxxxxxxxxxxx> writes: > > On Mon, 12 Mar 2018, Vitaly Kuznetsov wrote: > >> Thomas Gleixner <tglx@xxxxxxxxxxxxx> writes: > >> > On Fri, 9 Mar 2018, Vitaly Kuznetsov wrote: > >> >> Static key is being used to avoid performance penalty for non-Hyper-V > >> >> deployments. Tests show we add around 3 (three) CPU cycles on each > >> >> VMEXIT (1077.5 cycles before, 1080.7 cycles after for the same CPUID > >> >> loop on bare metal). We can probably avoid one test/jmp in vmx_vcpu_run() > >> >> but I don't see a clean way to use static key in assembly. > >> > > >> > STATIC_JUMP_IF_TRUE, STATIC_JUMP_IF_FALSE are your friends. > >> > > >> > >> Thanks for the tip, > >> > >> with a single kernel user of these APIs it was easy to miss :-) > >> > >> Unfortunately, these APIs are only present if HAVE_JUMP_LABEL and > >> (afaiu) we still care about KVM on !HAVE_JUMP_LABEL builds. It would be > >> nice if we can make them behave the same way static_branch_likely() and > >> friends do: compile into something else when !HAVE_JUMP_LABEL so we can > >> avoid nasty #ifdefs in C code. > > > > What's the reason for !jump label builds of a recent kernel? Old compilers? > > > > To be honest I don't see any, we can start depending on HAVE_JUMP_LABEL > for CONFIG_KVM I guess. We currently try to move the minimum compiler version to one which provides jump label support, so this should be a non issue. @Peter: What was the final conclusion of this discussion? Thanks, tglx
