Paolo Bonzini <pbonz...@redhat.com> writes: > On 14/07/2016 14:02, Alex Bennée wrote: >> Looking at the perf data it looks like the hotest part of the code now >> is cpu_get_tb_cpu_state which is required to get the initial hash to >> search for the next tb. > > It should be possible to keep the flags up-to-date in a separate field > of CPUARMState, similar to what x86 does.
Possibly, there is a tension between lazy and always upto date version of the flags. But this is work for another day I think ;-) > > Paolo -- Alex Bennée