On 06/22/2016 12:57 PM, Sergey Fedorov wrote: >> > @@ -1236,6 +1236,9 @@ uintptr_t tcg_qemu_tb_exec(CPUArchState *env, >> > uint8_t *tb_ptr) >> > tcg_abort(); >> > } >> > break; >> > + case INDEX_op_mb: >> > + smp_mb(); > I think we could distinguish smp_rmb(), smp_wmb() and smp_mb() here.
I don't. The extra unpredictable branch is not likely to be worthwhile. If we want anything finer grained for tci, would should revisit the complete rewrite that I did a year to two ago. r~
