On 03/07/2014 07:11 AM, Peter Maydell wrote: >> > As a bonus, you'll have accurate exceptions should the access throw, so you >> > don't need to force the save of PC before calling the helper. Which... I >> > don't >> > see you doing, so perhaps there's a bug here at the moment. > Mmm. (In system mode we'll save PC as a side effect of having > an accessfn defined for the DC_ZVA reginfo.)
Ah, I see it. I'll note that the GETRA/PC thing can be made to work with any helper. It's just a matter of passing down the outermost helper's retaddr to cpu_restore_state. C.f. target-alpha's dynamic_excp/arith_excp functions. So in theory there's no need for the accessfn to require storing the pc first. r~
