On 2/21/20 12:32 PM, Stefan Hajnoczi wrote: > Hi Andrzej, > After having looked at the PMRWBM part of the spec, I think that the > Bit 1 mode should be implemented for slightly better performance. Bit > 0 mode is not well-suited to virtualization for the reasons I > mentioned in the previous email. > > The spec describes Bit 1 mode as "The completion of a read to the > PMRSTS register shall ensure that all prior writes to the Persistent > Memory Region have completed and are persistent". > > Stefan
Make sense -- will incorporate that feedback in second version of patch.