A provably time-efficient parallel implementation of full speculation