Hybrid consistency, a consistency condition for shared memory multipro
cessors, attempts to capture the guarantees provided by contemporary h
igh-performance architectures. It combines the expressiveness of stron
g consistency conditions (e.g., sequential consistency, linearizabilit
y) and the efficiency of weak consistency conditions (e.g., pipelined
RAM, causal memory). Memory access operations are classified as either
strong or weak. A global ordering of strong operations at different p
rocesses is guaranteed, but there is very little guarantee on the orde
ring of weak operations at different processes, except for what is imp
lied by their interleaving with the strong operations. A formal and pr
ecise definition of this condition is given and an algorithm for provi
ding hybrid consistency on distributed memory machines is presented. T
he response time of the algorithm is proved to be within a constant mu
ltiplicative factor of the (theoretical) optimal time bounds.