The testing of very large RAMs is difficult. Simple write-read sequences in each cell will not generally catch pattern errors and are useless in detecting neighborhood faults. Deterministic algorithms are available for the more general test, but they tend to be time-consuming for large arrays. Random pattern tests are often more rapid but are suspect.
The authors examine Markovian sequence testing and conclude that by using suitable algorithms such tests can be at least as reliable as deterministic tests and are faster for large memories. They present evidence to justify their position and refer to an experimental paper based on real RAM experience; this paper, however, is not yet available for examination. Thus, although the analysis presented is sound, the validity of the method must await practical trial.