Glossary Concurrency / Term

[Parallel vs serial] Parallel computations have overlapping lifetimes, and true parallelism means that the overlap is physically instant or simultaneous, while pseudo-parallelism means that the overlap is just conceptual. True parallelism requires hardware support (a multi-core processor), while pseudo-parallelism means that the parallel computations are an abstraction over serially interleaved sub-computations. True parallelism is essentially an implementation detail, while pseudo-parallelism can be part of the computational model.

The colloquial meanings of "concurrent" and "parallel" are largely synonymous, which is a source of significant confusion that extends even to computer science literature, where concurrency may be misleadingly described in terms that imply or explicitly refer to overlapping lifetimes. The crux of the relation between concurrency and parallelism is just that order-independent (concurrent) computations can be parallelized without changing the correctness of the outcome, but concurrency alone doesn't imply parallelism, or vice versa. For example, computations can be parallel without being independent, as in the case of SIMD, which executes parallel computations in lockstep.

Permanent link Overlapping vs interleaved lifetimes (true parallelism vs pseudo-parallelism) - Creation date 2020-09-05