multiprocessor

A multiprocessor is a computer system that contains two or more central processing units (CPUs) capable of executing instructions concurrently. In such systems, CPUs may share memory and I/O resources, enabling parallel execution of tasks and improved throughput.

Architectural styles include tightly coupled shared-memory systems, often called symmetric multiprocessing (SMP), where all processors access a common address space and memory is centrally or uniformly accessible. In non-uniform memory access (NUMA) designs, memory is distributed among processors, and access latency differs by location. Loosely coupled configurations connect multiple independent processors or nodes via a network (clustered or distributed systems) and rely on message passing. Modern multiprocessors may also be multi-core chips, where several processing units reside on a single integrated circuit but are treated as separate processors by the system.

Key concerns include cache coherence, which ensures that cached copies of shared data remain consistent; memory bandwidth; interconnect topology; and operating system support for symmetric multiprocessing or message-passing paradigms. Scheduling, synchronization, and memory models are central to achieving scalable performance on multiprocessors.

Benefits include higher aggregate compute power and improved throughput for parallel workloads such as servers, databases, and scientific computing. Drawbacks include programming complexity and the potential for contention, synchronization overhead, and diminishing returns beyond certain processor counts; power, cooling, and cost considerations also rise with scale.

History and impact: Early multiprocessors appeared in the 1960s and 1970s, with subsequent generations expanding shared-memory and NUMA architectures. Today, multiprocessor systems are common in servers, workstations, and high-performance computing, reflecting a trend toward parallelism to exploit multicore and multi-processor capabilities.