What form of parallel processing is used by computers to improve computation speed?

Pipelining is a form of parallel processing that improves computation speed by allowing multiple instruction phases to overlap. In a pipelined architecture, the CPU breaks down the execution pathway of instructions into several stages, such as fetching, decoding, executing, and writing back. Each of these stages can be handled by different components of the CPU simultaneously, so while one instruction is being executed, another one can be decoded, and yet another can be fetched. This effectively keeps the processor constantly busy and reduces the total time to execute a sequence of instructions, thereby enhancing speed and throughput.

In contrast, multitasking tends to refer to the operating system's ability to manage multiple processes or users simultaneously, rather than improving the speed at which a single computation is performed. Concurrent processing, while it involves performing multiple calculations or processes at the same time, does not inherently focus on overlapping different stages of a single processing task in the same efficient manner that pipelining does. Vector processing, on the other hand, enhances the ability to process arrays or vectors of data through single instructions but doesn't specifically describe the pipeline methodology. Hence, pipelining is the most accurate choice for describing the specific method used to increase computation speed through parallel processing.