CPSC 418: Superscalar vs Parallel Processing
Eddie Wong writes:
Are super-scalar and parallel processing the same, or is there
a difference? I'm not quite sure what "issuing multiple
instruction simultaneously" means exactly in hardware
terms.
Short answer:
Parallel processing refers to having multiple CPUs
working in parallel and comes in two flavours:
SIMD = single instruction multiple data
MIMD = multiple instruction multiple data
it is generally used for large scientific computations,
although it is beginning to appear in workstations
such as the UltraSparc, which can have up to 4 CPUs
The CPUs can be working on separate programs, or can
be working on different parts of the same program (eg
dividing up a large matrix into smaller matrices and
having each CPU work on one of the smaller matrices.)
A super-scalar CPU can fetch two or more instructions
in one clock cycle. The instructions are typically
consecutive instructions in the same program.
Superscalar is called fine-grain parallelism
Parallel processing is called course parallelism
Long answer:
I'll try to describe these issues in more detail at
the beginning of lecture on Tue 9 Jan.
Last modified: 04 Jan 1996