A superscaler
processor is one that accepts more than one
instruction into the
pipeline per
clock cycle.
The
Pentium 4 processor accepts up to 6 instructions per clock cycle, while emitting the results of 3 instructions per clock cycle. Why it emits fewer than it gets per clock and still performs well is left as an
exercise for the reader.
This is different from a
VLIW processor. Very Long Instruction Word processors accept multiple instructions per cycle, but they must be formed in groups, and certian 'slots' of the very long instruction can only perform certian operations - so a series of add instructions may only use two of, say, the four execution slots available per long instruction. This processor design relies on
compiler intelligence to
parallelize code, whereas a superscaler processor parallelizes the code
dynamically, requiring no special compilation.