Describe the difference between superscalar and VLIW architectures.

Question

What will be an ideal response?

Answer 1

A superscalar processor achieved parallelism by internal means (i.e., within the processor). Instructions are placed in a window/queue and a batch of instructions executed in parallel as long as all operands are ready and there are sufficient free execution units. A VLIW uses the compiler to put a bundle of, typically, three operations into a single instruction slot. The VLIW processor does not have to check data dependencies because they have been checked by the compiler. In that sense a VLIW processor is easier to design than a superscalar processor. If a VLIW processor cannot find instructions that can be processed in the same time slot, the bundled instruction word must fill the empty slots with no?operations, NOPs.

Another difference between superscalar and VLIW is that superscalar is a matter of organization and not architecture. Any processor can be converted into superscalar form without the source code and instruction set changing. A VLIW processor has its own ISA because the instructions are bundles of individual operations.

Because VLIW processors have long instructions, they tend to be used in high performance roles. The most successful VLIW processor is the Itanium, which is found largely in workstations and servers.