VECTOR PROCESSORS :
- A processor can operate on an entire vector in one instruction
- Work done automatically in parallel (simultaneously)
- The operand to the instructions are complete vectors instead of one element
- Reduce the fetch and decode bandwidth
- Data parallelism
- Tasks usually consist of:
- Large active data sets
- Poor locality
- Long run times
The three major categories to exploit parallelism:
Instruction-level parallelism (ILP)
Multiple instructions from one instruction stream are executed simultaneously
Thread-level parallelism (TLP)
Multiple instruction streams are executed simultaneously
Vector data parallelism (DP)
The same operation is performed simultaneously on arrays of elements
Vector Length :
- VMIPS has a vector length of 64. But in real world applications vector lengths are not exactly 64. For example, adding just first n elements of a vector.
- Vector Length register is used for this purpose.
- VLR controls the length of any vector operation by defining their length.
- Its value cannot be greater than the length of the vector registers.(64 in this case)
- This works when the length of data is less than the Maximum Vector Length of a processor. But in real world applications, data in vectors in memory can be greater than the MVL of the processor.
- In this case, we use a technique called Strip Mining.