Efficiency of Vector Processing over Scalar Processing:
We know that, a sequential computer processes scalar operands one at a time. Thus, if we have to process a vector of length n through the sequential computer then the vector have to be broken into n scalar steps and implemented one by one.
For example, take the following vector addition:
A + B → C
The vectors are of length 500. This operation through the sequential computer can be specified by 500 add instructions as given below:
C[1] = A[1] + B[1]
C[2] = A[2] + B[2]
....................................
C[500] = A[500] + B[500]
If we execute the similar operation through a pipelined-vector computer then it does not split the vectors in 500 add statements. Because a vector processor having the set of vector instructions that allow the operations to be specified in one vector instruction as:
A (1:500) + B (1:500) → C (1:500)
Every vector operation may be split internally in scalar operations but they are implemented in parallel which results in mush quicker implementation as compared to sequential computer.
Therefore, the benefit of adopting vector processing over scalar processing is that it eliminates the overhead cause by the loop control need in a sequential computer.