We look at speed up for this simple case |
Suppose you run a problem on one processor - the sequential version
|
Now run the "same" problem (definition of "same" can be tricky but it is not subtle here) on a parallel machine with P nodes where C.P.U. same as sequential machine Let execution time be TP |
Then speed up is S=T1/TP |
We would like S to be P What is it in our case and why ? |