Parallel Computing: Performance

Suppose that the execution of this program on a single core takes $40$ milliseconds. If we have $2$ cores, we could split these $40$ tasks among these two cores, as we explained on the previous slide, which will make the program run approximately $2$ times faster (about $20$ milliseconds). On a 4-core CPU, the program will run close to $4$ times faster (about $10$ milliseconds), etc.

Running all 40 multiplications on one core (core #0).

Running all 40 multiplications on one core. Miriam Briskman, CC BY-NC 4.0.

Splitting the program between 2 cores.

Splitting the program between 2 cores. Miriam Briskman, CC BY-NC 4.0.

Splitting the program between 4 cores.

Splitting the program between 4 cores. Miriam Briskman, CC BY-NC 4.0.