Amdahl’s Law governs, the time it takes the program to execute using parallel processors/CPUs, versus using only one serial processor.
Naturally, it follows that:
- That the parallelizable part of the program can be executed faster by applying more processors to executing the program. The more threads/ CPUs the more concurrent work is done, so, the faster it goes.
- The non-parallelizable [single threaded] part can only be executed faster by optimizing the code or by recoding in a way, which can be supported by parallelization.