Performance

Amdahl’s law describes the maximum speedup you can get by improving or parallelizing only part of a system. It is commonly used in performance engineering to answer questions like: “If I make this function 10× faster, how much faster will the whole program be?” “If I parallelize this workload across 8 CPU cores, what is the best possible speedup?” “Why did adding more machines stop helping?” The key idea is that the parts you do not improve become the limiting factor. ...