Understanding Amdahl's Law

Share

Summary

An explanation of how serial and parallel components dictate the performance limits of parallelized computer programs.

Understanding Amdahl's Law

Highlights

Serial and Parallel Fractions

Amdahl's Law demonstrates that the **serial parts** of a program restrict potential speedup during parallelization. The law defines programs as consisting of two distinct segments: **completely serial sections (S)** and **completely parallelizable sections (P)**. On a single core, the total execution time (T₁) is calculated as the sum of these fractions, expressed by the formula **T₁ = S + P**, where **S** represents the **serial fraction** and **P** represents the **parallel fraction**.

Recently Summarized Articles

Loading...
Original text

Amdahl's Law shows that a program's serial parts limit the potential speedup from parallelizing code. Amdahl conceptualizes programs as consisting of two parts: the completely serial sections (S) and the completely parallelizable sections (P). When the program executes on a single core, its execution time (T₁) will be the sum of the time spent executing the serial fraction (S) and the time spent executing the parallel fraction (P), so T₁ = S + P.