Browse Teach Curate

Amdahl's Law and Parallel Speedup

The principle that the maximum speedup from parallelizing a computation is limited by the fraction of work that remains inherently sequential: if a fraction s of the work is serial, speedup on N processors is bounded by 1/(s + (1-s)/N). It frames the trade-off between adding parallel resources and the diminishing returns imposed by serial sections and coordination overhead.

jeremy Video

Amdahl's Law and the Sequential Fraction Ceiling on Speedup in Parallel Computing

Amdahl's Law is a theorem of parallel computing performance analysis that bounds the speedup obtainable from parallelization by the fraction of a computation that must execute sequentially. Formally,…