Improving software performance by deploying parallel software on multiple processors often comes at the cost of increasing energy consumption. This paper focuses on such energy-performance tradeoffs. Techniques for computing bounds on software speedup and energy factor that captures the energy cost are presented. Numeric examples for the bounding techniques lead to valuable insights regarding system behaviour, energy and performance.

Additional Metadata
Keywords Amdahl's Law, Average Parallelism, Software Parallelism, Speedup, Speedup Bounds
Persistent URL dx.doi.org/10.1007/978-3-642-15784-4_5
Citation
Majumdar, S. (2010). On the energy-performance tradeoff for parallel applications. doi:10.1007/978-3-642-15784-4_5