This preview shows pages 1–3. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.View Full Document
Unformatted text preview: Class Overview: 1. Concepts of parallel structures: Dependencies Granularity Locality 2. Performance metrics: Execution Time (latency) Speed Up EfFciency 3. Reading discussion 4. Task versus data parallelism Unlimited / ¡ixed / Scalable Parallelism 5. Working groups discussion Concepts of Parallel Structures: 3 topics on how to formally limit overhead. 1. Dependency: ordering relationship between computations. Examples: • Message passing system: must wait to use data • Protect critical region (count 3s example) DeFne a taxonomy by looking at the ordering of read and write operations: Data Dependencies: • Read after write: ¢ow dependency (“true” dependency) • Write after read: anti dependency (“false” dependency) • Write after write: output dependency (“false” dependency) 01/31/2011 Parallel & Distributed Computing Class Notes • Read after read: no constraint Consider the code sequence: (i) sum = a + 1; (ii) Frst_term = sum * scale1; (iii) sum = b + 1; (iv) second_term = sum * scale2;...
View Full Document
This note was uploaded on 02/18/2012 for the course CIS 4930 taught by Professor Staff during the Spring '08 term at University of South Florida - Tampa.
- Spring '08
- Distributed Computing