Department of Computer Science, Unit Catalogue 2009/10 |
CM30225: Parallel computing |
 Credits:
| 6 |
| Level:
| Honours |
| Period: | Semester 1 |
| Assessment:
| CW25EX75 |
| Supplementary Assessment: | Supplementary assessment information not currently available (this will be added shortly) |
| Requisites:
| Before taking this unit you must take CM20214 or equivalent. |
| Description: | Aims: To give students the ability to recognise and understand the problems and opportunities afforded by parallel systems; to recognise the differing types of parallelism available and make advised choices between them; and to take advantage of progress in technology as modern computers become ever more parallel. Learning Outcomes: Students will be able to: 1. write and debug simple parallel programs; 2. recognise the issues surrounding concurrent access to data; 3. Describe the various kinds of parallel hardware, parallel programming methodologies and the relationship between them. Skills: IT (T, A). Content: 1. Basics: supercomputers and the consequences of Moore's Law; bandwidth vs latency; speedup, efficiency, scalability; Amdahl's & Gustafson's Laws; Flynn's taxonomy, SPMD; distributed, shared, NUMA and other memory architectures. 2. Shared memory computing: multicore systems (cache coherence and bottlenecks); mutual exclusion and critical regions; low level constructs including POSIX threads and synchronisation methods such as barriers, locks, semaphores, etc.; language-level support including monitors, OpenMP; vector and array (SIMD), HPF, Cn. 3. Distributed computing: clusters, message passing, MPI. Programming using MPI (PBS). 4. Parallel algorithms and data structures. 5. Topics in Parallel Computing: examples might include HPC; MapReduce; distributed file systems; the Grid; GPGPU and OpenCL; instruction level parallelism (SWAR,VLIW). |