SPPU · Final Year Information Technology · 2019 Course
High Performance Computing
414444 · Elective III · Semester VII · 3 credits · In-semester 30 + end-semester 70
One of the Elective III options.
Unit 1: Introduction to Parallel Computing
6 hoursMotivation and scope; Flynn's classification; shared, distributed and hybrid memory; communication costs; interconnection networks and routing; process–processor mapping.
Marks: 15 in the in-semester exam
- Flynn's taxonomy
- Shared and distributed memory systems
- Interconnection networks
- Process-processor mapping
Unit 2: Parallel Algorithm Design
6 hoursTask forming, pipelining and divide-and-conquer; decomposition techniques; task characteristics and interactions; mapping for load balancing; parallel algorithm models; CUDA and OpenACC.
Marks: 15 in the in-semester exam
- Decomposition techniques
- Mapping for load balancing
- Parallel algorithm models
- OpenACC
Unit 3: Basic Communication
6 hoursSynchronous, asynchronous, blocking and non-blocking calls; MPI point-to-point and collective operations: broadcast, reduction, all-reduce, prefix sum, scatter, gather, all-to-all and circular shift; shared-memory synchronisation.
Marks: 17–18 in the end-semester exam
- MPI programming model
- Point-to-point communication
- Non-blocking communication
- Collective communication
Unit 4: Analytical Modelling of Parallel Programs
6 hoursSources of overhead; speedup, efficiency, Amdahl's and Gustafson's laws; granularity; scalability; minimum and cost-optimal execution time; asymptotic analysis.
Marks: 17–18 in the end-semester exam
- Speedup: Amdahl's and Gustafson's laws
- Sources of overhead in parallel programs
- Scalability and cost-optimality
Unit 5: Shared Memory Programming
6 hoursCUDA architecture; kernels, synchronisation, memory contention and device–host transfer; OpenMP programming.
Marks: 17–18 in the end-semester exam
- CUDA kernels, threads, blocks and grids
- CUDA memory management
- OpenMP fork–join model and directives
Unit 6: Parallel Algorithms and Applications
6 hoursMatrix–vector and matrix–matrix multiplication and Cannon's algorithm; Monte Carlo π; parallel bubble and quick sort; all-pairs shortest paths; parallel DFS and best-first search.
Marks: 17–18 in the end-semester exam
- Parallel matrix multiplication and Cannon's algorithm
- Parallel sorting
- Parallel all-pairs shortest paths
- Parallel search: BFS and DFS
Unit contents are summarised from the official SPPU syllabus . AiCanCode.org is not affiliated with Savitribai Phule Pune University.