Designing and Building Applications for Extreme Scale Systems
University of Wyoming
COSC 5010, 2015 Spring
Tuesday, 11:00, 241 Ross Hall
Professors William Gropp and Craig C. Douglas



Useful links

Link What
HPCWire HPCWire (April 6, 2015) article on parallel loops in compilers


Lecture Title Who
lecture01 Introduction. Course plan. Importance of performance models; Quantifying performance ccd
lecture02 Extreme scale systems. Basic architecture. Simple performance modeling ccd
Benchmarks. Sparse matrix-vector multiply performance nf
Cache memory. Add to performance model. Measuring performance ta
Spatial locality. Transpose as an example. Blocking for memory hierarchy mw
Cache oblvious approach and performance model kp
Aliasing: helping the compiler out jb
Instruction Execution and Pipelining sa
lecture 13 Vectors mb
lecture 14
lecture 15
Moore's Law and Speedup kp
lecture 16 Threads ta
lecture 17 OpenMP Basics ta
lecture 18 OpenMP and MAXLOC ta
lecture 19 OpenMP and General Synchronization jb
lecture 20 Distributed Memory Parallelism mb
lecture 21 Parallel Programming Models for Scientific Computing nf
lecture 22 MPI Basics kp
lecture 23 More on Point-to-Point Communication sa
lecture 24 Buffering and Message Protocols jb
lecture 25 Strategies for Parallelism and Halo Exchanges mb
lecture 26 Performance Models for Distributed Memory Parallel Computing nf
lecture 27 Halo Exchange and Contention nf
lecture 27a MPI Datatypes kp
lecture 28 Processor Topology and MPI ta
SMPCommModel Improving the Performance Model for SMPs jb
lecture 29 Collective Communication and Computation in MPI jb
lecture 30 Considerations When Using Collective Operations mb
lecture 31 Introduction to Parallel I/O ta
lecture 32 Introduction to MPI I/O nf
lecture 33 More on MPI I/O jb
lecture 34 One-sided Communication in MPI mb
lecture 35 More on One Sided Communication sa
lecture 36 MPI, Hybrid Programming, and Shared Memory kp
lecture 37 New Features of MPI-3 jb


Craig C. Douglas

Last modified: