Goals

• Lab 6: Shared Memory Computing Patterns

  • Choose an appropriate number of worker processors for your compute node & problem

  • Parallelize code using shared memory model & multiple threads, using tools such as:

    • ThreadsX.map and ThreadsX.mapreduce

    • Threads.@threads

    • ThreadsX.foreach

    • FLoops.jl and ThreadedEx (recommended)

    • Threads

    • Folds.jl (alternative)

  • Parallelize code using shared memory model & multiple processes, using tools such as:

    • pmap

    • SharedArrays

    • DistributedArrays.jl w/ using map and mapreduce

    • @distributed for loop

    • FLoops.jl and DistributedEx (recommended)

    • Folds.jl (alternative)

• Readings / Discussions

  • Evaluating the suitability of a problem for different parallel architectures

Project

• Peer Code Review due

Lab

Lab 6: Parallel Programming I: Shared-memory Systems (due Oct 24)

• Exercise 1: Parallelization for Multi-Core Workstations via Multiple-Threads

• Exercise 2: Parallelization for Multi-Core Workstations via Multiple Processes (static html version)

Readings

• Introduction to Parallel Computing Sec B-E -or-

• Introduction to High-Performance Computing Ch 2.1-2.4: Parallel Computing (19pg) -or-

• Introduction to High Performance Computing for Scientists and Engineers:

  • Ch 4 Parallel Computers (20pg) -and-

  • Ch 5 Basics of Parallelization (22pg)

Additional Resources

• Week 8, Monday: Parallelization: Q&A and Scaling Laws