M8: Advanced Parallel Scheduling Techniques (Part 2)

In this eighth module, we will begin to look at more ways to dynamically distribute work to threads using barriers and data parallelism.

Pre-recorded Lectures

The pre-recorded lectures are available here: M8 Videos. You can also find the videos under the “Panopto” tab on the MPCS 52060 canvas site.

The lectures are a series of approx 5-20 minute videos divided into 5 sections:

  • 8.1 - Barrier Motivating Example

  • 8.2 - Parallel Prefix Sum

  • 8.3 - Barrier Implementation

The slides/code presented in these videos are accessible in the upstream repository under the m8/resources directory. Please reachout to me if you do not have access to this repository.

Resources/Readings

  • The Art of Multiprocessor Programming readings:

    • Ch. 17.1, 17.2, 17.3

Synchronous Session (In-person)

As a reminder here are the dates and times for the synchronous session for this module:

  • Dates/Times

    • Section: Tuesday November 15th @ 12:30pm-2:20pm

  • Discussion Outline

    • Q&A on the pre-recorded lectures

    • Revisit Work-Stealing algorithm

    • Work Distribution: Map-Reduce

Assignment

Assignments are always out Wednesday evenings.