Lesson 1: Welcome & Overview

• discussion thread
• tasks due January 8

On readings: Recommended background readings are marked with (^) above. Optional historical readings are marked with (*).

Notes

• Welcome to ECE 4/599! I’m really excited to dive deep into some research with you all this quarter!

• Why are memory systems important, why do we care?

  • Simply: store data
  • But why is it challenging? Ultimately, physical constraints
  • We can have large memories or fast memories, but not both. This means if we have a lot of data we need fast access to, we need to use careful engineering.
  • Today’s workloads (e.g., AI) are memory hungry. They use a lot of data and need to crunch it quickly.

• How do we manage with disparate memory performance profiles?

  • Hierarchies
  • How to move between?
  • How much should be automatic?
  • How can we speculate?
  • What is the correct granularity of access and transfer?

• Course overview and the syllabus.

  • Communication will be on Discord and on GitHub.
  • Class sessions will usually follow this structure:
    • First part:
      • I’ll do some traditional “lecturing” in class
      • Some (but not all) sessions will include lessons that may or may not have associated tasks for you to complete.
    • Second part:
      • Everyone should read the relevant paper beforehand so we can have a good discussion. You’ll need to post a comment in the paper’s discussion thread on GitHub.
      • The paper leader’s job is to present the paper using slides, guide the discussion, answer discussion questions on GitHub, and organize the lab to work on a collective blog post.
      • The scribe takes notes from in-class discussions.
      • The blogger summarizes the paper (using input from the paper leader’s notes, online and in-class discussions, and their own interpretations) into a technical blog post.
      • The reviewer summarizes strengths and weaknesses of the paper.
      • Remember to keep track of the schedule!
    • Hacking sessions: in some cases I will give you time during class to hack on your project with me around to help.
  • The course project.
  • Policies, academic integrity, generative AI, respect, accessibility.

• This lesson’s task is about getting set up for the course.

  • Ask questions about the course setup in this lesson’s discussion topic!
  • Your first task is below: introduce yourself in Discord right now, and (offline) pick a paper discussion to lead.
    • This website and the blog, including the GitHub repository where you’ll send PRs. The idea is that we are working together in this class as a real research team might.

Tasks

• Read the syllabus!
• Sign up for Discord and introduce yourself
• Do the background reading for Wednesday and for next week.
• Decide on teams for projects. See the syllabus.
• Ask any questions about the course structure or content in this lesson’s discussion topic.
• Pick a paper from the schedule whose discussion you want to lead and volunteer for it. Claim it by opening a pull request that modifies content.toml to fill in your name on one of the leader = "TK" lines.