6.S894: Accelerated Computing

Term: Fall 2024
Instructor: Jonathan Ragan-Kelley
Lecture: Thursdays 2pm-3pm, room 34-303
Lab: Tuesdays 2pm-4pm, room 24-307
TAs: William Brandon, Nikita Lazarev

Welcome!

6.S894 is a new graduate-level course at MIT focused on performance engineering for specialized hardware accelerators. We aim to convey both:

1. First-principles understanding:
   Why do modern accelerator architectures look the way they do?

2. Hands-on experience:
   How do we write fast code for these platforms in practice?

Most of the assignments in this course focus on programming GPUs, for which we use NVIDIA’s CUDA programming language. Students are expected to complete weekly programming assignments, participate in weekly in-person lab sessions, and present a final project at the end of the semester on a topic of their choice.

Announcements

• Lab 5: Matrix Multiply (Improved Scheduling) has been released! (Instructions)

• Lab 5 Checkpoint due Monday, October 7 (Gradescope)

• Lab 5 Final due Tuesday, October 15 (Gradescope)

Site Map

• Calendar – Lecture, lab, and office hour schedule; assignment due dates.

• Labs – Descriptions and instructions for all lab assignments.

• Syllabus – Course policies, topics covered, general course information.

• Resources – Links to notes and external materials relevant to the course.

• Contact – How to get in touch with course staff.

• Piazza – Main hub for student discussion and staff communication.

Acknowledgements

Course infrastructure is made possible by generous support from Jane Street.

Thanks to CS6120 at Cornell for providing the base CSS stylesheet on which this website is built.