MIT 6.004: Computation Structures | Hacker News Reader

MIT 6.004: Computation Structures

6004.mit.edu

275 pointsskovorodkin19 comments

Somewhat related: looks like Prof. Harry Lewis is also teaching a class on "Classic CS" during the Spring 2020 term at MIT.

"This subject examines papers every computer scientist should have read, with an emphasis on the period from the 1930s to the 1980s. It is meant to be a synthesizing experience for advanced students in computer science: a way for them to see the field as a whole, not through a survey, but by reliving the experience of its creation, relating the original work to the field as it exists today. The aim is to create a unified view of the field by replaying its entire evolution at an accelerated rate, giving students the opportunity to become sophisticated generalists"

https://www.eecs.mit.edu/academics-admissions/academic-infor...

That sounds pretty awesome, does anyone happen to have a list of the papers the class covers, I can't seem to find it online.

It's probably similar to this one https://canvas.harvard.edu/courses/34992/assignments/syllabu...

Do you have the course schedule or content? I can't seem to find any past content/editions of the course. The course sounds amazing!

wow. setup and hold times have been demoted to 'demystification only' and not something that is part of the formal curriculum. when I helped teach this course in the 90's, it was a major section and if you could not answer basic questions about synchronous clock discipline, you could not get an A. It was as important as stack-based calling conventions.

I mean, I guess most 'computer science' folks today can have a fecund and profitable career and have never heard of these concepts, but... I hope some people still wonder, why do we have clock speeds, and what other alternatives might exist?

It was there till 2017. The change to the present form happened in 2018, when Arvind became a co instructor

Source I have been following this class online since 2015.

The previous version, with full labs, autograded tutorial problems and exercises is still available here:

https://computationstructures.org/lectures/info/info.html

Also, an archived version of that course is still running on edx ...

https://www.edx.org/course/computation-structures-part-1-dig...

.. where there is a forum available, so people can still ask questions about how to build a 32-bit CPU from scratch using MOSFETs :)

It's also on OCW,

https://ocw.mit.edu/courses/electrical-engineering-and-compu...

This seems very comprehensive and beautifully comprehensible.

It reminds me closely of The Elements of Computing Systems and its companion web-based incarnation Nand2Tetris available at https://www.nand2tetris.org/

It seems MIT changes the url:

https://6004.mit.edu/web/fall19FA19/resources/lectures

Thank you! I really needed this.

Last time I looked at MIT 6.004 the lecture slides seemed like they were useless without the speaker but there were no videos so I moved on.

Looks like there's been an update since so will take another look...

I started trying to read through the slides in the link and I felt the same way - this looks like fascinating stuff that I wish I understood better, but I got lost trying to just read the slides without a lecturer going over them. There were a few textbooks listed (including Hennesy & Patterson's excellent "Computer Organization"); I might check out the other two instead.

I'm finding nand2tetris much better for covering what I gather is mostly the same.

The 6.004 for 09 has more of the physics than nand2tetris but I can live without that

I kind of like their Minispec HDL - I don't think I've seen it before though I guess it derives from Bluespec. I have always liked Wirth's Lola as well.

Verilog and VHDL are serviceable, but I think there is an advantage to having a simple, friendly syntax without the verbosity and overhead of VHDL.

I also liked how Wirth's course involved running on FPGA hardware. It looks like you might be able to do that in the MIT course as well although I didn't see specific labs for it.

I was hoping to learn about GPU and neural network accelerator architectures. This seems like a conspicuous absence in the course.

This is an undergrad course, "200 level" in the standard course numbering (which MIT does not follow). It's for people who know how to program, but relatively new to the field. It's mostly going to be getting people used to assembly and basic computational building blocks.

You might be interested in the new class: 6.812 Hardware Architecture for Deep Learning.

Indeed. Thank you!

Is the site just temporarily down? I was watching one of the lecture videos and when I came back to the site I got a 502 error.

I get 502 error

Reply on news.ycombinator.com