UNIX Weapons School: Difference between revisions

Lecture slides are available [http://nick-black.com/lectures.html here]. These slides are very much a work in progress; LaTeX source is available on [https://github.com/dankamongmen/unix-weapons-school GitHub].

''update from 2021: this class would now absolutely go into [[Io_uring|io_uring]] and [[eBPF]]. also SSDs and the resultant changes to the I/O paradigm. there would probably be some talk of Spectre/Meltdown in conjunction with the architecture lecture, also BIG.little and core scheduling. probably a bit more coverage of [[ARM]] details, though i'll not yet yeet the x86 stuff. the "future of systems programming section" would be updated. a bit more talk about distributed systems, probably, and i'd discard the current material about machine learning. other than that, this still pretty much stands.''

 

 

* "[http://nick-black.com/comparch.pdf Your Friend the Computer]"

 

* "[http://nick-black.com/x86.pdf The x86 is dead. Long live the x86!]"

 

* "[http://nick-black.com/practicum.pdf Not Sucking in the UNIX Environment]"

 

<blockquote>"The TOPGUN course is designed to train already experienced Navy and Marine Corps aircrews at the graduate level in all aspects of strike-fighter aircraft employment, including tactics, hardware, techniques and the current world threat for air-to-air and air-to-ground missions. The course includes eighty hours of lectures and twenty-five flights that pit students against TOPGUN instructors. When a pilot or WSO completes the TOPGUN course he/she will return as a Training Officer carrying the latest tactical doctrine back to their operational squadron, or go directly to an FRS squadron to teach new aircrews."</blockquote>

Without systems programmers, hardware sits silent, stillborn. Without systems programmers, application designers shout mutely into a void.

The reconfigurability of software combined with the system bounding of hardware allows the systems programmer the broadest space with which to innovate. Unfortunately, most of these possible innovations are just bugs, or (within the innermost ring) superbugs. When that signed vs unsigned comparison means the wrong hard drive sector is selected, it can ruin your whole day. Systems programming unites the most fundamental realms of our science: [[Programming Language Theory|programming language theory]], [[Compiler Design|compiler design]], [[architecture]], and algorithms come together in practical application. These theories are complex enough to demand classes of their own. Here, we will integrate these prerequisite theories to deliver efficient, robust systems software solutions. Furthermore, you'll become acquainted with the gritty details of cutting-edge technique.

[https://www.youtube.com/watch?v=tHxf17yJsKs Because I am hard, you will not like me. But the more you hate me, the more you will learn.]

Each project will be evaluated for both correctness (in terms of unit tests passed) and performance (on several reference platforms). Performance of incorrect code is, of course, moot. I will make each project public two weeks prior to its due date, and provide a set of basic unit tests and a web-accessible oracular reference implementation (you can submit inputs and receive outputs). For details of grading, see the first week's slides. I am using a scheme designed to maximize competition and conflict between class members, which I hope will reduce cheating, lead to great personal exertions, and possibly get a chair thrown.

* David Kanter of [http://realworldtech.com Real World Tech] for use of his superb processor diagrams, and his years of insightful writing besides.

** ...and through them, of course, to the United States Navy's [http://en.wikipedia.org/wiki/United_States_Navy_Fighter_Weapons_School Fighter Weapons School] aka "Top Gun".