Performance Engineering of Software Systems (MIT 6.172)

reading · we frame, you read MIT or the canonical taught · we author, no canonical fits ↺ spirals back to earlier lessons

Course locked

Complete Compilers and Interpreters first.

This course unlocks once you've finished its prerequisite. Open prerequisite →

Prerequisites

Introduction to Algorithms and Complexity not yet

Systems Programming (C and Zig) not yet

Digital Logic and Computer Organization (Nand2Tetris + Theory) not yet

Processes and Concurrency not yet

Complete prerequisites first to enroll.

reading · 30 min + ~276 min canonical

The Performance Engineering Mindset: Measure, Don't Guess

Performance engineering is the one area of software where your intuition is most confidently, most consistently wrong — and the entire discipline begins with accepting that. You w…

primary MIT 6.172 Performance Engineering of Software Systems (YouTube)

preview →

reading · 30 min + ~276 min canonical L1

Bentley Rules and Bit Hacks: Optimizing the Work

Once measurement (lesson 1) has pointed you at a genuine bottleneck, the most fundamental optimization is also the most obvious once stated: do less work. Not "do the same work fa…

primary MIT 6.172 Performance Engineering of Software Systems (YouTube)

after L1 locked

reading · 35 min + ~276 min canonical L1

Programming the Machine: Caches, Pipelines, and What Compilers Can't Do

Here is the fact that reorganizes everything you thought you knew about performance: on modern hardware, the cost of an operation is dominated not by the operation itself but by w…

primary MIT 6.172 Performance Engineering of Software Systems (YouTube)

after L2 locked

reading · 35 min + ~276 min canonical L1

Cache-Efficient and Cache-Oblivious Algorithms

The previous lesson delivered the uncomfortable truth: memory access dominates performance, and cache misses cost hundreds of cycles. This lesson delivers the empowering consequen…

primary MIT 6.172 Performance Engineering of Software Systems (YouTube)

after L3 locked

reading · 35 min + ~276 min canonical L1

Parallel Performance in Practice: Cilk, Work-Stealing, and Scalable Memory

You have multiple cores; the question is how to actually use them well — and "well" is harder than it looks, because parallel speedup is easy to leave on the table and easy to acc…

primary MIT 6.172 Performance Engineering of Software Systems (YouTube)

after L4 locked

exercise reading · 15 min L1

Exercises: Easy (Warm-up)

- [ ] Bentley rules: identify optimization opportunities in 5 C functions using work reduction, logic reduction, and loop optimization - [ ] Bit hacks: implement population count,…

after L5 locked

exercise reading · 30 min L1

Exercises: Medium (Core)

- [ ] Cilk parallelism: convert a sequential matrix multiply to parallel using cilkspawn/cilksync, measure speedup on multicore - [ ] Work-span analysis: compute work T1, span Tin…

after L6 locked

exercise reading · 45 min L1

Exercises: Hard (Stretch)

- [ ] Cache-oblivious algorithm: implement a cache-oblivious matrix transpose and compare performance to naive and cache-aware versions across different cache sizes - [ ] Lock-fre…

after L7 locked

exercise reading · 60 min L1

Mini-Project

- [ ] Performance optimization challenge — Take a provided slow C program (e.g., N-body simulation), profile it, and optimize it to run at least 10x faster using techniques from t…

after L8 locked

L10

project reading · 600 min L1

Project: Performance Optimization Lab

Take a provided C program that performs matrix operations and is deliberately unoptimized. Profile it using perf/cachegrind, then apply Bentley rules, cache blocking, and Cilk par…

after L9 locked

L11

project reading · 480 min L1

Project: Dynamic Memory Allocator

Implement a high-performance storage allocator in C with malloc, free, and realloc. Use a segregated free list with coalescing. Benchmark against glibc malloc on several workloads…

after L10 locked

L12

assessment reading · 63 min L1

Self-Assessment: Performance Engineering of Software Systems (MIT 6.172)

- [ ] What are Bentley's rules for optimizing work? Give 3 examples of each category (work, logic, loop). - [ ] What is work-span analysis? Define work (T1), span (Tinf), and para…

after L11 locked

L13

assessment reading · 60 min L1

Quarter 11 Gate Check: Software Engineering, Compilers, and Performance Engineering

- [ ] Course 11.1 knowledge check passed - [ ] Course 11.2 knowledge check passed - [ ] Course 11.3 knowledge check passed - [ ] Course 11.4 knowledge check passed - [ ] Lox Inter…

after L12 locked

13 lessons. Read in order; spiral back when you need to. By the end you'll have used the core ideas twice — once on the abstract, once on something you'll meet at work next week.