Performance Engineering of Software Systems
Performance engineering in C: bit hacks, vectorization, multicore parallelism, cache-efficient algorithms, storage allocation and lock-free synchronization.
Modern machines offer huge computing power, but making software use even a fraction of it takes insight into every layer of the platform. This hands-on, project-based course teaches how to build scalable, high-performance software in C, in recorded lectures with notes: performance analysis, Bentley’s rules and bit hacks, vectorization and compiling C to assembly, multicore programming and races, storage allocation, cache-efficient and cache-oblivious algorithms, synchronization without locks, and graph optimization.
What You Will Achieve
- Measure and analyze the performance of C programs and find their bottlenecks
- Apply algorithmic, instruction-level and caching optimizations to speed up code
- Write parallel programs for multicore machines and reason about races and parallelism
- Design cache-efficient and lock-free algorithms and data structures
Program Overview
Single-Core Performance
Making one processor go faster.
- Matrix multiplication, Bentley’s rules and bit hacks
- Architecture, vectorization and C to assembly
- What compilers can and cannot do
- Measurement and timing
Parallelism and Memory
Using many cores and the memory hierarchy well.
- Multicore programming, races and parallelism
- Analysis of multithreaded algorithms and the Cilk runtime
- Storage allocation, serial and parallel
- Caching, cache-efficient and cache-oblivious algorithms
Advanced Techniques
Beyond the basics of performance engineering.
- Nondeterministic programming and synchronization without locks
- Domain-specific languages and autotuning
- Speculative parallelism and a chess-engine code walk
- Graph optimization
Course Completion
Job Outlook
Performance engineering is used in systems and infrastructure software, high-performance and scientific computing, game engines, trading systems and compiler work.
Good for: Programmers who want to learn how to make real software run much faster on modern hardware.
Keep in mind: The assignments rely on AWS and Git features that may not be available outside the class, and the team projects with code reviews were for enrolled students.
Bottom line: A demanding, hands-on course in software performance that goes all the way down to the hardware.
Frequently Asked Questions
What are the projects like?
Team projects with a beta submission, a design and code review, and a final submission that must meet a performance goal, each with a write-up of the bottlenecks found and the speedups gained.
Are there practice quizzes?
Yes. Four practice quizzes are published, and three of them come with solutions, so you can test yourself on the lecture material.
Is the course taught in C?
Yes. C is the course language throughout, with the Cilk runtime for parallel programming, and the homework comes as C code files, headers and makefiles.



There are no reviews yet.