CS 61CL/61C. Great Ideas of Computer Architecture (Formerly Machine Structure)
Current Schedule (Spring 2016)
Starting this semester, Fall 2010, we are reinventing CS61C, starting with a blank page as to what makes sense to teach of computer architecture and hardware in 2010 that will give a solid foundation on the topic on which to build that should last for decades.
Rather the be something of a catch-all, as in the past, the goal is to learn the great ideas of computer design and implementation:
- Memory Hierarchy (e.g., Caches)
- Thread Level Parallelism (e.g., Multicore)
- Data Level Parallelism (e.g., MapReduce and Graphical Processing Units or GPUs)
- Instruction Level Parallelism (e.g., Pipelining)
- The Transistor and its rate of change (e.g., Moore's Law)
- Quantitative Evaluation (e.g., GFLOPS, Clocks Per Instruction or CPI)
- Layering of Hardware Levels of Abstraction (e.g., AND gates, Arithmetic Logic Unit or ALU, Central Processing Units or CPU)
- Compilation vs. Interpretation (e.g., C compiler, Java interpreter)
- Hardware Instruction Set Interpretation (e.g., instructions as binary numbers)
The idea is to go over the big ideas at a higher level in the first two-thirds of the course, and then go back in more depth in the last third of the course.
We use a running example through the whole course to illustrate the ideas, and will be the basis of a programming contest in the last third of the course to see who can make the fastest version running on the latest multicore hardware.
We use the C programming language and MIPS assembly language to demonstrate these great ideas. The course closely follows the Patterson and Hennessy textbook supplemented by material on the C programming language. A sample week-by-week outline follows.
- 1 - Introduction - Mobile Client vs Cloud Server
- 2, 3 - C programming language vs MIPS assembly language
- 4 - Computer Components and Compilation vs Interpretation
- 5 - Quantitative Evaluation
- 6 - Memory Hierarchy
- 7 - Thread Level Parallelism
- 8 - Data Level Parallelism
- 9 - Transistors and Logic
- 10 - Layers of HW Abstraction
- 11 - Instruction Level Parallelism
- 12 - In More Depth: Cache associativity, Cache coherence, Locks
- 13 - In More Depth: Illusion of machine to yourself - Virtual Memory, Virtual Machines
- 14 - In More Depth: Dependability via Redundancy - Error Correcting Codes, Redundant Array of Inexpensive Disks
- 15 - Contest Results and Conclusion
Coursework will involve weekly two-hour in-laboratory exercises designed to learn the big ideas by doing experiments.