CS 4121/5121 Introduction to Compilers Practicum Spring 2018

2018 CS 4121/5121 Compiler Bakeoff

The “Best Compiler” title gives members of the winning team both bragging rights and an attractive plaque announcing their excellence in the compilation domain. PA7 submissions were treated as submissions to the contest; however, performance in the bakeoff did not affect course grades.

Criteria

Compilers are compared using the following criteria, listed roughly in order of importance:

• Correctness of generated code and of compile-time error checking.
• Speed of the generated code across a wide selection of benchmarks, including object-oriented benchmarks.
• Length and clarity of the compiler implementation.

Benchmarks

Compilers are evaluated using several Xi and oXi programs, testing both language features and performance. These benchmarks test the performance of objects, arrays, loops, recursion, redundant computations, and more.

Contest results

We had a remarkable number of excellent compiler projects this year, so picking the best one was a real challenge. The winning compiler project was that of Drew Dunne, Jacob Glueck, Alex Libman, and Daniel Wisner. Across our set of benchmarks, their compiler compiled the most programs correctly and the code it generated was 19% faster, on average, than the next fastest compiler. They also implemented some tricky language extensions, including first-class functions and safe downcasting. Perhaps surprisingly, their compiler, written in Java, contained less code than those of several other groups.

Honorable mentions go to three other groups who did an outstanding job of writing compilers that worked well and generated good code:

• Ariel Fong, Danny Qiu, Chesley Tan, and Evan Zhao
• Varun Gandhi, Daniel Glus, Harrison Goldstein, and Devin Lehmacher
• Stephen Buttolph, David Chu, Kevin Wang, and Daniel Weber