A collection of benchmarks for Flix.
Some benchmarks are implemented in other languages (C++, Java, Scala, Ruby) for comparison. These implementations are written "naturally." They are written clearly in a style that fits with the language, are not optimized, and do not intentionally mimic the Flix code. Thus, they will use mutability and iteration instead of immutability and recursion, but the overall data structures and algorithms are the same. The Scala implementations are written in a functional style, and are therefore more similar to Flix than Java.
The Flix-only benchmarks are ones that would not make sense for comparisons with other languages, since they rely heavily on the solver and its datastore. (When rewritten for other languages, these benchmarks would probably use arrays, which are very efficient due to cache locality and fewer memory allocations.)
The Strong Update analysis is a real benchmark (that is, not a toy program or contrived example), and was one of the benchmarks used for the PLDI 2016 Flix paper. The inputs for the analysis are kept in a separate, private repository.
The scripts directory contains scripts for running these benchmarks.
Includes:
- fibonacci
- matrixmult (Flix only)
- nbody
- pidigits
- shortestpaths (Flix only)
- strongupdate (Flix only, see above)