This benchmark suite is used as a reference for BQN development. It has precise within-language benchmarks and a slow and sometimes unreliable cross-language interface to BQN, Dyalog APL, and J.
There are benchmarks for most primitives that have significant performance considerations, mostly focused on rank-1 numeric arrays at the moment.
BQN result graphs with commentary here.
To use:
$ ./benchmark.bqn # Benchmark and plot everything (slow)
$ ./benchmark.bqn help # See options
$ ./benchmark.bqn list # See functions that can be benchmarked
$ ./benchmark.bqn arith # Record and plot arithmetic benchmarks
Strength reduction tests (very incomplete) test primitives against equivalent expressions to find cases where alternate implementations would be faster.
$ ./testreduct.bqn # Run all tests
$ ./testreduct.bqn -h # Options
$ ls reductions # Test specification files
