Trying out the mtl and baseline benchmarks stated in Alexis King's talk. This is the result I get:
~/g/mtl-bench $ stack bench
mtl-bench> benchmarks
Running 1 benchmarks...
Benchmark mtlbench: RUNNING...
benchmarked MTL Benchmark/Single module
time 992.6 μs (970.5 μs .. 1.023 ms)
0.975 R² (0.933 R² .. 0.994 R²)
mean 1.162 ms (1.105 ms .. 1.231 ms)
std dev 227.0 μs (178.2 μs .. 257.3 μs)
variance introduced by outliers: 90% (severely inflated)
benchmarked MTL Benchmark/Multi module
time 51.21 ms (49.26 ms .. 53.62 ms)
0.993 R² (0.983 R² .. 1.000 R²)
mean 51.71 ms (50.45 ms .. 54.77 ms)
std dev 3.743 ms (1.876 ms .. 6.450 ms)
variance introduced by outliers: 27% (moderately inflated)
benchmarked Base Benchmark/Single module
time 546.4 μs (516.5 μs .. 608.2 μs)
0.910 R² (0.789 R² .. 0.998 R²)
mean 532.7 μs (519.7 μs .. 585.4 μs)
std dev 71.94 μs (15.79 μs .. 160.3 μs)
variance introduced by outliers: 76% (severely inflated)
benchmarked Base Benchmark/Multi module
time 460.2 μs (428.8 μs .. 490.3 μs)
0.890 R² (0.787 R² .. 0.962 R²)
mean 623.8 μs (580.2 μs .. 693.3 μs)
std dev 193.1 μs (151.6 μs .. 233.3 μs)
variance introduced by outliers: 95% (severely inflated)
Benchmark mtlbench: FINISH