What's really surprising is that once you add more cores to the benchmarks, it becomes orders of magnitude slower.
With -N1:
benchmarking connections/100
time 14.78 ms (14.66 ms .. 14.86 ms)
1.000 R² (0.999 R² .. 1.000 R²)
mean 15.11 ms (14.99 ms .. 15.33 ms)
std dev 388.5 μs (236.6 μs .. 611.0 μs)
benchmarking connections/1000
time 146.5 ms (143.9 ms .. 149.7 ms)
1.000 R² (0.999 R² .. 1.000 R²)
mean 150.5 ms (148.9 ms .. 152.0 ms)
std dev 2.242 ms (1.489 ms .. 3.410 ms)
variance introduced by outliers: 12% (moderately inflated)
benchmarking connections/10000
time 1.523 s (1.500 s .. 1.567 s)
1.000 R² (1.000 R² .. 1.000 R²)
mean 1.541 s (1.532 s .. 1.551 s)
std dev 10.84 ms (3.937 ms .. 14.65 ms)
variance introduced by outliers: 19% (moderately inflated)
With -N2:
benchmarking connections/100
time 15.90 ms (8.682 ms .. 29.22 ms)
0.395 R² (0.345 R² .. 0.999 R²)
mean 9.890 ms (8.624 ms .. 14.89 ms)
std dev 6.383 ms (164.5 μs .. 13.09 ms)
variance introduced by outliers: 96% (severely inflated)
benchmarking connections/1000
time 86.76 ms (83.65 ms .. 93.80 ms)
0.995 R² (0.984 R² .. 1.000 R²)
mean 88.68 ms (87.43 ms .. 91.65 ms)
std dev 3.044 ms (1.942 ms .. 4.282 ms)
benchmarking connections/10000
time 860.6 ms (636.3 ms .. 1.077 s)
0.987 R² (0.987 R² .. 1.000 R²)
mean 931.1 ms (890.3 ms .. 971.8 ms)
std dev 52.23 ms (24.33 ms .. 69.72 ms)
variance introduced by outliers: 19% (moderately inflated)
With -N4:
Benchmark bench-connections: RUNNING...
benchmarking connections/100
time 80.88 ms (1.093 ms .. 159.9 ms)
0.263 R² (0.004 R² .. 0.799 R²)
mean 75.93 ms (34.91 ms .. 128.1 ms)
std dev 78.69 ms (56.97 ms .. 108.9 ms)
variance introduced by outliers: 91% (severely inflated)
benchmarking connections/1000
time 864.1 ms (-959.9 ms .. 3.006 s)
0.580 R² (0.000 R² .. 1.000 R²)
mean 893.1 ms (405.7 ms .. 1.064 s)
std dev 326.1 ms (16.56 ms .. 394.2 ms)
variance introduced by outliers: 73% (severely inflated)
With -N8:
benchmarking connections/100
time 1.331 s (1.024 s .. 2.048 s)
0.966 R² (0.947 R² .. 1.000 R²)
mean 1.088 s (1.024 s .. 1.216 s)
std dev 128.0 ms (86.18 μs .. 148.1 ms)
variance introduced by outliers: 23% (moderately inflated)
What's really surprising is that once you add more cores to the benchmarks, it becomes orders of magnitude slower.
With
-N1
:With
-N2
:With
-N4
:With
-N8
: