The stddev of erubi-rails is so much higher than all other headline benchmarks. As a headline benchmark, I want it to produce reliable results. The duration of each iteration in erubi-rails is so slow that it could be spending non-trivial time in the harness's loop implementation too.
To address that, this PR increases the loop count of erubi-rails so that the duration of each iteration becomes closer to other headline benchmarks. I used railsbench and lobsters to pick a target duration.
The stddev of erubi-rails is much smaller, and the duration of each iteration is shorter too, making the output less verbose and the experience closer to other headline benchmarks.
The stddev of erubi-rails is so much higher than all other headline benchmarks. As a headline benchmark, I want it to produce reliable results. The duration of each iteration in erubi-rails is so slow that it could be spending non-trivial time in the harness's loop implementation too.
To address that, this PR increases the loop count of erubi-rails so that the duration of each iteration becomes closer to other headline benchmarks. I used railsbench and lobsters to pick a target duration.
Before
After
The stddev of erubi-rails is much smaller, and the duration of each iteration is shorter too, making the output less verbose and the experience closer to other headline benchmarks.