You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Reduce the numbers of iterations in benchmarkshttp/check_is_http_token #15218
Currently benchmark/http/check_is_http_token takes several hours to complete if we try to compare two different Node versions. I think we should lower the number of iterations to simplify benchmarking the whole of http.
Specifically, I propose to revert this commit 39fdf07, or to reduce the number to something in between.
This was part of #6570.
To be honest, what the benchmarks in general could use is either some way to have specific n values for different mixes of inputs (would require some extra logic in each benchmark file) or we just switch to some sort of time-based mechanism for these synchronous benchmarks (would require some coordination with the benchmark runner).
Activity
mscdex commentedon Sep 6, 2017
To be honest, what the benchmarks in general could use is either some way to have specific
nvalues for different mixes of inputs (would require some extra logic in each benchmark file) or we just switch to some sort of time-based mechanism for these synchronous benchmarks (would require some coordination with the benchmark runner).bnoordhuis commentedon Sep 21, 2017
@mcollina Ping. Do you want to follow up?
mcollina commentedon Sep 21, 2017
I'm open to discuss this, but I do not see any actionable comment so far. What do we want to do?
BridgeAR commentedon Sep 21, 2017
This changed recently again in #14002. Closing this therefore.