Buy commercial curl support. We
help you work out your issues, debug your libcurl applications, use the API,
port to new platforms, add new features and more. With a team lead by the
curl founder Daniel himself.
Re: First baby steps at performance testing
- Contemporary messages sorted: [ by date ] [ by thread ] [ by subject ] [ by author ] [ by messages with attachments ]
From: Daniel Stenberg via curl-library <curl-library_at_lists.haxx.se>
Date: Tue, 11 Aug 2026 23:23:26 +0200 (CEST)
On Tue, 11 Aug 2026, Patrick Schlangen wrote:
>> interval and appears on https://curl.se/perf/
>
> What’s the x axis?
Yeah, I need to add some explanations! :-)
The X axis are simply different builds. For a while now at least I run them
pretty frequently so that I get more data to plot and to learn how big
variation we see even without any real code changes.
And to figure out how to best visualize these things.
> The run number? Could there be multiple runs for the same build?
For the same commit, yes. I do them every 20 minutes right now, and it'll do
the same dance wether the same commit or not. This way it might of course also
*not* to a build for a specific commit if more than one commit has been done
within that 20 minute window.
In the data plots, the oldest runs are on the left and then later runs are
added on the right, and they show the commit hash when there is a new commit -
so you can see there how many runs there are per commit.
> To correlate changes in the plot with commits, would it make sense to also
> show the commit hash on the x axis legend?
The embeded commit hashes show this.
> Also, maybe aggregate multiple runs/measurements for the same commit hash
> into the same x axis ‘bucket' and show them as average + markers for error /
> standard deviation (or p50 and markers for p3/p97)?
That's a cool idea. I need to experiment a little with that!
Thanks!
Date: Tue, 11 Aug 2026 23:23:26 +0200 (CEST)
On Tue, 11 Aug 2026, Patrick Schlangen wrote:
>> interval and appears on https://curl.se/perf/
>
> What’s the x axis?
Yeah, I need to add some explanations! :-)
The X axis are simply different builds. For a while now at least I run them
pretty frequently so that I get more data to plot and to learn how big
variation we see even without any real code changes.
And to figure out how to best visualize these things.
> The run number? Could there be multiple runs for the same build?
For the same commit, yes. I do them every 20 minutes right now, and it'll do
the same dance wether the same commit or not. This way it might of course also
*not* to a build for a specific commit if more than one commit has been done
within that 20 minute window.
In the data plots, the oldest runs are on the left and then later runs are
added on the right, and they show the commit hash when there is a new commit -
so you can see there how many runs there are per commit.
> To correlate changes in the plot with commits, would it make sense to also
> show the commit hash on the x axis legend?
The embeded commit hashes show this.
> Also, maybe aggregate multiple runs/measurements for the same commit hash
> into the same x axis ‘bucket' and show them as average + markers for error /
> standard deviation (or p50 and markers for p3/p97)?
That's a cool idea. I need to experiment a little with that!
Thanks!
-- / daniel.haxx.se || https://rock-solid.curl.dev
-- Unsubscribe: https://lists.haxx.se/mailman/listinfo/curl-library Etiquette: https://curl.se/mail/etiquette.htmlReceived on 2026-08-11