This is not perfect, because we don't see which machine is which easily, it's also just adding clutter. While we can click on the CPU icon to get the details, it's hidden, and not sure the sorting is consistent everywhere (it might be).
It's also not clear how different environments are handled for the executor comparison and the overview plots.
Before implementing something new/different, I think it would be good to sketch options.
Possible options might be:
stick with multiple rows per benchmark, but make the machines more easily recognizable. Perhaps color coding the CPU icons? or showing the host name?
or a more condensed view, having the environments on the same inline plot, and show the time difference for both environments next to each other. so diff column per environment.
There are probably other options.
Other work arounds to the whole issue could be:
having separate projects for different environments
having different executors to signify different environments
Currently, the compare view shows data from different environments/machines by having the same benchmark multiple times, see for instance: https://rebench.dev/RPySOM/compare/47442d40216b6240aafe9a0a01668db08441b7f9..503a3c9bc7b7a3f0e4fe37cb518e4c289978f1e8
This is not perfect, because we don't see which machine is which easily, it's also just adding clutter. While we can click on the CPU icon to get the details, it's hidden, and not sure the sorting is consistent everywhere (it might be). It's also not clear how different environments are handled for the executor comparison and the overview plots.
Before implementing something new/different, I think it would be good to sketch options.
Possible options might be:
stick with multiple rows per benchmark, but make the machines more easily recognizable. Perhaps color coding the CPU icons? or showing the host name?
or a more condensed view, having the environments on the same inline plot, and show the time difference for both environments next to each other. so diff column per environment.
There are probably other options.
Other work arounds to the whole issue could be: