version control for results (and analysis?) directories?

[tinaok]

we have several version of results based on different computational methods co-exisiting in the results directory.
How can we better organise this?

[tinaok]

For example we now have 3 different computational results co-exisiting in the benchmarking/results directory

Cheyenne/2019-05-22 (before the update of 22thAug=version 0.1)
hal24/* (before the update of 7th Sept=version 0.2)
Cheyenne/2019-09* (after the update of 7th Sept-version 0.3)
hal1H (after the update of 7th Sept=version 0.3)

(&I am also creating hal1D with my new update=version 0.4, and I have in my mind to make another series of weak scaling analysis =version 0.5)

What is the best way to organise this so that we do not get lost, and can make correct analysis?

[andersy005]

What is the best way to organise this so that we do not get lost, and can make correct analysis?

My suggestion is to tag stable versions of the repository and running the benchmarks with the stable releases/tags instead of master branch.

[tinaok]

summary of the discussion with @kmpaul in SC19
we propose that

1. separate benchmarking code and result. results should stay in this repository, and source codes and running environment should go in a new repository.
2. sourcode-enviroment repository should have a version numbers