benchpark
benchpark copied to clipboard
LLNL
An open collaborative repository for reproducible specifications of HPC benchmarks and cross site benchmarking environments
Currently the csv files that populate the [system](https://software.llnl.gov/benchpark/system-list.html) and [benchmark](https://software.llnl.gov/benchpark/benchmark-list.html) lists in the docs are generated by iterating through the old directory structure. We need to rewrite the generation scripts...
- [ ] System provides @becker33 - [ ] Experiment requires @becker33 - [ ] Concretizer @becker33 - [ ] Add to all systems and experiments @team
- [x] System init, docs. #298 - merging 8/19 - [x] Experiment init, docs. #331 @alecbcs @becker33 - merge before 8/22 - [ ] Re-create all existing LC systems (manual...
We do need to compare the versions of all of the software in the stack (mpi, dependencies). What the compiler and optimization flags? What about the resources and process mapping?...
We need to be able to reason whether the new experiment.py is equivalent to the experiments we previously had written in yaml.
We want to be able to specify strong and weak scaling experiments: - what stays constant, - what gets zipped, - what range of process counts to use (i.e., 2...
`benchpark init` needs to perform `system init` and `experiment init`. Followed by `benchpark setup` command. Later, may want `benchpark setup` to do `benchpark init` at the same time.