We need to adapt the workflow profile_emissions.Rmd to show the noise added to the co2e* columns.
The most expensive step is likely profile_emissions_impl(). It would be nice to cache this intermediate result so we can iterate cheaply on the downstream step that adds noise, so we can experiment with different amount of jitter until we approach a satisfactory noise for each (most?) benchmarks.
Also there are a handful of details added to the workflows used to release tiltOutput that have not yet been incorporated here.
summarize_noise()tiltAddCO2#8We need to adapt the workflow profile_emissions.Rmd to show the noise added to the
co2e*columns.The most expensive step is likely
profile_emissions_impl(). It would be nice to cache this intermediate result so we can iterate cheaply on the downstream step that adds noise, so we can experiment with differentamountof jitter until we approach a satisfactory noise for each (most?) benchmarks.Also there are a handful of details added to the workflows used to release tiltOutput that have not yet been incorporated here.