Sample Comparison #56302

New Issue

Francesco Siddi · 2018-08-11T01:19:46+02:00

Francesco Siddi commented

2018-08-11 01:19:46 +02:00

When viewing a sample (for example after submitting it) we would like to be able to compare it to similar samples. Ideally:

Best result
Slighly better result
Our sample
Slightly worse result
Worst result

When viewing a sample (for example after submitting it) we would like to be able to compare it to similar samples. Ideally: * Best result * Slighly better result * Our sample * Slightly worse result * Worst result

Francesco Siddi commented

2018-08-11 01:19:46 +02:00

Added subscriber: @fsiddi

Francesco Siddi commented

2018-09-09 20:48:08 +02:00

Added subscriber: @dr.sybren

Francesco Siddi commented

2018-09-09 20:48:08 +02:00

Addressed feedback on the implementation in the wip-compare-results. Can we have another round of feedback and merge this @dr.sybren?

Regarding the question:

Do you have any idea to extend the query to "the N benchmarks slower/faster than yours"?

We could do the following. For each scene in groupValues:

query for the fasted and slowest, near fastest and near slowest results (near values can be an 80% near interpolation from current to extreme - the query should return a device name)
each result per scene is properly regrouped and pushed to the chart

This probably does not make much sense, I can elaborate later.

Addressed feedback on the implementation in the [wip-compare-results](https://developer.blender.org/source/blender-open-data/history/wip-compare-results/). Can we have another round of feedback and merge this @dr.sybren? Regarding the question: > Do you have any idea to extend the query to "the N benchmarks slower/faster than yours"? We could do the following. For each scene in groupValues: - query for the fasted and slowest, near fastest and near slowest results (near values can be an 80% near interpolation from current to extreme - the query should return a device name) - each result per scene is properly regrouped and pushed to the chart This probably does not make much sense, I can elaborate later.

Francesco Siddi commented

2018-10-21 19:50:24 +02:00

This is now partially implemented (when viewing a result, we can compare with the average).