Dumping raw data and interpreting it

[Vlad Ureche]

Hi!

First of all, thanks to the scalameter developers for a cool application that has come in handy on numerous occasions.

I constantly bump into the following scenario: I prepare a scalameter benchmark, I set an aggregator, I set the parameters and then run the benchmark. And then I figure the parameters are not exactly what I wanted...

Wouldn't it be better to have the raw experiment data dumped and a mechanism to load it up and process it with different aggregators, noise, outer elimination etc? I know some of the parameters affect online diecisions, such as re-running a benchmark, but there are also many parameters which don't interfere with the execution and act as post-processing instructions. It would be really nice to have those in a second pass.

WDYT?

Cheers,
Vlad

[Aleksandar Prokopec]

Aggregators are a utility that people thought should be removed. Most people use aggregators to get the history and the mean value. Using min and max aggs is rare, and I also wonder if it is useful.

You could dump the data into a dsv file with the dsvreporter and postprocess it outside scalameter?

--
Sent from my Android phone with K-9 Mail. Please excuse my brevity.

[Aleksandar Prokopec]

Also, I did not understand what kind of postprocessing you would like to do?
Are you looking to aggregate your values differently, or do something else?
Can you give me an example?

Note that noise and outer elimination are measurement methodologies used by ScalaMeter.
ScalaMeter uses outer elimination to internally decide if he should discard some measurements, and then do _additional measurements_.
These affect online decisions.

--
Aleksandar Prokopec
Doctoral Assistant
LAMP, IC, EPFL
http://people.epfl.ch/aleksandar.prokopec

Sent from my Android phone with K-9 Mail. Please excuse my brevity. --
You received this message because you are subscribed to the Google Groups "ScalaMeter" group.
To unsubscribe from this group and stop receiving emails from it, send an email to scalameter+...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.