mozilla / firefox-hardware-report

The Firefox Hardware Report was the precursor to the Firefox Public Data Report, to which it now redirects
https://data.firefox.com/dashboard/hardware
27 stars 13 forks source link

Could the Hardware Report also include logical cores (i.e. hyperthreads)? #60

Open lukewagner opened 6 years ago

lukewagner commented 6 years ago

The Hardware Report currently reports the number of physical processors (which is a great start), but it would be really helpful when tuning for parallelism to know the distribution of logical cores (i.e., hyperthreads). In particular, one big question I have is what % of our 2-physical-core users have hyperthreads (so 4 logical cores).

fbertsch commented 6 years ago

This is absolutely doable - we already have this info in the environment section of main pings, and it's also available in the Longitudinal table.

A generic logical cores section would not specifically answer the question you posed, however - it would only give you the distribution of number of logical cores; i.e. what % of users have 2 logical cores, what % have 4, etc.

lukewagner commented 6 years ago

That's a good point. I think the most ideal data would be, for each physical core count bucket, what % of that bucket had hyperthreading enabled. Thus, the labels would be something like "% of 2-core with HT". The most interesting datum for me is what % of 2-physical-cores systems have hyperthreading, given that the Core 2 arch did not have HT.

fbertsch commented 6 years ago

That makes sense, but the downside is it doesn't conform with the rest of the figures in the report. We meet Friday as a team, and we'll discuss it there.

@lukewagner in the meantime, would it be helpful to have some ad-hoc analysis for this?

lukewagner commented 6 years ago

Thanks for putting it on the agenda!

If I could get a snapshot of the logical/physical distribution at the moment, yes, that'd be lovely.

fbertsch commented 6 years ago

@lukewagner this should help: https://sql.telemetry.mozilla.org/queries/47219/source. Note that is based on a 1% sample, so multiply counts by 100 to get more accurate numbers - the fractions will be fine though.

lukewagner commented 6 years ago

Thanks, that's so useful!

jeffgca commented 6 years ago

bump, this is pretty useful, for example in bugs like this one.

openjck commented 6 years ago

This is not something we already have in the dataset, is it @fbertsch?

openjck commented 6 years ago

I'll open a corresponding issue for the new site.

openjck commented 6 years ago

Moved to:

https://github.com/mozilla/Fx_Usage_Report/issues/63