Navigation3 reaches API stability this week with the 1.0.0-beta01 release! There's a few important bug fixes in this release as well, but this beta means we're only a few weeks away from rc01 and then stable!

If you run into any issues, please let us know ASAP
developer.android.com/jetpack/andr...

I mostly agree with what your observations are except - Frame data is not dependent either. Some frames are (especially the ones that are slow). There are some experiments on how you can minimize the effects of these interrelated frames and once you do you can still bootstrap.

Here are the references you need.

youtu.be/Iq9DzN6mvYA?... is the actual talk that inspired some of the math.

The actual code lives in: cs.android.com/androidx/pla...

I also tells you that ! There is a fun paper I read that I used as inspiration when I wrote the implementation for computing P values and its confidence interval.

No, I meant rather than use bins with the actual values of the benchmark - you can use % offsets from the min value and then bin them - helps remove noise from the distributions when you are trying to look for patterns. Especially relevant when you see fluctuations in values (noise).

Absolutely. I should have a mode where you are just comparing values.

I should have better onboarding. You can totally drag > 1 file for comparing. I also compute P values for statistical significance.

You can also normalize the distributions before you compare them.

This is what the actual data looks like: cs.android.com/androidx/pla...

This is amazing. I will test this as well and report back.

Saw the same video and it resonated with me too. Learning from mistakes and the journey is so important

Nice title 😁. Looking forward to your talk.

So many GitHub stars 🤩. I would just keep it around for bragging rights 😄

Porkbun is pretty great.

All I am hearing is an excuse to buy a new mouse 😃

This is incredibly exciting ! Looking forward to your talk 🙏

Just wanted to congratulate you folks on some of the delightful animations. Also thank you for Trainline 🙂 another beautiful app.

I should really make this tool easier to discover, and use.

Being able to compare perfetto traces is fantastic. I have been working on writing something to make this part simpler as well. Hopefully will have something to share soon.

I agree. We wrote plot-benchmarks (cs.android.com/androidx/pla...) to make it simpler to read these distributions. This also makes it easy to compare benchmarks before / after normalization between runs; and also provides a very valuable metric of statistical significance.