Latest iOS, iPad Air 2 performance is bad

[ghost]

For phetsims/qa#264. Performance isn't very good. See video.

https://drive.google.com/open?id=1hV6cgK10tV_uVH7XnzRJNGIyCbDjusGB

[ariel-phet]

@samreid while this sim is less likely to be used on iPads considering its more advanced nature, iPad Air performance being poor is disconcerting, so I would say this issue warrants attention. We do not want to get in a performance rathole, but I am marking high priority to indicate it is worth a couple hours of investigation to see if the source of the performance bottleneck can at least be identified.

[arouinfar]

@lmulhall-phet which device were you testing on? I spent quite a bit of time testing on the iPad Air (can't remember the device name, but I think it was the the one with the sticker that warns not to update iOS), for #315 (comment) and it was running smoothly. @samreid is your iPad Air running the latest iOS? Looks like I was testing on iOS 11.4.

[KatieWoe]

@lmulhall-phet which device were you testing on?

@arouinfar Leibniz

[samreid]

I've been testing on iPad Air 2 "Dalton" which is running iOS 11.4.1. I'm a bit confused because I thought iPad Air 1 was our lowest platform (and that's what I thought I was using) but was surprised to learn I have iPad Air 2 instead. @KatieWoe or @arouinfar or @ariel-phet can you help clarify?

When I go use dev.66 Interference Screen, Sound + Particles, I'm getting 45FPS on iPad Air 2 "Dalton" 11.4.1.

According to the PhET Asset Inventory, Leibniz is running iOS 9.3.5. However in the issue title, @lmulhall-phet said the test was run on "Latest iOS". @KatieWoe or @lmulhall-phet can you please clarify which is correct?

[KatieWoe]

Leibniz is running iOS 12, it was upgraded when iOS 12 came out. The simplified inventory reflects this, and I'll go in and change the full inventory. We don't have any iPad Air 1's that I am aware of. Our lowest device is Tycho, an iOS 9 iPad 2. Beyond that, it is an iPad 4 with iOS 10.

[arouinfar]

I'm a bit surprised that the iOS 11.4 performance is so much better than the iOS 12 performance. @KatieWoe is Leibniz our only iOS 12 device?

[KatieWoe]

We have one other. Give me a minute to check it.

[KatieWoe]

It seems ok on our other device. I'm going to go troubleshoot Leibniz

[arouinfar]

Thanks @KatieWoe!

[KatieWoe]

From slack:

Kathryn Woessner [9:06 AM]
Even after restarting Leibniz I'm still seeing the same behavior. I'll keep working at it, and there's a new update so I'll see if that does anything. I only saw the performance issue with sound on the first two screens.

Amy Rouinfar [9:17 AM]
So strange that the two iOS 12 devices are behaving differently
Are they different iPad models?

Kathryn Woessner [9:19 AM]
Yes. Leibniz is iPad Air 2. The other is iPad (2018)
Updating, restarting, and clearing Safari history hasn't cleared it up, but it may be a bit better than when Liam posted the video.

Further info: Confirmed that iPad Air 2 iOS 11 device in office did not show issue. Issue still present in iPad Air 2 iOS 12, but not in iPad (2018) iOS 12 device. After methods it seems like the behavior might be a bit better, but is still present. See video.
https://drive.google.com/file/d/1zjiypySCLIr_S5nrVQWRj1O2It0OR8aT/view?usp=sharing

[samreid]

Does the screen capture slow down the frame rate? What is the reported FPS with ?profiler (while not recording)?

[KatieWoe]

Not much if it does. That's fairly representative. With profiler it usually reads 59 or 60 fps, but when playing sound on Waves or Interference it drops to about 5 fps, or even 3 if using particles. Checking the other iOS 12 device it drops to about 14 fps.

[KatieWoe]

I decided to do a bit more profiler testing and saw that on my laptop there was no fps drop and on the iOS 11 it drops to anywhere from the high 30s to the low 50s, or more of a drop if on interference screen. I think it may be an iOS 12 problem that our second device is better at handling.

[samreid]

@ariel-phet and @arouinfar do you want me to check out an iOS 12 iPad Air for investigation before RC.1?

[ariel-phet]

@samreid my suggestion would be to look at this on campus on Thursday during when design time meeting would be (since you are not on the agenda). See what you can find at that time on campus and lets discuss. We can also plan to finish dev meeting early, so perhaps you and @jonathanolson can look together.

[KatieWoe]

May not be connected, but also saw some variant of #319. If you have the continuous wave going with Both particles and switch to pulse and fire a pulse, end of the continuous wave disappears too soon.

[samreid]

From slack:

Sam Reid [3:00 PM]
Katie and Liam pointed out 3 issues:

1. clipping problem on iPad
2. smeared particle problem on iPad
3. variant of Wave clears after some time #319 where the wave clears

I’ll take a quick look at each one to see if I can correct it. But to clarify, which ones of those are acceptable for 1.0?
Also, what is the soonest time that you would start testing a Wave Interference RC? That will give me a window of time for investigation. (edited)

Kathryn Woessner [3:02 PM]
Maybe Monday if the dev for efac goes well today. But Fractions also has its RC

Sam Reid [3:03 PM]
OK please keep me in the loop so I can make sure the Wave Interference RC is ready when you are.

Ariel Paul [3:58 PM]
My understanding was none of these were easy to reproduce (they happened but no clear steps)

Kathryn Woessner [3:59 PM]
3. Is fairly easy to reproduce
As is 2, though we already thought it should be left. We just got a better look while testing

Ariel Paul [4:00 PM]
Yes, 2 is fine, that is the "flash"

Kathryn Woessner [4:02 PM]
quality assurance
Posted using /giphy
As for 1.

Ariel Paul [4:08 PM]
Please prioritize 1 and 3 for Sam
@samreid realistically Tuesday morning would be appropriate for RC

Kathryn Woessner [4:10 PM]
Yeah. I don't know how to feel about 1. Since we still don't know what caused it I don't know how common it will be. May be worth it to have someone not expecting it use it for awhile and see if they see it.
But it probably will be fine for 1.0 since it will be updated soon. Depends on which we want less. Poor iOS 12 performance or the chance of this bug.

Ariel Paul [4:12 PM]
I want poor performance less. Doe Reset fix 1?

Kathryn Woessner [4:12 PM]
No
You need to refresh the page

Ariel Paul [4:12 PM]
But Refreshing the browser does...fair enough
I still want the chance of the bug more than poor performance

Sam Reid [4:18 PM]
We could disable clipping for ipads. It would look like this:

Ariel Paul [4:19 PM]
@samreid ask @arouinfar if she is OK with that. I could live with it for 1.0 (I think on sound it is not too crazy and I expect iPads to be lower use for this sim)

[samreid]

QA found this buggy clipping behavior: #322 (comment)
@ariel and I are wondering how you feel about disabling clipping altogether for iPads. It would look like this (see above):

Amy Rouinfar [4:22 PM]
I could definitely live with that

Sam Reid [4:22 PM]
@ariel said:

@samreid ask @arouinfar if she is OK with that. I could live with it for 1.0 (I think on sound it is not too crazy and I expect iPads to be lower use for this sim)

Amy Rouinfar [4:23 PM]
Seems better than the bugs QA found

Sam Reid [4:23 PM]
OK I’ll disable clipping for the particles, only for iPad. Thanks for your quick consultation!

Amy Rouinfar [4:23 PM]
Sounds good 🙂

[samreid]

Clipping removed, and I opened a new issue about restoring the clip area (at lower priority). Closing.

[samreid]

I tried removing the empty Rectangle from the sound screen particle view, and could not tell any difference, with or without webgl enabled, so I removed it. Flagging for a closer look during next QA test.

[KatieWoe]

For Dev.24. This issue was seen mostly on iOS 12 if I recall correctly. This platform is less supported now. @samreid should we discuss what this test should look like?

[KatieWoe]

Things look ok in dev.24

[KatieWoe]

This seems ok and has been published. Closing