957925 - [LowMemory] Gallery app scroll shows stutter if enable APZ for all Apps

Status: RESOLVED DUPLICATE of bug 959719

(Core :: Panning and Zooming, defect, P1)

Description

[Tapas[:tkundu on #b2g/gaia/memshrink/gfx] (always NI me)]

Attachment: IMG_0001.jpg

Gallary app scroll shows stutter if enable APZ for all Apps

STR:

1) Copy attached image 200 times on device's sdcard. You will have 200 images on sdcard in this way.
2) Enable APZ for all apps 
3) Reboot
4) Open gallery app and try to scroll it fully.

Observation: It shows big stutters.

Comment 1

[Jason Smith [:jsmith]]

I think this is a dupe of bug 942750.

Comment 2

[Milan Sreckovic [:milan] (needinfo for best results)]

Tapas, is what you're talking about the same as described in bug 942750 as Jason mentions?

Comment 3

[Jason Smith [:jsmith]]

(In reply to Jason Smith [:jsmith] from comment #1)
> I think this is a dupe of bug 942750.

Vivien - Can you confirm that this is a dupe of that bug?

Comment 4

[Mandyam Vikram]

Actually, we see bad visual artifacts (blanking, black bands etc). Attaching amateur video.

Comment 5

[Milan Sreckovic [:milan] (needinfo for best results)]

Is there a video attached?

Comment 6

[Mandyam Vikram]

Attachment: 20140109_100258.mp4

Video demonstrating bad UI artifacts with AZPC enabled.

Comment 7

[Jason Smith [:jsmith]]

(In reply to Mandyam Vikram from comment #6)
> Created attachment 8357893 [details]
> 20140109_100258.mp4
> 
> Video demonstrating bad UI artifacts with AZPC enabled.

Okay, that's definitely a dupe of the bug I cited. It's the same behavior seen in bug 951259.

Comment 8

[Kartikaya Gupta (email:[e-mail])]

Interesting. The video here looks like what happens when memory is running low. There's probably pmem warnings in logcat when this happens. I thought bug 942750 was tracking checkerboarding but if it's tracking pmem issues then we should make that clear.

Comment 9

[Jason Smith [:jsmith]]

(In reply to Kartikaya Gupta (email:[e-mail]) from comment #8)
> Interesting. The video here looks like what happens when memory is running
> low. There's probably pmem warnings in logcat when this happens. I thought
> bug 942750 was tracking checkerboarding but if it's tracking pmem issues
> then we should make that clear.

So this isn't a dupe? I thought this issue was the same as what was seen in bug 951259, as the video looked similar to what was seen in that bug.

Comment 10

[Kartikaya Gupta (email:[e-mail])]

It depends. This bug and bug 951259 sound like the same issue. However bug 942750 comment 0 says nothing about flickering and flashing. That comment describes symptoms consistent with checkerboarding whereas this bug and bug 951259 describe symptoms consistent with low-memory issues.

Comment 11

[Kartikaya Gupta (email:[e-mail])]

Also just to note, the two will look somewhat different visually although it might be hard to discern unless you know what to look for. In the low-memory case entire layers will disappear (probably replaced with black) and things will flicker on and off. In the checkerboarding case things generally don't flicker except when panning at high speed as you are racing gecko in and out of painted areas. Also with the checkerboarding scenario you can and often do end up with part of the content showing while the rest of the visible area is blank with the background color.

Comment 12

[Jason Smith [:jsmith]]

Okay - I'm going to reopen bug 951259 to track the low memory issue & dupe this bug over there.

Comment 13

[Kartikaya Gupta (email:[e-mail])]

Reopening; see bug 951259 comment 10.

Comment 14

[Kartikaya Gupta (email:[e-mail])]

Not sure if this should block or not, but it doesn't look very nice so. See video in comment 6.

Comment 15

[Preeti Raghunath(:Preeti)]

Kats,

Are you looking into this blocker?

Please provide next steps.

Comment 16

[Kartikaya Gupta (email:[e-mail])]

It's not #1 on my list but I can look into it after I deal with bug 951113. In the meantime I need to talk to Sotaro to understand better what the HWcomposer is doing because I see logcat output from HWComposer whenever I run into this (which isn't frequent, but I have seen it).

Comment 17

[Sotaro Ikeda [:sotaro]]

Tapas, can you provide a logcat log?

Comment 18

[Milan Sreckovic [:milan] (needinfo for best results)]

Assign to Sotaro while we're waiting for info.

Comment 19

[Mandyam Vikram]

Attachment: gallery.txt

Attaching logcat. The log captures loading the pictures from the SD card for the first time (which demonstrates a lot of the same UI artifacts) as well as scrolling through the pictures.

Comment 20

[Sotaro Ikeda [:sotaro]]

Mandyam, does the big stutters happens after all the data load complete? And which device and b2g version are you using? I checked the STR in Comment 0 on v1.3 hamachi and master hamachi. But I did not saw the  big stutters.

Comment 21

[Sotaro Ikeda [:sotaro]]

attachment 8359897 [details] does not have a log related to the following. From the log, I can not find the problematic things.
- It does not have a log related to pmem fallback.
    The hw might not using pmem, but using ion. And the hw might have more gralloc buffers than hamachi.
- The log does not have an information about the rendering is done by OpenGL/hw composer.

Comment 22

[Mandyam Vikram]

This is done on a 8610 MTP device (dual core, WVGA).

We see the black banding while the pictures are being loaded for the first time as well as during the scroll. Hopefully, the video demonstrates that.

Please send us a patch with the traces that you like to see.

Comment 23

[Mandyam Vikram]

Attachment: 20140114_185240.mp4

Attaching a video of gallery app importing pictures the first time.

Comment 24

[Sotaro Ikeda [:sotaro]]

(In reply to Mandyam Vikram from comment #22)
> This is done on a 8610 MTP device (dual core, WVGA).

Mandyam, Thanks for the info. Which version of andorid(gonk), does it use? ICS? JB? KK?

Comment 25

[Mandyam Vikram]

It is 1.3, JB MR2 based build.
By including the patch for https://bugzilla.mozilla.org/show_bug.cgi?id=959719 locally, the issue goes away.