Spike in Flash Game Client Memory with Chrome 58.0.3029.96 stable release
Reported by
shrilol...@gmail.com,
May 11 2017
|
||||||
Issue descriptionUserAgent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_12_4) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/58.0.3029.110 Safari/537.36 Steps to reproduce the problem: 1. Load https://apps.facebook.com/farmville-two/ or https://apps.facebook.com/onthefarm on Chrome versions 58.0.3029.96 and 58.0.3029.81 versions 2. Check Used Memory on Activity monitor and any other flash memory profiling tools like Scout. 3. We can see an increase in memory on 58.0.3029.96 version. What is the expected behavior? What went wrong? We are seeing a spike in Farmville/Farmville2 (flash Games) Client side memory usage by more than 120MB. Firefox and IE memory remains flat and seeing spike with Chrome version 58.0.3029.96 release. We have around ~4M+ MAU coming from chrome. Increasing the memory leading to flash crash and players not being able to play the game. Can you please look into this issue and release the fix asap. Did this work before? N/A Chrome version: 58.0.3029.96 Channel: stable OS Version: Flash Version: Shockwave Flash 25.0 r0
,
May 12 2017
@shrilola: Can you please be more specific about which tool do we need to use to get the memory usage report of the Farmville/Farmville2 game. That will help us in further triaging of the issue. Thanks,
,
May 12 2017
@Sandeep: We have updated profiled scout file in google drive https://drive.google.com/drive/folders/0ByAw0MvLVeI0OGNrNzFJOXN0d0U You need scout app to open this flm file, you can download adobe scout from here. http://www.adobe.com/in/products/scout.html We have profiled a same level blob on Chrome 57 and 58 versions. Here is our observations on 57 version Total Memory used : 2256.233 MB Actionscript object memory: 1024.355 MB Bitmap : 414.488 MB ByteArrays: 427.849 MB SWF file: 41.37 MB Other(Network Buffers and some other overheads): 348.171 58 Version: Total Memory used : 2399.86 MB Actionscript object memory: 1081.513 MB Bitmap : 447.66 MB ByteArrays: 451.684 MB SWF file: 41.37 MB Other(Network Buffers and some other overheads): 376.572 Looks like there is an increase in ByteArray and Bitmap memory. Please let us know if you need any more details regarding this.
,
May 12 2017
Thank you for providing more feedback. Adding requester "sandeepkumars@chromium.org" to the cc list and removing "Needs-Feedback" label. For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot
,
May 12 2017
We have also profiled memory on Chrome 58.0.3029.110 and we do see a spike in this version too.
,
May 15 2017
This issue seems to be out of TE-scope. Hence, adding label TE-NeedsTriageHelp for further investigation. Thanks...!!
,
May 18 2017
Any update on this?
,
May 18 2017
Hi, We have created a test app to check the memory on different versions, we are consistently able to reproduce the issue on Chrome Version 58.0.3029.96 and above. Steps to reproduce the issue: 1. Load http://zlive-aws.s3.amazonaws.com/images/goat.html on Chrome version 57.0.2987.133 and 58.0.3029.96. 2. Open developer tool, check verbose log to get the memory. Take a reading of AmfTest Memory End. Note: You have to enable flash to get the data. On Chrome 57.0.2987.133 version total memory is 148115456 bytes (148.11 MB) and on Chrome 58.0.3029.96 version total memory is 186839040 bytes (186.83 MB). We are seeing around 38MB memory spike in a small app. system details: 1. Google Chrome 57.0.2987.133 (Official Build) (64-bit) Revision ec33cd0c06881d919ac0de419d829ad914e0be8f-refs/branch-heads/2987@{#887} OS Windows JavaScript V8 5.7.492.71 Flash 25.0.0.148 2. Google Chrome 58.0.3029.96 (Official Build) (64-bit) Revision 2a03c99a5f45c6d507af8eb2345ad68a565d1518-refs/branch-heads/3029@{#787} OS Windows JavaScript V8 5.8.283.37 Flash 25.0.0.148 Also attaching total game memory graph with this thread.
,
Jun 20 2017
lfg: Can you take a look? It seems like we have a reproducible Flash-related memory leak that occurs in a specific Chrome rev.
,
Jun 21 2017
I tried to reproduce your testcase but I can't find any measurable difference between M57 and M58. I can get the 148/186MBs test cases on both versions, and they aren't consistent, likely due to garbage collection timings. I strongly suspect that the numbers you are seeing from the clients are due to the fact that we migrated users to 64-bit from 32-bit starting with chrome 58. We only migrated users on 64-bit systems with >= 4GBs of RAM, and our internal data show that despite higher memory usage of the 64-bit version, we have fewer out-of-memory crashes due to having a larger address space. |
||||||
►
Sign in to add a comment |
||||||
Comment 1 by ligim...@chromium.org
, May 11 2017Labels: -Pri-2 Needs-Bisect Performance-Memory Needs-Triage-M58 Pri-1