Failure tracing OpenGL ES application in gapid\gapii\cc\pool.cpp "Out of memory allocating 0x1f8000 bytes" #1554
Comments
|
This looks suspiciously like: If you are using external images, this is likely the cause. I will leave this open so we can track it in case it is not the same issue. |
|
Thanks for the quick response. I am not aware that we are using external images, unless it is something SDL is doing underneath the hood. |
|
We're very close to releasing V1.0.1, which fixes #1508. I'll ping this bug once it is released. |
|
v1.0.1 is now public. Please let me know if this fixes your crash. Cheers, |
|
Thanks. Unfortunately I'm still seeing the same crash: Here is the log: |
|
By the way, I notice this which looks a little suspicious and related to that fix you pointed me to: [T:\src\github\src\github.com\google\gapid\gapii\cc\gles_extras.cpp:548] Created AndroidNativeBufferExtra: width=576, height=896, layers=690044 |
|
I was able to capture the same application on Samsung Galaxy S7 Android 6.0.1 and GAPID is working properly with it so it seems specific to the NVIDIA Tegra K1. |
|
Hi @danginsburg, Sorry the fix didn't work for you. It would appear that for this particular device a reserved field (which in ASOP should be 0), has been used for something else. This reserved field is now in use in Android O to describe the number of layers in, which appears to hold the value 690044. Attempting to allocate 690044 576x896 images, unsurprisingly, is running out of memory. I'm knocking together another fix which will ignore this field for devices running Android N or lower (where the field was not introduced). We'll spin v1.0.2 to contain just this fix. |
ben-clayton
added a commit
to ben-clayton/gapid
that referenced
this issue
Jan 18, 2018
|
@danginsburg - can I ask what host OS you're running? |
|
@ben-clayton Win10 x64 |
|
Hi @danginsburg, The custom build is here. Due to Google policy, I will need to grant you permission once you click the link. Please note this build is unsigned, and may require clicking a few run-program-downloaded-from-internet dialogs to get going. Please let me know how you get on. |
|
@ben-clayton I can confirm that I am now able to capture a trace so that build fixed the problem. This bug can be closed. Thanks! There are some issues in the trace itself, but I'm still debugging some rendering issues so I'll file those once I'm sure they are not my bugs. |
|
Hooray! Please file bugs for anything weird or broken. If you are able to share Many thanks, |
ben-clayton
added a commit
that referenced
this issue
Jan 18, 2018
|
Hi @ben-clayton, do you mind pushing a new release with the fix above? I ran into same issue using 1.0.1 and verified that current master fixes it. |
|
Hi @marcinkwiatkowski - sure, give me a couple of working days to get it out - the team have been traveling around a lot the past couple of weeks, and we have a lot of stuff to catch up on. If you want to request access to the file above, please feel free to. Unfortunately there's no option to just make it public, otherwise I would. |
ben-clayton
added a commit
that referenced
this issue
Jan 26, 2018
|
@marcinkwiatkowski - I've pushed 1.0.2 which contains this fix. Please let me know how you get on. |
|
Thanks @ben-clayton, much appreciated ❤️. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
I am trying to capture a trace of an OpenGL ES application using gapid on NVIDIA Shield Tablet K1 using GAPID v1.0.0 release and it is crashing trying to capture a trace. I'm unfortunately not able to provide a repro at this time, but perhaps some of the info here will help.
This is the backtrace:
The text was updated successfully, but these errors were encountered: