I have an Oculus Go app that is crashing for people with disturbing frequency. Some users know how to adb logcat, and what they send me looks like this.
So it seems like it’s running out of GPU memory, I guess? But I’m not certain I’m interpreting it correctly. Nor do I see why that causes a crash rather than something like pink materials.
Any tips on how to narrow this down, and what I can do about it?
In case it’s useful, here’s the result of adb shell dumpsys meminfo on my device after playing a while. (I did not experience a crash, however.)
Applications Memory Usage (in Kilobytes):
Uptime: 739145165 Realtime: 925678597
** MEMINFO in pid 28970 [com.alvios.quivr] **
Pss Private Private SwapPss Heap Heap Heap
Total Dirty Clean Dirty Size Alloc Free
------ ------ ------ ------ ------ ------ ------
Native Heap 602902 602772 0 0 642560 621237 21322
Dalvik Heap 4577 4156 0 0 9933 5960 3973
Dalvik Other 711 708 0 0
Stack 1004 1004 0 0
Ashmem 16 4 0 0
Gfx dev 470520 409952 0 0
Other dev 8 0 8 0
.so mmap 16584 500 14060 0
.jar mmap 4 0 4 0
.apk mmap 7700 340 3368 0
.dex mmap 2197 12 1920 0
.oat mmap 4954 0 2352 0
.art mmap 3388 1568 588 0
Other mmap 68 4 28 0
EGL mtrack 29600 29600 0 0
Unknown 25326 25312 0 0
TOTAL 1169559 1075932 22328 0 652493 627197 25295
App Summary
Pss(KB)
------
Java Heap: 6312
Native Heap: 602772
Code: 22556
Stack: 1004
Graphics: 439552
Private Other: 26064
System: 71299
TOTAL: 1169559 TOTAL SWAP PSS: 0
Objects
Views: 6 ViewRootImpl: 1
AppContexts: 2 Activities: 1
Assets: 3 AssetManagers: 2
Local Binders: 29 Proxy Binders: 26
Parcel memory: 15 Parcel count: 60
Death Recipients: 2 OpenSSL Sockets: 2
WebViews: 0
No ideas?
What about some way to measure how much GPU memory is in use (ideally, as a percentage of total), at runtime, so I can see whether my efforts at optimization are having any effect?
(I did go through with Resource Checker and reduce all my texture sizes; most of them are 1k by 1k now. But the crash is still occurring.)