Search Issue Tracker

By Design

Votes

0

Found in

2020.3.35f1

2021.3.5f1

2022.1.5f1

2022.2.0a17

2023.1.0a1

Issue ID

UUM-1365

Regression

No

GPU hang or exceedingly long wait when Baking Lights

--

-

Reproduction steps:
1. Open the attached project
2. Open Window -> Rendering > Lighting
3. Click Generate Lighting

Expected result: The Lighting is baked in a reasonable amount of time
Actual result: Baking the Lighting takes exceedingly long and sometimes crashes

Reproducible with: 2020.1.17f1, 2020.2.0f1, 2021.1.0a10
Could not test with: 2018.4, 2019.4 (Compiler errors, assets cannot be loaded)

Notes:
The issue might not reproduce steadily
Tested with URP 8.3.1, 10.1.0, 10.2.2
Baking process seems to behave recursively, reaching a 40 sec ETA, moving to finalize after the 40 sec, then resetting back to 40 sec ETA, and repeating for about 2 hours.
Reproduces on both Mac and Windows
Reproducible with CPU and GPU PLM

From the log:
GPU hang occurred, CoreAnalytics returned false
Obtained 19 stack frames.
Thread 0x700007976000 may have been prematurely finalized
#0 0x007fff2030d462 in __pthread_kill
Thread 0x700007976000 may have been prematurely finalized
#1 0x007fff2028e720 in abort
Thread 0x700007976000 may have been prematurely finalized
#2 0x007fff461310a8 in gpumInitializeIOData.cold.1
Thread 0x700007976000 may have been prematurely finalized
#3 0x007fff4612820d in gpumCreateComputeContext
Thread 0x700007976000 may have been prematurely finalized
#4 0x000001653328fa in gpusKillClientExt
Thread 0x700007976000 may have been prematurely finalized
#5 0x007fff46128bac in gpusQueueSubmitDataBuffers
Thread 0x700007976000 may have been prematurely finalized
#6 0x00000165302ab2 in glrAMDCompSubmitPackets
Thread 0x700007976000 may have been prematurely finalized
#7 0x0000016530299a in glrAMDFlush
Thread 0x700007976000 may have been prematurely finalized
#8 0x00000165354306 in glrAMDExecuteKernel
Thread 0x700007976000 may have been prematurely finalized
#9 0x00000165356fbf in gldExecuteKernel
Thread 0x700007976000 may have been prematurely finalized
#10 0x007fff6ca46428 in clFlush
Thread 0x700007976000 may have been prematurely finalized
#11 0x007fff6ca5e227 in clSetEventCallback
Thread 0x700007976000 may have been prematurely finalized
#12 0x007fff6ca60110 in clCreateCommandQueueWithPropertiesAPPLE
Thread 0x700007976000 may have been prematurely finalized
#13 0x007fff201907c7 in _dispatch_client_callout
Thread 0x700007976000 may have been prematurely finalized
#14 0x007fff201965fe in _dispatch_lane_serial_drain
Thread 0x700007976000 may have been prematurely finalized
#15 0x007fff201970cb in _dispatch_lane_invoke
Thread 0x700007976000 may have been prematurely finalized
#16 0x007fff201a0c5d in _dispatch_workloop_worker_thread
Thread 0x700007976000 may have been prematurely finalized
#17 0x007fff20338499 in _pthread_wqthread
Thread 0x700007976000 may have been prematurely finalized
#18 0x007fff20337467 in start_wqthread

  1. Resolution Note:

    The project has a probe sample multiplier of 100 and uses 2K indirect samples. So roughly 230000 samples per probe. It has 10 bounces and 150+ baked lights. So the bake is relatively complex in nature.

    With LightBaker in 2023.1:

    The project bakes in 20 min on CPU Lightmapper with LightBaker (with a reasonably stable eta).

    The project bakes in 1 min on GPU Lightmapper (RTX 2080) with LightBaker (with a reasonably stable eta).

    This is to be expected, given the complexity of the bake.

Add comment

Log in to post comment

All about bugs

View bugs we have successfully reproduced, and vote for the bugs you want to see fixed most urgently.