Search Issue Tracker
Fixed in 5.4.0
Votes
0
Found in
5.4.0b10
Issue ID
780820
Regression
No
[GLCore][ComputeShader] Dispatching specificing shader results in crash
Info:
I couldn't find realy compute shader crashing problem. But its related with loop and gpu memory accessing in it. Check gif.
My Findings:
"After some investigation its compute shader crash for sure.
For now I know that it crashes, because of loop using count from memory. It might be corrupted not sure.
At least not accessing memory bug.
So might be a high number, going to do a bit simulations.
Trying max unit 4294967295. No crash
Trying 0. (kind obvious) No crash.
Mhh maybe the compiler optimizes my constant values, got to try dynamic ones. **** still no crash…
Maybe value have to be supplied by buffer (Don’t remember how its called in gpu, got that fancy name).
Got another crazy idea, maybe they release compute buffer during the compute shader job (And glcore doesn’t have protection for it). Trying. Nop it still crashes without the buffer releases. Yea that would was stupid idea, not sure how it would be related to loops…
I should probably try with other gpu, maybe its amd problem (I Hope). Nop Nvidia Crash. Damn I love nvidia crashes, because when AMD crashes it looks like your computer going to explode."
Project Steps:
1. Open scene "AOBake"
2. Click select "EthanBody" in game view
3. Open window Tools->Vagabond-POLM->Open POLM Tool
4. Click button "Use Selected"
5. Click "Precompute"
6. Click "Bake AO"
7. Driver crash.
All about bugs
View bugs we have successfully reproduced, and vote for the bugs you want to see fixed most urgently.
Latest issues
- Bad Naming Convention in Shortcuts Window for Sprite Shape Editing
- Bad Naming Convention in Shortcuts Window for Shader Graph
- Bad Naming Convention in Shortcuts Window for Particle System
- "Cameras" tooltip is too far from the window
- Crash on D3D12CommonShader::ApplyGpuProgram when attaching material which samples "_UnityFBInput0" to "Full Screen Pass Renderer Feature" Component
Add comment