Segmentation fault when rendering

[Zod-L]

Hi,
I encountered a segmentation fault when trying to render the bunny.xml in src/optixRenderer/cbox. My environment is ubuntu 20.04, CUDA 10.2, optix 5.11.0. The error message is listed as follows:
Input file name: /home/zod/OptixRenderer/src/optixRenderer/cbox/bunny.xml
Output file name: /home/zod/OptixRenderer/src/optixRenderer/cbox/bunny.rgbe
Material num: 5
Shape num: 8
Vertex num: 72
Camera Parameters:
Camera Field of View: 60
Width: 512
Height: 512
FovAxis: x
Target: 278 273 -799
Origin: 278 273 -800
Up: 0 1 0
Sample Count: 8192
Sampler: independent
Sampler Count: 8192
No Environmental maps.
No point light source.
GPU Num: 1
The length of diagonal of bouncding box: 960.74
Max X: 556.000 Max Y: 548.800 Max Z: 559.200
Min X: 0.000 Min Y: 0.000 Min Z: 0.000
Total number of materials: 5
段错误 (核心已转储)

No output file is generated. I tried all examples in cbox, all encountering the same problem. Could you please help me with it?

[lzqsd]

That's very weird. I suggest trying different rendering modes, to render diffuse albedo and normal and see what's going on.

[Zod-L]

I have tried all other modes, all reporting the same segment fault problem.

[songxujay]

hello, what is your GPU volumn?

[jiangjiechu]

I encountered a similar problem, changing CMAKE_BUILD_TYPE from Release to Debug can solve this without loss of speed. I have no idea of how this happens. I guess it is caused by the mismatch of cudadriver and deprecated OptiX6 library. I once had a Release build working fine, but got Segmentation fault error after I upgraded my nvidia driver from 470 to 510 and linux headers from 5.4 to 5.10. The error persists even though I downgraded them.

[XiaoKangW]

Hi, i'm benefit from you r talking. and i have a problem, Are you using CUDA 10.2, optix 5.11.0 on RTX3090 ? I can't use the GPU properly when using the provided docker

[HospitableHost]

Hi, i'm benefit from you r talking. and i have a problem, Are you using CUDA 10.2, optix 5.11.0 on RTX3090 ? I can't use the GPU properly when using the provided docker

RTX3090 do not support cuda10

[xiao2021]

thank you very much [jiangjiechu],it takes well