Problem: integrating OptiX into an existing project

Hello,
I’d like to integrate OptiX into one of my existing project. I have installed version 8 and all the SDKs work correctly. I’m just adding the optixHello to my project. Everything compiles perfectly and go for “optixHello_generated_draw_solid_color.cu.optixir” in the /lib/build/ptx folder. The getInputData recognises the .optixir file correctly. optixModuleCreate works and so does optixProgramGroupCreate. However I get an error at optixPipelineCreate like this:

[ 4][ DISKCACHE]: Cache data size: “7.1 MiB”
[ 4][ DISKCACHE]: Cache hit for key: ptx-…-drv535.129.03

[ 4][ COMPILER]:

*** Break *** floating point exception

Thread 3 (Thread “cuda-EvtHandlr”):
#0 in poll () from /lib/x86_64-linux-gnu/libc.so.6
#1 in ?? () from /lib/x86_64-linux-gnu/libcuda.so.1
#2 in ?? () from /lib/x86_64-linux-gnu/libcuda.so.1
#3 in ?? () from /lib/x86_64-linux-gnu/libcuda.so.1
#4 in ?? () from /lib/x86_64-linux-gnu/libc.so.6
#5 in ?? () from /lib/x86_64-linux-gnu/libc.so.6

Thread 2 (Thread “cuda-EvtHandlr”):
#0 in poll () from /lib/x86_64-linux-gnu/libc.so.6
#1 in ?? () from /lib/x86_64-linux-gnu/libcuda.so.1
#2 in ?? () from /lib/x86_64-linux-gnu/libcuda.so.1
#3 in ?? () from /lib/x86_64-linux-gnu/libcuda.so.1
#4 in ?? () from /lib/x86_64-linux-gnu/libc.so.6
#5 in ?? () from /lib/x86_64-linux-gnu/libc.so.6

#5
#6 in ?? () from /lib/x86_64-linux-gnu/libnvidia-rtcore.so.535.129.03
#7 in ?? () from /lib/x86_64-linux-gnu/libnvidia-rtcore.so.535.129.03
#8 in ?? () from /lib/x86_64-linux-gnu/libnvidia-rtcore.so.535.129.03
#9 in ?? () from /lib/x86_64-linux-gnu/libnvidia-rtcore.so.535.129.03
#10 in ?? () from /lib/x86_64-linux-gnu/libnvidia-rtcore.so.535.129.03
#11 in ?? () from /lib/x86_64-linux-gnu/libnvidia-rtcore.so.535.129.03
#12 in ?? () from /lib/x86_64-linux-gnu/libnvidia-rtcore.so.535.129.03
#13 in ?? () from /lib/x86_64-linux-gnu/libnvidia-rtcore.so.535.129.03
#14 in ?? () from /lib/x86_64-linux-gnu/libnvidia-rtcore.so.535.129.03
#15 in ?? () from /lib/x86_64-linux-gnu/libnvidia-rtcore.so.535.129.03
#16 in ?? () from /lib/x86_64-linux-gnu/libnvidia-rtcore.so.535.129.03
#17 in ?? () from /lib/x86_64-linux-gnu/libnvidia-rtcore.so.535.129.03
#18 in ?? () from /lib/x86_64-linux-gnu/libnvidia-rtcore.so.535.129.03
#19 in ?? () from /lib/x86_64-linux-gnu/libnvidia-rtcore.so.535.129.03
#20 in ?? () from /lib/x86_64-linux-gnu/libnvidia-rtcore.so.535.129.03
#21 in ?? () from /lib/x86_64-linux-gnu/libnvidia-rtcore.so.535.129.03
#22 in ?? () from /lib/x86_64-linux-gnu/libnvidia-rtcore.so.535.129.03
#23 in ?? () from /lib/x86_64-linux-gnu/libnvidia-rtcore.so.535.129.03
#24 in ?? () from /lib/x86_64-linux-gnu/libnvidia-rtcore.so.535.129.03
#25 in ?? () from /lib/x86_64-linux-gnu/libnvoptix.so.1
#26 in optixPipelineCreate ()

The symbolic link between libnvoptix.so.1 and libnvidia-rtcore.so.535.129.03 is done.
Do you know where this could come from, knowing that nothing differs from optixHello, which works well independently ?

Thanks.

If I understand your question correctly, you only see that inside your own project and not inside the OptiX SDK examples you built yourself on the same machine.
Did you build release and/or debug targets and did either work?
(Note that debug OptiX SDK examples are really slow because of the nvcc option -G for device debug code used only in debug targets.)

Are you running this from inside a debugger?
Do you have exceptions enabled inside the debugger which would be captured and result in that *** Break *** floating point exception?
If that is just a break and not a crash, does the program continue when stepping over that breakpoint?
Do release mode builds trigger the same break?

I have never seen a floating point exception during some pipeline compilation before.
(Disclaimer: I’m not a Linux user.)