can't run the vpt

[chena1982]

Win 10, cuda 10.1, vs 2017

GVDB CUDA ERROR:
Launch status: file not found
Kernel status: no error
Caller: VolumeGVDB::LoadFunction
Call: cuModuleLoad
Args: D:/PROJECTS/gvdb-voxels/_output/lib/cuda_gvdb_module.ptx
Error. Application will exit.

i don't compile gvdb-voxels myself, so this is because some hardcode the gvdb path?

ps, how to compile shared_cudpp which in the gvdb-voxels?
use the cmake, can't compile the shared _cudpp.

Thanks.

[sergeneren]

Yes the file path is hardcoded at main.cpp. Try full path if you are having problems at finding the file.

For compilation of cudapp, please refer to https://github.com/NVIDIA/gvdb-voxels

[chena1982]

Could you please upload a vpt.exe, Thanks.

[sergeneren]

I can but the exe will still be my file system specific. I will add the vdb file path as an argument later.

Could you try a full path to vdb in "gvdb.LoadVDB(scnpath);" replacing scnpath.
main.cpp line 877

[chena1982]

Could you try a full path to vdb in "gvdb.LoadVDB(scnpath);" replacing scnpath.
main.cpp line 877

gvdb.LoadVDB("C:/GitHub/Volumetric - Path - Tracer/assets/wdas_cloud_quarter_filled.vdb");

Don't work, it don't failed here.
It's failed when initialize the gvdb, line 867, init_gvdb(); the "cuda_gvdb_module.ptx" path "D:/PROJECTS/gvdb-voxels/_output/lib" is hardcode in the libgvdb.lib.

I compile the libgvdb.lib my self now, but it still failed, when call cudaMalloc, it will lead to out of memory, it's strange, because cudaMalloc should alloc memory on GPU.

[sergeneren]

Ok. Will look into that.

[chena1982]

The crash is beed fixed, it's a bug of shared_cudpp.

But the vdb can't been loaded, becuase when generate the project sln file, the macro "BUILD_OPENVDB" didn't been add to the project.

I find the "FindOpenVDB.cmake" in folder thirdparty\gvdb\sample_utils, but in the source\CMakeLists.txt, it didn't use openvdb lib.

[sergeneren]

Can you try building this fork https://github.com/sergeneren/gvdb-voxels

I've automated cmake a little bit to grab openvdb libraries from vcpkg.

[chena1982]

Thanks.
Finally, it works now.

ps, nvidia update some makefile recently, fix the error.