CUDPP issue with CUDA 10

Hello,

I got an issue with the latest version CUDA (10.2).
I rebuild the CUDPP lib in my machine and run the example case g3Dprint.
It takes a very long time to “creating CUDPP” and can not visualize the result correctly.
But it works well if I direclty use the executable in ./bin.
I find that the executable in ./bin using the cudpp lib compiled by CUDA 9.
So I just wonder whether GVDB 1.1 is compatible with the latest version CUDA?
Anyway, I will try my build in another machine with CUDA 9.

Thanks for advice advanced.

Trigolds

Hello,

Actually I tested my case with CUDA 9 and got the same issue.
And the example case always work slow at the first time.
Well, this should be caused by the compilation of cuda kernel.
The visualization of result shows only the part of grid but not voxels.
I think that maybe because the USE_OPENVDB is false.
I try to build the openvdb lib on my win 64 machine.
I hope this problem can be solved.

Trigolds

Hi
Did you solve a problem?
I have the same problem.
cudaMalloc takes a few minutes, increases using RAM to 12GB, c:/pagefile.sys file grows to 50GB, and than fails.
Leonid

Yes, the problem has been solved.
Actually, it is not a real problem.
The issue only happens at the first time when you run the executable.
I think that is because nvcc is compiling the kernel.
So, your issue is not the same as mine.