mcmingchang
mcmingchang
Although my solution solves GPU memory overflow, CUDAFree will slow down the entire computation speed. Is there a better solution?
I am still researching how to call knn and obtain the first N nearest sequences
I need to rewrite the algorithm into a GPU version for the project. I have already completed voxelization and normal vector calculation, but I still need the final ICP and...
The reason is that the construction of the KD tree failed，tree.numPrims=10000，but tree.numNodes=2
I found that the problems with low version compilation can only be solved using the new version.Then I found that there is a low probability of read-write conflicts in the...
> I found that the problems with low version compilation can only be solved using the new version.Then I found that there is a low probability of read-write conflicts in...
The reason why I use Spatial KDTREE is that it can preserve the constructed tree and create multiple different trees, which I can selectively call at different times. In cukd:...
The parameter for thrust should be used to apply for video memory using CUDAMallocManaged. According to the official documentation, using CUDADeviceSynchronize allows the CPU to directly access the values inside....
When building a SpatialKDTree, using compute_75 will result in calculation failure, while compute_80 will result in everything being normal
error ：too many resources requested for launch The selectSplits and updatePrims calls under the builder function of SpatialKDTree are suspected to be due to the data type atomic_row not being...