
Advertisement
This topic is 2204 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.
If you intended to correct an error in the post then please contact us.
Recommended Posts
Advertisement
Thanks for the sorting algorithm links.
I've run into another roadblock:
My octree construction requires that I check each parent node in parallel if it is flagged for subdivision. Then the createNodes function needs to be run in parallel based on these flags.
For instance, after my data is sorted, and the first four nodes are flagged for subdivision, I want createNodes to run for each one simultaneously so that:
Node 1 checks voxBuf[0~50,000] for which side of each axis it lies on.
Node 2 checks voxBuf[50,001~100,000] for which side of each axis it lies on.
..etc.
In order to do this, I will have to allocate the first 50,000 threads to node 1 and the next 50,000 to node 2 and so on...
Is it possible to allocate threads in this way? Does anyone know how to do this? Edited by gboxentertainment
I've run into another roadblock:
My octree construction requires that I check each parent node in parallel if it is flagged for subdivision. Then the createNodes function needs to be run in parallel based on these flags.
For instance, after my data is sorted, and the first four nodes are flagged for subdivision, I want createNodes to run for each one simultaneously so that:
Node 1 checks voxBuf[0~50,000] for which side of each axis it lies on.
Node 2 checks voxBuf[50,001~100,000] for which side of each axis it lies on.
..etc.
In order to do this, I will have to allocate the first 50,000 threads to node 1 and the next 50,000 to node 2 and so on...
Is it possible to allocate threads in this way? Does anyone know how to do this? Edited by gboxentertainment
0
Okay, so now I've decided to go with the MortonCode sorting octree method.
I have interlaced all of my xyz coordinates into a singledimensional list of morton codes. Then I have used the Compute Shader Bitonic Sort example to sort this list of morton codes in ascending order.
I can see that the way the MortonCode is sorted from ascending order essentially represents an Octree. Does anyone know of any examples that implement the octree structure using MortonCodes in parallel on the gpu?
I have interlaced all of my xyz coordinates into a singledimensional list of morton codes. Then I have used the Compute Shader Bitonic Sort example to sort this list of morton codes in ascending order.
I can see that the way the MortonCode is sorted from ascending order essentially represents an Octree. Does anyone know of any examples that implement the octree structure using MortonCodes in parallel on the gpu?
0
Sign in to follow this
Followers
0

Advertisement
see here for example for an OpenCL implementation: http://www.bealto.co...ting_intro.html
Share this post
Link to post
Share on other sites