[go: up one dir, main page]

Skip to content
/ libcbt Public

Concurrent data-structure to compute binary trees in parallel

License

Notifications You must be signed in to change notification settings

jdupuy/libcbt

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

53 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Concurrent Binary Tree Library

This library provides a concurrent binary tree data-structure suitable for accelerating the processing of subdivision algorithms on multicore processors, including GPUs. More details are available in my paper "Concurrent Binary Trees (with application to Longest Edge Bisection)".

Usage

Initialization A CBT requires a maximum depth (typically the maximum depth of the subdivision algorithm you're interested in accelerating).

cbt_Tree *cbt = cbt_Create(myMaximumDepth);

Once this depth chosen, it can not be changed throughout the lifetime of the CBT instance. You can query the maximum depth of the CBT as follows:

maximumDepth = cbt_MaxDepth(cbt);

By default, the CBT will be initialized at the root node. You can also initialize it at an explicit subdivision depth as follows:

cbt_Tree *cbt = cbt_CreateAtDepth(myMaximumDepth, myInitializationDepth);

Note that the initialization depth must be less or equal to the maximum depth of the CBT. Always remember to release the meomory once you're done with your CBT:

cbt_Release(cbt);

Resetting the tree Additionally, you can reset the subdivision by using any of the following routines:

cbt_ResetToRoot(cbt); // resets the CBT to its root
cbt_ResetToCeil(cbt); // resets the CBT to its maximum depth
cbt_ResetToDepth(cbt, myInitializationDepth); // resets the CBT to a custom depth

Updating the tree in parallel The main advantage of CBTs is their ability to update their topology in parallel. Nodes can be split or merged using respectively cbt_SplitNode(cbt, node) and cbt_MergeNode(cbt, node). In order to process the operations in parallel, you can provide a custom callback that will be executed in parallel within an OpenMP parallel for loop. Here is a simple example that splits or merges nodes if their index is even:

// update callback
void UpdateCallback(cbt_Tree *cbt, const cbt_Node node, const void *userData)
{
    if ((node.id & 1) == 0) {
#ifdef SPLIT
        cbt_SplitNode(cbt, node);
#else
        cbt_MergeNode(cbt, node);
#endif        
    }
}

// execute the update callback in parallel
cbt_Update(cbt, &UpdateCallback, NULL);

For a more complex example, see this repo.

Queries You can query the number of leaf nodes in the CBT using

int64_t nodeCount = cbt_NodeCount(cbt);

You can retrieve the i-th leaf node using

cbt_Node node = cbt_DecodeNode(cbt, i);

Conversely, you can retrieve the index of an existing leaf node using

int64_t nodeID = cbt_EncodeNode(cbt, node);

Serialization Internally, the CBT uses a compact binary heap data-structure, i.e., a 1D array. This makes the CBT trivial to serialize. To access the heap, use

int64_t cbtByteSize = cbt_HeapByteSize(cbt); // size in Bytes of the CBT
char *cbtMemory = cbt_GetHeap(cbt); // CBT raw-data

GPU implementation The GLSL folder provides a GLSL implementation of the library. An HLSL port of the library would also be welcome. For a GPU implementation example, see this repo.

License

The code from this repository is released in public domain. You can do anything you want with them. You have no legal obligation to do anything else, although I appreciate attribution.

It is also licensed under the MIT open source license, if you have lawyers who are unhappy with public domain.

About

Concurrent data-structure to compute binary trees in parallel

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published