CUDA Stream Compaction

CUDA, C++. I implement GPU stream compaction algorithms, and compare their performance.

By Michael Mason

1 min read

In this project, I implemented a GPU stream compaction algorithm using CUDA, with the goal of removing zeros from an array of integers.

Stream compaction is a critical technique in GPU programming, widely applicable in various scenarios (i.e. path tracing where terminated rays are compacted out of the working set).

This project involved both CPU and GPU implementations of the all-prefix-sums (or simply scan) algorithm as a foundation for the compaction process.

I used Chapter 39 from GPU Gems 3 as the basis for the GPU implementations.

See my GitHub Repo…

This post is licensed under CC BY 4.0 by the author.