Output details

InK-Compact: In-Kernel Stream Compaction and Its Application to Multi-Kernel Data Visualization on General-Purpose GPUs

<26>Stream compaction is an important parallel computing primitive among others (e.g. prefix-sum and parallel sorting). Producing a reduced (compacted) output stream of only valid elements from an mixed input stream, stream compaction leads to improvements in performance, load balancing, and memory footprint and thus has significant implications to various application domains. We present an innovative In-Kernel stream compaction method, where compaction is completed before leaving an operating kernel. The proposed stream compaction is demonstrated to outperform the state-of-art methods that require leaving the kernel and running a prefix sum kernel followed by a scatter kernel