Abstract Stream compaction is an important parallel computing primitive that produces a reduced (compacted) output stream consisting of only valid elements from an input stream containing both invalid and valid elements. Computing on this compacted stream rather than the mixed input stream leads to improvements in performance, load balancing and memory footprint. Stream compaction has numerous applications in a wide range of domains: e.g. deferred shading, isosurface extraction and surface voxelization in computer graphics and visualization. We present a novel In‐Kernel stream compaction method, where compaction is completed before leaving an operating kernel. This contrasts with conventional parallel compaction methods that require leaving the kernel and running a prefix sum kernel followed by a scatter kernel. We apply our compaction methods to ray‐tracing‐based visualization of volumetric data. We demonstrate that the proposed In‐Kernel compaction outperforms the standard out‐of‐kernel Thrust parallel‐scan method for performing stream compaction in this real‐world application. For the data visualization, we also propose a novel multi‐kernel ray‐tracing pipeline for increased thread coherency and show that it outperforms a conventional single‐kernel approach.

Stackless traversal techniques are often used to circumvent memory bottlenecks by avoiding a stack and replacing return traversal with extra computation. This paper addresses whether the stackless traversal approaches are useful on newer hardware and technology (such as CUDA). To this end, we present a novel stackless approach for implicit kd-trees, which exploits the benefits of index-based node traversal, without incurring extra node visitation. This approach, which we term Kd-Jump, enables the traversal to immediately return to the next valid node, like a stack, without incurring extra node visitation (kd-restart). Also, Kd-Jump does not require global memory (stack) at all and only requires a small matrix in fast constant-memory. We report that Kd-Jump outperforms a stack by 10 to 20% and kd-restart by 100%. We also present a Hybrid Kd-Jump, which utilizes a volume stepper for leaf testing and a run-time depth threshold to define where kd-tree traversal stops and volume-stepping occurs. By using both methods, we gain the benefits of empty space removal, fast texture-caching and realtime ability to determine the best threshold for current isosurface and view direction.