A CUDA implementations of work-efficient parallel scan (Blelloch, O(n) work), stream compaction, and LSD radix sort - built from scratch with shared memory, bank conflict avoidance, and recursive scanning for arbitrary array sizes up to 1B elements.
cpp gpu parallel-computing cuda high-performance-computing prefix-sum shared-memory radix-sort parallel-algorithms stream-compaction
-
Updated
Sep 18, 2025 - Python