SSE FlockingSSE Tech Demo in Custom C++ Engine
About
Normally flocking would be implemented on the GPU for large sets of boids. However this project is an experiment to see whether SSE intrinsics could be used in other applications other than library functions. In the demo I implemented flocking on the CPU in C++ and in intrinsics. The shared/duplicate code between them are implemented in C++, which include the data structures that hold the boids and the octree for spatial partitioning. No logic has been changed between the approaches for a reasonably fair performance gauge. There is very minimal graphics optimizations since the primary goal of this experiment is testing the performance of SSE intrinsics in alternate applications. The difference in performance is very minimal. Only with around 50k boids the difference in performance becomes slightly more visible. |
IMAGES
|
