Simulating thousands of sparks during a high-speed blade clash quickly chokes the CPU if handled in game logic. By offloading particle state updates directly to OpenGL 4.3 Compute Shaders via Shader Storage Buffer Objects (SSBOs), we bypass CPU-GPU bandwidth bottlenecks.
Each particle's position, velocity, lifespan, and color gradient are updated entirely in parallel on GPU execution units. A single dispatch call computes physics for 250,000 sparks with sub-millisecond overhead.
This architecture delivers visceral visual feedback during parries while keeping the game thread locked at a silky 144 FPS.