Do you have an idea if it’s possible that this fast SIMD sort could be applied for faster sorting on GPUs?