SIMD

Header	Content	CCCL Availability	CUDA Toolkit Availability
cuda::simd::saturating_add	Performs element-wise saturating addition of two basic_vec objects	CCCL 3.5.0	CUDA 13.5
cuda::simd::dot	Computes the dot product of two basic_vec objects and an accumulator	CCCL 3.6.0	CUDA 13.6
cuda::simd::abs_diff	Performs element-wise absolute difference of two integer basic_vec objects	CCCL 3.6.0	CUDA 13.6
cuda::simd::min_relu and cuda::simd::max_relu	Perform rectified element-wise minimum or maximum of two or three signed integer basic_vec objects	CCCL 3.6.0	CUDA 13.6
cuda::simd::add_min, cuda::simd::add_max, cuda::simd::add_min_relu, and cuda::simd::add_max_relu	Perform element-wise addition followed by a minimum or maximum, optionally followed by ReLU	CCCL 3.6.0	CUDA 13.6