pystencils.sympyextensions.fast_approximation.fast_inv_sqrt

class pystencils.sympyextensions.fast_approximation.fast_inv_sqrt(*args)

Perform a fast approximate inverse square root on GPU targets.

This function is only supported on CUDA and HIP targets, where it will resolve to __frsqrt_rn. Using it on other targets will lead to an error during code generation.

See also

• CUDA Fp32 intrinsics

• HIP math API