8#ifndef KERNEL_HELPERS_H
9#define KERNEL_HELPERS_H
17 template <
typename T, my_
size_t Bits,
typename Arch>
32 if constexpr (
requires { K::fmadd(a, b, c); })
34 return K::fmadd(a, b, c);
Global configuration for the tesseract tensor library.
#define FORCE_INLINE
Hint the compiler to always inline a function.
Definition config.h:26
Definition microkernel_base.h:16
T VecType
Definition microkernel_base.h:18
static FORCE_INLINE VecType mul(VecType a, VecType b) noexcept
static FORCE_INLINE VecType add(VecType a, VecType b) noexcept
Definition kernel_helpers.h:19
static FORCE_INLINE K::VecType fmadd_safe(typename K::VecType a, typename K::VecType b, typename K::VecType c) noexcept
Fused multiply-add with fallback for architectures without native FMA.
Definition kernel_helpers.h:27