kernel_ops.h

Go to the documentation of this file.

 
#ifndef KERNEL_OPS_H
#define KERNEL_OPS_H
 
#include "config.h"
#include "fused/microkernels/microkernel_base.h"
#include "fused/kernel_ops/kernel_helpers.h"
#include "fused/kernel_ops/kernel_eval.h"
#include "fused/kernel_ops/kernel_reduce.h"
#include "fused/kernel_ops/kernel_compare.h"
#include "fused/kernel_ops/kernel_dot.h"
#include "fused/kernel_ops/kernel_gemm.h"
 
template <typename T, my_size_t Bits, typename Arch>
struct KernelOps
{
    using K = Microkernel<T, Bits, Arch>;
    static constexpr my_size_t simdWidth = K::simdWidth;
 
    // ========================================================================
    // Evaluation
    // ========================================================================
 
    template <typename Expr>
    FORCE_INLINE static void eval(T *output, const Expr &expr) noexcept
    {
        detail::KernelEval<T, Bits, Arch>::eval(output, expr);
    }
 
    // ========================================================================
    // Reductions
    // ========================================================================
 
    template <typename Expr>
    FORCE_INLINE static T reduce_min(const Expr &expr) noexcept
    {
        return detail::KernelReduce<T, Bits, Arch>::reduce_min(expr);
    }
 
    template <typename Expr>
    FORCE_INLINE static T reduce_max(const Expr &expr) noexcept
    {
        return detail::KernelReduce<T, Bits, Arch>::reduce_max(expr);
    }
 
    template <typename Expr>
    FORCE_INLINE static T reduce_sum(const Expr &expr) noexcept
    {
        return detail::KernelReduce<T, Bits, Arch>::reduce_sum(expr);
    }
 
    // ========================================================================
    // Comparisons
    // ========================================================================
 
    template <typename Expr1, typename Expr2>
    FORCE_INLINE static bool reduce_all_approx_equal(
        const Expr1 &lhs,
        const Expr2 &rhs,
        T tolerance) noexcept
    {
        return detail::KernelCompare<T, Bits, Arch>::reduce_all_approx_equal(lhs, rhs, tolerance);
    }
 
    // ========================================================================
    // Dot Products
    // ========================================================================
 
    template <typename Expr1, typename Expr2>
    FORCE_INLINE static T dot(
        const Expr1 &expr1, my_size_t base1, my_size_t stride1,
        const Expr2 &expr2, my_size_t base2, my_size_t stride2,
        my_size_t len) noexcept
    {
        return detail::KernelDot<T, Bits, Arch>::dot(
            expr1, base1, stride1,
            expr2, base2, stride2,
            len);
    }
 
    template <typename Expr1, typename Expr2>
    FORCE_INLINE static T naive_dot_physical(
        const Expr1 &expr1, my_size_t base1, my_size_t stride1,
        const Expr2 &expr2, my_size_t base2, my_size_t stride2,
        my_size_t len) noexcept
    {
        return detail::KernelDot<T, Bits, Arch>::naive_dot_physical(
            expr1, base1, stride1,
            expr2, base2, stride2,
            len);
    }
};
 
#endif // KERNEL_OPS_H