Arch Dispatching

xsimd provides a generic way to dispatch a function call based on the architecture the code was compiled for and the architectures available at runtime. The xsimd::dispatch() function takes a functor whose call operator takes an architecture parameter as first operand, followed by any number of arguments Args... and turn it into a dispatching functor that takes Args... as arguments.

Warning

doxygenfunction: Unable to resolve function “xsimd::dispatch” with arguments “None”. Candidate function could not be parsed. Parsing error is Error when parsing function declaration. If the function has no return type: Error in declarator or parameters-and-qualifiers Invalid C++ declaration: Expecting “(” in parameters-and-qualifiers. [error at 73] template<class ArchList = supported_architectures, class F> XSIMD_INLINE detail::dispatcher< F, ArchList > dispatch (F &&f) noexcept ————————————————————————-^ If the function has a return type: Error in declarator or parameters-and-qualifiers If pointer to member declarator: Invalid C++ declaration: Expected ‘::’ in pointer to member (function). [error at 107] template<class ArchList = supported_architectures, class F> XSIMD_INLINE detail::dispatcher< F, ArchList > dispatch (F &&f) noexcept ———————————————————————————————————–^ If declarator-id: Invalid C++ declaration: Expecting “(” in parameters-and-qualifiers. [error at 107] template<class ArchList = supported_architectures, class F> XSIMD_INLINE detail::dispatcher< F, ArchList > dispatch (F &&f) noexcept ———————————————————————————————————–^

Following code showcases a usage of the xsimd::dispatch() function:

#include "sum.hpp"

// Create the dispatching function, specifying the architecture we want to
// target.
auto dispatched = xsimd::dispatch<xsimd::arch_list<xsimd::avx2, xsimd::sse2>>(sum{});

// Call the appropriate implementation based on runtime information.
float res = dispatched(data, 17);

This code does not require any architecture-specific flags. The architecture specific details follow.

The sum.hpp header contains the function being actually called, in an architecture-agnostic description:

#ifndef _SUM_HPP
#define _SUM_HPP
#include "xsimd/xsimd.hpp"

// functor with a call method that depends on `Arch`
struct sum
{
    // It's critical not to use an in-class definition here.
    // In-class and inline definition bypass extern template mechanism.
    template <class Arch, class T>
    T operator()(Arch, T const* data, unsigned size);
};

template <class Arch, class T>
T sum::operator()(Arch, T const* data, unsigned size)
{
    using batch = xsimd::batch<T, Arch>;
    batch acc(static_cast<T>(0));
    const unsigned n = size / batch::size * batch::size;
    for (unsigned i = 0; i != n; i += batch::size)
        acc += batch::load_unaligned(data + i);
    T star_acc = xsimd::reduce_add(acc);
    for (unsigned i = n; i < size; ++i)
        star_acc += data[i];
    return star_acc;
}

// Inform the compiler that sse2 and avx2 implementation are to be found in another compilation unit.
extern template float sum::operator()<xsimd::avx2, float>(xsimd::avx2, float const*, unsigned);
extern template float sum::operator()<xsimd::sse2, float>(xsimd::sse2, float const*, unsigned);
#endif

The SSE2 and AVX2 version needs to be provided in other compilation units, compiled with the appropriate flags, for instance:

// compile with -mavx2
#include "sum.hpp"
template float sum::operator()<xsimd::avx2, float>(xsimd::avx2, float const*, unsigned);

// compile with -msse2
#include "sum.hpp"
template float sum::operator()<xsimd::sse2, float>(xsimd::sse2, float const*, unsigned);