KokkosBlas::mult

Defined in header: KokkosBlas1_mult.hpp

template <class execution_space, class YMV, class AV, class XMV>
void mult(const execution_space& space, typename YMV::const_value_type& beta, const YMV& Y,
          typename AV::const_value_type& alpha, const AV& A, const XMV& X);

template <class YMV, class AV, class XMV>
void mult(typename YMV::const_value_type& beta, const YMV& Y, typename AV::const_value_type& alpha, const AV& A,
          const XMV& X);

computes an element wise multiplication of A and X scaled by alpha, then accumulate into Y scaled by beta.

\[Y = \beta * Y + \alpha * A * X\]
  1. Perform the entry wise multipliaction and addition using space instance.

  2. Perform the entry wise multipliaction and addition using the default instance of typename XMV::execution_space.

The function will throw a runtime exception if 1. the vectors do not have same length: Y.extent(0) != A.extent(0) || Y.extent(0) != X.extent(0), 2. the number of vectors in X and Y does not match: Y.extent(2) != X.extent(1) for the rank-2 case.

Parameters

space:

execution space instance

beta:

scaling factor for Y

Y:

output vector(s)

alpha:

scaling factor for the A, X product

A:

input vector to multiply element wise with X

X:

input vector(s)

Type Requirements

  • execution_space must be a Kokkos execution space

  • YMV must be a Kokkos View of rank 1 or 2, that satisfies

    • Kokkos::SpaceAccessibility<execution_space, typename XMV::memory_space>::accessible

    • std::is_same_v<typename YMV::value_type, typename YMV::non_const_value_type>

  • AV must be a Kokkos View of rank 1

  • XMV must be a Kokkos View of rank 1 or 2, that satisfies

    • Kokkos::SpaceAccessibility<execution_space, typename XMV::memory_space>::accessible

    • XMV::rank == YMV::rank

Example

 1#include <iostream>
 2#include <Kokkos_Core.hpp>
 3#include <KokkosBlas1_mult.hpp>
 4
 5int main(void) {
 6  Kokkos::initialize();
 7  {
 8    Kokkos::View<double*> a("a", 5), x("x", 5), y("y", 5);
 9    auto h_y = Kokkos::create_mirror_view(y);
10    h_y(0)   = 3.0;
11    h_y(1)   = 2.0;
12    h_y(2)   = 3.0;
13    h_y(3)   = -2.0;
14    h_y(4)   = 4.0;
15    auto h_a = Kokkos::create_mirror_view(a);
16    h_a(0)   = -3.0;
17    h_a(1)   = -1.0;
18    h_a(2)   = 2.0;
19    h_a(3)   = -2.0;
20    h_a(4)   = -3.0;
21    auto h_x = Kokkos::create_mirror_view(x);
22    h_x(0)   = 0.0;
23    h_x(1)   = -1.0;
24    h_x(2)   = -2.0;
25    h_x(3)   = 3.0;
26    h_x(4)   = -4.0;
27    Kokkos::deep_copy(y, h_y);
28    Kokkos::deep_copy(a, h_a);
29    Kokkos::deep_copy(x, h_x);
30
31    std::cout << "orignal values:\n"
32              << "   beta: 2.0,  alpha: -1.0\n"
33              << "   y: {" << h_y(0) << ", " << h_y(1) << ", " << h_y(2) << ", " << h_y(3) << ", " << h_y(4) << "}\n"
34              << "   a: {" << h_a(0) << ", " << h_a(1) << ", " << h_a(2) << ", " << h_a(3) << ", " << h_a(4) << "}\n"
35              << "   x: {" << h_x(0) << ", " << h_x(1) << ", " << h_x(2) << ", " << h_x(3) << ", " << h_x(4) << "}\n"
36              << std::endl;
37
38    KokkosBlas::mult(2.0, y, -1.0, a, x);
39
40    Kokkos::deep_copy(h_y, y);
41    std::cout << "final values:\n"
42              << "   y: {" << h_y(0) << ", " << h_y(1) << ", " << h_y(2) << ", " << h_y(3) << ", " << h_y(4) << "}"
43              << std::endl;
44  }
45  Kokkos::finalize();
46}

output:

orignal values:
   beta: 2.0,  alpha: -1.0
   y: {3, 2, 3, -2, 4}
   a: {-3, -1, 2, -2, -3}
   x: {0, -1, -2, 3, -4}
final values:
   y: {6, 3, 10, 2, -4}