SIMD library

From cppreference.com
 
C++
 
Technical specifications
Filesystem library (filesystem TS)
Library fundamentals (library fundamentals TS)
Library fundamentals 2 (library fundamentals TS v2)
Library fundamentals 3 (library fundamentals TS v3)
Extensions for parallelism (parallelism TS)
Extensions for parallelism 2 (parallelism TS v2)
Extensions for concurrency (concurrency TS)
Extensions for concurrency 2 (concurrency TS v2)
Concepts (concepts TS)
Ranges (ranges TS)
Reflection (reflection TS)
Mathematical special functions (special functions TR)
 
Extensions for parallelism v2
Parallel exceptions
Additional execution policies
Algorithms
                                        
                                        
Task blocks
Data-parallel vectors
 
SIMD library
 

The SIMD library provides portable types for explicitly stating data-parallelism and structuring data for more efficient SIMD access.

An object of type simd<T> behaves analogue to objects of type T. But while T stores and manipulates one value, simd<T> stores and manipulates multiple values (called width but identified as size for consistency with the rest of the standard library; cf. simd_size).

Every operator and operation on simd<T> acts element-wise (except for horizontal operations, which are clearly marked as such). This simple rule expresses data-parallelism and will be used by the compiler to generate SIMD instructions and/or independent execution streams.

The width of the types simd<T> and native_simd<T> is determined by the implementation at compile-time. In contrast, the width of the type fixed_size_simd<T, N> is fixed by the developer to a certain size.

A recommended pattern for using a mix of different SIMD types with high efficiency uses native_simd and rebind_simd: 

#include <experimental/simd>
namespace stdx = std::experimental;
 
using floatv  = stdx::native_simd<float>;
using doublev = stdx::rebind_simd_t<double, floatv>;
using intv    = stdx::rebind_simd_t<int, floatv>;

This ensures that the set of types all have the same width and thus can be interconverted. A conversion with mismatching width is not defined because it would either drop values or have to invent values. For resizing operations, the SIMD library provides the split and concat functions.

Defined in header <experimental/simd>

Main classes

(parallelism TS v2)
 data-parallel vector type
(class template)
(parallelism TS v2)
 data-parallel type with the element type bool
(class template)

ABI tags

Defined in namespace std::experimental::simd_abi
(parallelism TS v2)
 tag type for storing a single element
(typedef)
(parallelism TS v2)
 tag type for storing specified number of elements
(alias template)
(parallelism TS v2)
 tag type that ensures ABI compatibility
(alias template)
(parallelism TS v2)
 tag type that is most efficient
(alias template)
(parallelism TS v2)
 the maximum number of elements guaranteed to be supported by fixed
(constant)
(parallelism TS v2)
 obtains an ABI type for given element type and number of elements
(class template)

Alignment tags

(parallelism TS v2)
 flag indicating alignment of the load/store address to element alignment
(class)
(parallelism TS v2)
 flag indicating alignment of the load/store address to vector alignment
(class)
(parallelism TS v2)
 flag indicating alignment of the load/store address to the specified alignment
(class template)

Where expression

(parallelism TS v2)
 selected elements with non-mutating operations
(class template)
(parallelism TS v2)
 selected elements with mutating operations
(class template)
(parallelism TS v2)
 produces const_where_expression and where_expression
(function template)

Casts

(parallelism TS v2)
 element-wise static_cast
(function template)
(parallelism TS v2)
 element-wise ABI cast
(function template)
(parallelism TS v2)
 splits single simd object to multiple ones
(function template)
(parallelism TS v2)
 concatenates multiple simd objects to a single one
(function template)

Algorithms

(parallelism TS v2)
 element-wise min operation
(function template)
(parallelism TS v2)
 element-wise max operation
(function template)
(parallelism TS v2)
 element-wise minmax operation
(function template)
(parallelism TS v2)
 element-wise clamp operation
(function template)

Reduction

(parallelism TS v2)
 reduces the vector to a single element
(function template)

Mask reduction

(parallelism TS v2)
 reductions of simd_mask to bool
(function template)
(parallelism TS v2)
 reduction of simd_mask to the number of true values
(function template)
(parallelism TS v2)
 reductions of simd_mask to the index of the first or last true value
(function template)

Traits

(parallelism TS v2)
 checks if a type is a simd or simd_mask type
(class template)
(parallelism TS v2)
 checks if a type is an ABI tag type
(class template)
(parallelism TS v2)
 checks if a type is a simd flag type
(class template)
(parallelism TS v2)
 obtains the number of elements of a given element type and ABI tag
(class template)
(parallelism TS v2)
 obtains an appropriate alignment for vector_aligned
(class template)
(parallelism TS v2)
 change element type or the number of elements of simd or simd_mask
(class template)

Math functions

All functions in <cmath>, except for the special math functions, are overloaded for simd.

Example

Run this code
#include <experimental/simd>
#include <iostream>
namespace stdx = std::experimental;
 
void print(auto const& a)
{
    for (std::size_t i{}; i != std::size(a); ++i)
        std::cout << a[i] << ' ';
    std::cout << '\n';
}
 
template<class A>
stdx::simd<int, A> my_abs(stdx::simd<int, A> x)
{
    where(x < 0, x) = -x;
    return x;
}
 
int main()
{
    const stdx::native_simd<int> a = 1;
    print(a);
    const stdx::native_simd<int> b([](int i) { return i - 2; });
    print(b);
    const auto c = a + b;
    print(c);
    const auto d = my_abs(c);
    print(d);
    const auto e = d * d;
    print(e);
    const auto inner_product = stdx::reduce(e);
    std::cout << inner_product << '\n';
}

Output: 

1 1 1 1 
-2 -1 0 1 
-1 0 1 2 
1 0 1 2 
1 0 1 4 
6
Retrieved from "https://en.cppreference.com/mwiki/index.php?title=cpp/experimental/simd&oldid=155812"