* Data parallel devices such as GPUs, SIMD CPUs and other accelerators are vector processors. * This means they can execute vector instructions. * Vector instructions are single instructions which perform loads, stores, or operations such as add or multiply on multiple elements at once.

* Vectorization is the process of converting scalar code into vectorized code. * In a SPMD programming model like SYCL vectorization is important. * Vectorization can be performed in two ways, and it depends on how you write your code and can impact the mapping to hardware.

* Both horizontal and vertical vectorization generally achieve the same result. * It can be useful to specify vectorization explicitly, particularly for describing aligned loads and stores. * An important distinction to make is that whether a kernel function uses explicit vector types can impact the mapping of work-items to processing elements. * It's not always a 1:1 mapping.

template <typename dataT, int numElements>
class vec;
						

* The `vec` class template is used to represent explicit vectors in SYCL. * It has a type which represents the type of elements it stores and a number of elements. * The valid number of elements are 1, 2, 3, 4, 8, 16. * Note that vectors of 3 elements are padded to the size of 4.

using float4 = vec<float, 4>;
...
						

* A number of aliases are provided for shorthand with the notation of the type followed by the size, such as `float4`.

auto f4 = sycl::float4{1.0f, 2.0f, 3.0f, 4.0f}; // {1.0f, 2.0f, 3.0f, 4.0f}
						

auto f2 = sycl::float4{2.0f, 3.0f}; // {2.0f, 3.0f}
auto f4 = sycl::float4{1.0f, f2, 4.0f}; // {1.0f, 2.0f, 3.0f, 4.0f}
						

auto f4 = sycl::float4{0.0f};  // {0.0f, 0.0f, 0.0f, 0.0f}
						

* A `vec` can be constructed with any combination of scalar and vector values which add up to the correct number of elements. * A `vec`can also be constructed from a single scalar in which case it will initialize ever element to that value.

auto f4a = sycl::float4{1.0f, 2.0f, 3.0f, 4.0f}; // {1.0f, 2.0f, 3.0f, 4.0f}

auto f4b = sycl::float4{2.0f}; // {2.0f, 2.0f, 2.0f, 2.0f}

auto f4r = f4a * f4b; // {2.0f, 4.0f, 6.0f, 8.0f}
						

* The `vec` class provides a number of operators such as `+`, `-`, `*`, `/` and many more, which perform the operation elemeent-wise.

auto f4 = sycl::float4{1.0f, 2.0f, 3.0f, 4.0f}; // {1.0f, 2.0f, 3.0f, 4.0f}
auto f2 = f4.swizzle<0, 3>(); // {1.0f, 4.0f}
						

auto f4 = sycl::float4{1.0f, 2.0f, 3.0f, 4.0f}; // {1.0f, 2.0f, 3.0f, 4.0f}
f4.swizzle<1, 2>() = sycl::float2{9.0f, 9.0f}; // f4 becomes {1.0f, 9.0f, 9.0f, 4.0f}
						

* The `swizzle` function returns a representation of the specified elements of a `vec` which can be used on the lhs or rhs of an expression.

auto f4 = sycl::float4{1.0f, 2.0f, 3.0f, 4.0f}; // {1.0f, 2.0f, 3.0f, 4.0f}
auto f2 = f4.xw(); // {1.0f, 4.0f}
						

auto f4 = sycl::float4{1.0f, 2.0f, 3.0f, 4.0f}; // {1.0f, 2.0f, 3.0f, 4.0f}
f4.yz() = sycl::float2{9.0f, 9.0f}; // f4 becomes {1.0f, 9.0f, 9.0f, 4.0f}
						

* If `SYCL_SIMPLE_SWIZZLES` is defined before including `sycl/sycl.hpp` simplified swizzle member functions can also be used in place of `swizzle`.


Code_Exercises/Vectors/source

Update the image convolution application to use vectors types.