Primitives

map

template<typename F, typename ...Args>
inline map_type<F, Args...> kernel_float::map(F fun, const Args&... args)

Apply the function F to each element from the vector input and return the results as a new vector.

Examples

vec<float, 4> input = {1.0f, 2.0f, 3.0f, 4.0f};
vec<float, 4> squared = map([](auto x) { return x * x; }, input); // [1.0f, 4.0f, 9.0f, 16.0f]

reduce

template<typename F, typename V>
inline vector_value_type<V> kernel_float::reduce(F fun, const V &input)

Reduce the elements of the given vector input into a single value using the function fun. This function should be a binary function that takes two elements and returns one element. The order in which the elements are reduced is not specified and depends on both the reduction function and the vector type.

Example

vec<int, 3> x = {5, 2, 1};
int y = reduce(x, [](int a, int b) { return a + b; }); // returns 5+2+1=8

zip

template<typename F, typename L, typename R>
inline zip_type<F, L, R> kernel_float::zip(F fun, const L &left, const R &right)

Combines the elements from the two inputs (left and right) element-wise, applying a provided binary function (fun) to each pair of corresponding elements.

Example

vec<bool, 3> make_negative = {true, false, true};
vec<int, 3> input = {1, 2, 3};
vec<int, 3> output = zip([](bool b, int n){ return b ? -n : +n; }, make_negative, input); // returns [-1, 2, -3]

zip_common

template<typename F, typename L, typename R>
inline zip_common_type<F, L, R> kernel_float::zip_common(F fun, const L &left, const R &right)

Combines the elements from the two inputs (left and right) element-wise, applying a provided binary function (fun) to each pair of corresponding elements. The elements are promoted to a common type before applying the binary function.

Example

vec<float, 3> a = {1.0f, 2.0f, 3.0f};
vec<int, 3> b = {4, 5, 6};
vec<float, 3> c = zip_common([](float x, float y){ return x + y; }, a, b); // returns [5.0f, 7.0f, 9.0f]

make_vec

template<typename ...Args>
inline vec<promote_t<Args...>, sizeof...(Args)> kernel_float::make_vec(Args&&... args)

Create a vector from a variable number of input values.

The resulting vector type is determined by promoting the types of the input values into a common type. The number of input values determines the dimension of the resulting vector.

Example

auto v1 = make_vec(1.0f, 2.0f, 3.0f); // Creates a vec<float, 3> [1.0f, 2.0f, 3.0f]
auto v2 = make_vec(1, 2, 3, 4);       // Creates a vec<int, 4> [1, 2, 3, 4]

into_vec

template<typename V>
inline into_vector_type<V> kernel_float::into_vec(V &&input)

Convert the given input into a vector. This function can perform one of the following actions:

• For vectors vec<T, N>, it simply returns the original vector.

• For primitive types T (e.g., int, float, double), it returns a vec<T, 1>.

• For array-like types (e.g., std::array<T, N>, T[N]), it returns vec<T, N>.

• For vector-like types (e.g., int2, dim3), it returns vec<T, N>.

concat

template<typename ...Vs>
inline concat_type<Vs...> kernel_float::concat(const Vs&... inputs)

Concatenates the provided input values into a single one-dimensional vector.

This function works in three steps:

• All input values are converted into vectors using the into_vector operation.

• The resulting vectors’ elements are then promoted into a shared value type.

• The resultant vectors are finally concatenated together.

For instance, when invoking this function with arguments of types float, double2, double:

• After the first step: vec<float, 1>, vec<double, 2>, vec<double, 1>

• After the second step: vec<double, 1>, vec<double, 2>, vec<double, 1>

• After the third step: vec<double, 4>

Example

double vec1 = 1.0;
double3 vec2 = {2.0, 3.0, 4.0);
double4 vec3 = {5.0, 6.0, 7.0, 8.0};
vec<double, 8> concatenated = concat(vec1, vec2, vec3); // contains [1, 2, 3, 4, 5, 6, 7, 8]

int num1 = 42;
float num2 = 3.14159;
int2 num3 = {-10, 10};
vec<float, 4> concatenated = concat(num1, num2, num3); // contains [42, 3.14159, -10, 10]

select

template<typename V, typename ...Is>
inline select_type<V, Is...> kernel_float::select(const V &input, const Is&... indices)

Selects elements from the this vector based on the specified indices.

Example

vec<float, 6> input = {0, 10, 20, 30, 40, 50};
vec<float, 4> vec1 = select(input, 0, 4, 4, 2); // [0, 40, 40, 20]

vec<int, 4> indices = {0, 4, 4, 2};
vec<float, 4> vec2 = select(input, indices); // [0, 40, 40, 20]

for_each

template<typename V, typename F>
inline void kernel_float::for_each(V &&input, F fun)

Apply the function fun for each element from input.

Example

for_each(range<int, 3>(), [&](auto i) {
   printf("element: %d\n", i);
});