 |
AI Engine API User Guide (AIE-API) 2025.2
|
Loading...
Searching...
No Matches
Reduction
Overview
Vector reduction operations.
Functions | |
| template<ElemBaseType T, Accum Acc> requires (Acc::value_class() == detail::accum_class_for_type_v<T>) | |
| T | aie::reduce_add (const Acc &acc) |
| Returns the sum of all input vector elements. | |
| template<Vector Vec> | |
| Vec::value_type | aie::reduce_add (const Vec &v) |
| Returns the sum of all input vector elements. | |
| template<RealVector Vec> | |
| Vec::value_type | aie::reduce_max (const Vec &v) |
| Returns the element from the input vector with the largest value. | |
| template<RealVector Vec> | |
| Vec::value_type | aie::reduce_min (const Vec &v) |
| Returns the element from the input vector with the smallest value. | |
| template<Vector Vec> | |
| constexpr Vec::value_type | aie::reduce_mul (const Vec &v) |
| Returns a product of the elements in the input vector. | |
Function Documentation
◆ reduce_add() [1/2]
template<ElemBaseType T, Accum Acc>
requires (Acc::value_class() == detail::accum_class_for_type_v<T>)
requires (Acc::value_class() == detail::accum_class_for_type_v<T>)
| T aie::reduce_add | ( | const Acc & | acc | ) |
Returns the sum of all input vector elements.
This overload is useful when the result must be converted to a lower precision format and the input values are already available in an accumulator (for example, as a result of a previous operation such as a multiplication).
- Parameters
-
acc Input accumulator. The type must meet aie::Accum.
- Template Parameters
-
Acc Input accumulator type. T Result type after reduction. The type must be compatible with the input accumulator.
Example:
aie::accum<accfloat, 16> acc = ...;
bfloat16 result = aie::reduce_add<bfloat16>(acc);
Vec::value_type reduce_add(const Vec &v)
Returns the sum of all input vector elements.
Definition aie.hpp:4220
bfloat16
Definition types.hpp:88
◆ reduce_add() [2/2]
template<Vector Vec>
| Vec::value_type aie::reduce_add | ( | const Vec & | v | ) |
Returns the sum of all input vector elements.
Higher precision intermediate values may be used to perform this reduction.
- Parameters
-
v Input vector. The type must meet aie::Vector.
◆ reduce_max()
template<RealVector Vec>
| Vec::value_type aie::reduce_max | ( | const Vec & | v | ) |
Returns the element from the input vector with the largest value.
- Parameters
-
v Input vector. The type must meet aie::RealVector.
◆ reduce_min()
template<RealVector Vec>
| Vec::value_type aie::reduce_min | ( | const Vec & | v | ) |
Returns the element from the input vector with the smallest value.
- Parameters
-
v Input vector. The type must meet aie::RealVector.
◆ reduce_mul()
template<Vector Vec>
|
constexpr |
Returns a product of the elements in the input vector.
- Parameters
-
v Input vector. The type must meet aie::Vector.