BFloat16Decoder Class Reference

#include <BaseIterator.hpp>

Inheritance diagram for BFloat16Decoder:

Public Member Functions
	BFloat16Decoder (const BFloat16 *data)
	BFloat16Decoder ()
float	Get () const override
std::vector< float >	DecodeTensor (const TensorShape &tensorShape, const unsigned int channelMultiplier, const bool isDepthwise) override
Public Member Functions inherited from TypedIterator< const BFloat16, Decoder< float > >
	TypedIterator (const BFloat16 *data=nullptr)
void	Reset (void *data) override
TypedIterator &	operator++ () override
TypedIterator &	operator+= (const unsigned int increment) override
TypedIterator &	operator-= (const unsigned int increment) override
TypedIterator &	operator[] (const unsigned int index) override
TypedIterator &	SetIndex (unsigned int index, unsigned int axisIndex=0) override
Public Member Functions inherited from Decoder< float >
	Decoder ()
virtual	~Decoder ()
Public Member Functions inherited from BaseIterator
	BaseIterator ()
virtual	~BaseIterator ()

Additional Inherited Members
Protected Attributes inherited from TypedIterator< const BFloat16, Decoder< float > >
const BFloat16 *	m_Iterator
const BFloat16 *	m_Start

Detailed Description

Definition at line 276 of file BaseIterator.hpp.

Constructor & Destructor Documentation

BFloat16Decoder() [1/2]

BFloat16Decoder ( const BFloat16 *  data )

inline

Definition at line 279 of file BaseIterator.hpp.

        : TypedIterator(data) {}

BFloat16Decoder() [2/2]

BFloat16Decoder ( )

inline

Definition at line 282 of file BaseIterator.hpp.

        : BFloat16Decoder(nullptr) {}

Member Function Documentation

DecodeTensor()

std::vector<float> DecodeTensor ( const TensorShape &  tensorShape, const unsigned int  channelMultiplier, const bool  isDepthwise  )

inlineoverridevirtual

Implements Decoder< float >.

Definition at line 291 of file BaseIterator.hpp.

References FloatingPointConverter::ConvertBFloat16ToFloat32(), TensorShape::GetNumElements(), armnn::IgnoreUnused(), and BaseIterator::operator[]().

    {
        IgnoreUnused(channelMultiplier, isDepthwise);

        const unsigned int size = tensorShape.GetNumElements();
        std::vector<float> decodedTensor;
        decodedTensor.reserve(size);

        for (uint32_t i = 0; i < size; ++i)
        {
            this->operator[](i);

            float val = 0.f;
            armnnUtils::FloatingPointConverter::ConvertBFloat16ToFloat32(m_Iterator, 1, &val);
            decodedTensor.emplace_back(val);
        }

        return decodedTensor;
    }

Get()

float Get ( ) const

inlineoverridevirtual

Implements Decoder< float >.

Definition at line 285 of file BaseIterator.hpp.

References FloatingPointConverter::ConvertBFloat16ToFloat32().

    {
        float val = 0.f;
        armnnUtils::FloatingPointConverter::ConvertBFloat16ToFloat32(m_Iterator, 1, &val);
        return val;
    }

---

The documentation for this class was generated from the following file:

• src/backends/reference/workloads/BaseIterator.hpp