FloatingPointConverter Class Reference

#include <FloatingPointConverter.hpp>

Static Public Member Functions
static void	ConvertFloat32To16 (const float *srcFloat32Buffer, size_t numElements, void *dstFloat16Buffer)
	Converts a buffer of FP32 values to FP16, and stores in the given dstFloat16Buffer. More...
static void	ConvertFloat16To32 (const void *srcFloat16Buffer, size_t numElements, float *dstFloat32Buffer)
static void	ConvertFloat32ToBFloat16 (const float *srcFloat32Buffer, size_t numElements, void *dstBFloat16Buffer)
static void	ConvertBFloat16ToFloat32 (const void *srcBFloat16Buffer, size_t numElements, float *dstFloat32Buffer)

Detailed Description

Definition at line 13 of file FloatingPointConverter.hpp.

Member Function Documentation

ConvertBFloat16ToFloat32()

static void ConvertBFloat16ToFloat32 ( const void *  srcBFloat16Buffer, size_t  numElements, float *  dstFloat32Buffer  )

static

ConvertFloat16To32()

void ConvertFloat16To32 ( const void *  srcFloat16Buffer, size_t  numElements, float *  dstFloat32Buffer  )

static

Definition at line 31 of file FloatingPointConverter.cpp.

{
    ARMNN_ASSERT(srcFloat16Buffer != nullptr);
    ARMNN_ASSERT(dstFloat32Buffer != nullptr);
 
    const armnn::Half* pHalf = static_cast<const armnn::Half*>(srcFloat16Buffer);
 
    for (size_t i = 0; i < numElements; i++)
    {
        dstFloat32Buffer[i] = pHalf[i];
    }
}

References ARMNN_ASSERT.

Referenced by armnn::AttemptBackendAssignment(), Float16Decoder::DecodeTensor(), NeonConvertFp16ToFp32Workload::Execute(), Float16ToFloat32::Func(), Float16Decoder::Get(), and Float16Encoder::Get().

ConvertFloat32To16()

void ConvertFloat32To16 ( const float *  srcFloat32Buffer, size_t  numElements, void *  dstFloat16Buffer  )

static

Converts a buffer of FP32 values to FP16, and stores in the given dstFloat16Buffer.

dstFloat16Buffer should be (numElements * 2) in size

Definition at line 16 of file FloatingPointConverter.cpp.

{
    ARMNN_ASSERT(srcFloat32Buffer != nullptr);
    ARMNN_ASSERT(dstFloat16Buffer != nullptr);
 
    armnn::Half* pHalf = static_cast<armnn::Half*>(dstFloat16Buffer);
 
    for (size_t i = 0; i < numElements; i++)
    {
        pHalf[i] = armnn::Half(srcFloat32Buffer[i]);
    }
}

References ARMNN_ASSERT.

Referenced by NeonConvertFp32ToFp16Workload::Execute(), Float32ToFloat16::Func(), and Float16Encoder::Set().

ConvertFloat32ToBFloat16()

static void ConvertFloat32ToBFloat16 ( const float *  srcFloat32Buffer, size_t  numElements, void *  dstBFloat16Buffer  )

static

---

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

• include/armnnUtils/FloatingPointConverter.hpp
• src/armnnUtils/FloatingPointConverter.cpp