blob: ac6a69d21aa0cab570d37a6380f9139b8000469e (
plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
|
//
// Copyright © 2020 Arm Ltd and Contributors. All rights reserved.
// SPDX-License-Identifier: MIT
//
#include "NeonConvertFp32ToBf16Workload.hpp"
#include <BFloat16.hpp>
#include <Profiling.hpp>
#include <armnnUtils/FloatingPointConverter.hpp>
#include <backendsCommon/WorkloadUtils.hpp>
namespace armnn
{
NeonConvertFp32ToBf16Workload::NeonConvertFp32ToBf16Workload(const ConvertFp32ToBf16QueueDescriptor& descriptor,
const WorkloadInfo& info)
: Float32ToBFloat16Workload<ConvertFp32ToBf16QueueDescriptor>(descriptor, info)
{
this->m_Data.ValidateInputsOutputs("NeonConvertFp32ToBf16Workload", 1, 1);
GatherTensorHandlePairs(descriptor, m_TensorHandlePairs);
}
void NeonConvertFp32ToBf16Workload::Execute() const
{
ARMNN_SCOPED_PROFILING_EVENT_NEON_GUID("NeonConvertFp32ToBf16Workload_Execute", this->GetGuid());
auto convertFunc = [](uint8_t* dst, const uint8_t* src, size_t size)
{
auto input = reinterpret_cast<const float*>(src);
auto output = reinterpret_cast<BFloat16*>(dst);
size_t numElements = size/2; // 2 bytes per bf16
armnnUtils::FloatingPointConverter::ConvertFloat32ToBFloat16(input, numElements, output);
};
for (const auto& pair : m_TensorHandlePairs)
{
CopyTensorContentsGeneric(pair.first, pair.second, convertFunc);
}
}
} //namespace armnn
|
