telsoa01 | c577f2c | 2018-08-31 09:22:23 +0100 | [diff] [blame] | 1 | // |
| 2 | // Copyright © 2017 Arm Ltd. All rights reserved. |
David Beck | ecb56cd | 2018-09-05 12:52:57 +0100 | [diff] [blame] | 3 | // SPDX-License-Identifier: MIT |
telsoa01 | c577f2c | 2018-08-31 09:22:23 +0100 | [diff] [blame] | 4 | // |
| 5 | |
| 6 | #include "NeonConvertFp16ToFp32Workload.hpp" |
Matteo Martincigh | e011d20 | 2019-11-28 11:35:47 +0000 | [diff] [blame] | 7 | |
| 8 | #include <armnnUtils/FloatingPointConverter.hpp> |
telsoa01 | c577f2c | 2018-08-31 09:22:23 +0100 | [diff] [blame] | 9 | |
Aron Virginas-Tar | c9cc804 | 2018-11-01 16:15:57 +0000 | [diff] [blame] | 10 | #include <Half.hpp> |
Matteo Martincigh | e011d20 | 2019-11-28 11:35:47 +0000 | [diff] [blame] | 11 | |
Aron Virginas-Tar | c9cc804 | 2018-11-01 16:15:57 +0000 | [diff] [blame] | 12 | #include <backendsCommon/WorkloadUtils.hpp> |
telsoa01 | c577f2c | 2018-08-31 09:22:23 +0100 | [diff] [blame] | 13 | |
| 14 | namespace armnn |
| 15 | { |
| 16 | |
| 17 | NeonConvertFp16ToFp32Workload::NeonConvertFp16ToFp32Workload(const ConvertFp16ToFp32QueueDescriptor& descriptor, |
| 18 | const WorkloadInfo& info) |
| 19 | : Float16ToFloat32Workload<ConvertFp16ToFp32QueueDescriptor>(descriptor, info) |
| 20 | { |
| 21 | this->m_Data.ValidateInputsOutputs("NeonConvertFp16ToFp32Workload", 1, 1); |
| 22 | GatherTensorHandlePairs(descriptor, m_TensorHandlePairs); |
| 23 | } |
| 24 | |
| 25 | void NeonConvertFp16ToFp32Workload::Execute() const |
| 26 | { |
| 27 | ARMNN_SCOPED_PROFILING_EVENT_NEON("NeonConvertFp16ToFp32Workload_Execute"); |
| 28 | |
| 29 | auto convertFunc = [](uint8_t* dst, const uint8_t* src, size_t size) |
| 30 | { |
| 31 | auto input = reinterpret_cast<const Half*>(src); |
| 32 | auto output = reinterpret_cast<float*>(dst); |
| 33 | size_t numElements = size/2; // 2 bytes per fp16 |
| 34 | armnnUtils::FloatingPointConverter::ConvertFloat16To32(input, numElements, output); |
| 35 | }; |
| 36 | |
| 37 | for (const auto& pair : m_TensorHandlePairs) |
| 38 | { |
| 39 | CopyTensorContentsGeneric(pair.first, pair.second, convertFunc); |
| 40 | } |
| 41 | } |
| 42 | |
| 43 | } //namespace armnn |