From fd627ffaec8fd8801d980b4c91ee7c0607ab6aaf Mon Sep 17 00:00:00 2001 From: Jan Eilers Date: Thu, 25 Feb 2021 17:44:00 +0000 Subject: IVGCVSW-5687 Update Doxygen Docu * Update Doxygen Documentation for 21.02 release Signed-off-by: Jan Eilers Change-Id: I9ed2f9caab038836ea99d7b378d7899fe431a4e5 --- ...convert_bf16_to_fp32_workload_8cpp_source.xhtml | 130 +++++++++++++++++++++ 1 file changed, 130 insertions(+) create mode 100644 21.02/_neon_convert_bf16_to_fp32_workload_8cpp_source.xhtml (limited to '21.02/_neon_convert_bf16_to_fp32_workload_8cpp_source.xhtml') diff --git a/21.02/_neon_convert_bf16_to_fp32_workload_8cpp_source.xhtml b/21.02/_neon_convert_bf16_to_fp32_workload_8cpp_source.xhtml new file mode 100644 index 0000000000..430c6e9e07 --- /dev/null +++ b/21.02/_neon_convert_bf16_to_fp32_workload_8cpp_source.xhtml @@ -0,0 +1,130 @@ + + + + + + + + + + + + + +ArmNN: src/backends/neon/workloads/NeonConvertBf16ToFp32Workload.cpp Source File + + + + + + + + + + + + + + + + +
+
+ + + + ArmNN + + + +
+
+  21.02 +
+
+
+ + + + + + + +
+
+ +
+
+
+ +
+ +
+
+ + +
+ +
+ +
+
+
NeonConvertBf16ToFp32Workload.cpp
+
+
+Go to the documentation of this file.
1 //
2 // Copyright © 2020 Arm Ltd. All rights reserved.
3 // SPDX-License-Identifier: MIT
4 //
5 
7 
9 
10 #include <BFloat16.hpp>
11 
13 
14 namespace armnn
15 {
16 
18  const WorkloadInfo& info)
20 {
21  this->m_Data.ValidateInputsOutputs("NeonConvertBf16ToFp32Workload", 1, 1);
22  GatherTensorHandlePairs(descriptor, m_TensorHandlePairs);
23 }
24 
26 {
27  ARMNN_SCOPED_PROFILING_EVENT_NEON("NeonConvertBf16ToFp32Workload_Execute");
28 
29  auto convertFunc = [](uint8_t* dst, const uint8_t* src, size_t size)
30  {
31  auto input = reinterpret_cast<const BFloat16*>(src);
32  auto output = reinterpret_cast<float*>(dst);
33  size_t numElements = size/2; // 2 bytes per Bf16
35  };
36 
37  for (const auto& pair : m_TensorHandlePairs)
38  {
39  CopyTensorContentsGeneric(pair.first, pair.second, convertFunc);
40  }
41 }
42 
43 } //namespace armnn
+ +
static void ConvertBFloat16ToFloat32(const void *srcBFloat16Buffer, size_t numElements, float *dstFloat32Buffer)
+ + +
NeonConvertBf16ToFp32Workload(const ConvertBf16ToFp32QueueDescriptor &descriptor, const WorkloadInfo &info)
+
const QueueDescriptor m_Data
Definition: Workload.hpp:46
+
#define ARMNN_SCOPED_PROFILING_EVENT_NEON(name)
+
void ValidateInputsOutputs(const std::string &descName, unsigned int numExpectedIn, unsigned int numExpectedOut) const
+
Copyright (c) 2021 ARM Limited and Contributors.
+ + + +
Contains information about inputs and outputs to a layer.
+
void CopyTensorContentsGeneric(const ITensorHandle *srcTensor, ITensorHandle *dstTensor, CopyFunc copy)
+
void GatherTensorHandlePairs(const DescriptorType &descriptor, std::vector< std::pair< SrcTensorHandleType *, DstTensorHandleType *>> &tensorHandlePairs)
+ +
+
+ + + + -- cgit v1.2.1