///
/// Copyright (c) 2021-2022 Arm Limited.
///
/// SPDX-License-Identifier: MIT
///
/// Permission is hereby granted, free of charge, to any person obtaining a copy
/// of this software and associated documentation files (the "Software"), to
/// deal in the Software without restriction, including without limitation the
/// rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
/// sell copies of the Software, and to permit persons to whom the Software is
/// furnished to do so, subject to the following conditions:
///
/// The above copyright notice and this permission notice shall be included in all
/// copies or substantial portions of the Software.
///
/// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
/// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
/// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
/// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
/// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
/// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
/// SOFTWARE.
///
namespace arm_compute
{
/**
@page operators_list Supported Operators

@tableofcontents

@section S9_1_operators_list Supported Operators

Compute Library supports operators that are listed in below table.

Compute Library supports a wide list of data-types, information can been directly found in the documentation of each kernel/function.
The main data-types that the Machine Learning functions support are the following:
  <ul>
    <li>BFLOAT16: 16-bit non-standard brain floating point
    <li>QASYMM8: 8-bit unsigned asymmetric quantized
    <li>QASYMM8_SIGNED: 8-bit signed asymmetric quantized
    <li>QSYMM8_PER_CHANNEL: 8-bit signed symmetric quantized (Used for the weights)
    <li>QSYMM8: 8-bit unsigned symmetric quantized
    <li>QSYMM16: 16-bit unsigned symmetric quantized
    <li>F32: 32-bit single precision floating point
    <li>F16: 16-bit half precision floating point
    <li>S32: 32-bit signed integer
    <li>U8: 8-bit unsigned char
    <li>All: Agnostic to any specific data type
  </ul>

Compute Library supports the following data layouts (fast changing dimension from right to left):
  <ul>
    <li>NHWC: The native layout of Compute Library that delivers the best performance where channels are in the fastest changing dimension
    <li>NCHW: Legacy layout where width is in the fastest changing dimension
    <li>NDHWC: New data layout for supporting 3D operators
    <li>All: Agnostic to any specific data layout
  </ul>
where N = batches, C = channels, H = height, W = width, D = depth

<table>
<caption id="multi_row"></caption>
<tr>
  <th>Function
  <th>Description
  <th>Equivalent Android NNAPI Op
  <th>Backends
  <th>Data Layouts
  <th>Data Types
<tr>
  <td rowspan="2">ActivationLayer
  <td rowspan="2" style="width:200px;"> Function to simulate an activation layer with the specified activation function.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_ELU
       <li>ANEURALNETWORKS_HARD_SWISH
       <li>ANEURALNETWORKS_LOGISTIC
       <li>ANEURALNETWORKS_RELU
       <li>ANEURALNETWORKS_RELU1
       <li>ANEURALNETWORKS_RELU6
       <li>ANEURALNETWORKS_TANH
      </ul>
  <td>NEActivationLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>QASYMM8<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED
    <tr><td>QSYMM16<td>QSYMM16
    <tr><td>F16<td>F16
    <tr><td>F32<td>F32
    </table>
<tr>
  <td>CLActivationLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>QASYMM8<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED
    <tr><td>QSYMM16<td>QSYMM16
    <tr><td>F16<td>F16
    <tr><td>F32<td>F32
    </table>
<tr>
  <td rowspan="2">ArgMinMaxLayer
  <td rowspan="2" style="width:200px;"> Function to calculate the index of the minimum or maximum values in a tensor based on an axis.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_ARGMAX
       <li>ANEURALNETWORKS_ARGMIN
      </ul>
  <td>NEArgMinMaxLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>QASYMM8<td>U32, S32
    <tr><td>QASYMM8_SIGNED<td>U32, S32
    <tr><td>S32<td>U32, S32
    <tr><td>F16<td>U32, S32
    <tr><td>F32<td>U32, S32
    </table>
<tr>
  <td>CLArgMinMaxLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>QASYMM8<td>U32, S32
    <tr><td>QASYMM8_SIGNED<td>U32, S32
    <tr><td>S32<td>U32, S32
    <tr><td>F16<td>U32, S32
    <tr><td>F32<td>U32, S32
    </table>
<tr>
  <td rowspan="1">ArithmeticAddition
  <td rowspan="1" style="width:200px;"> Function to add 2 tensors.
  <td rowspan="1">
      <ul>
       <li>ANEURALNETWORKS_ADD
      </ul>
  <td>NEArithmeticAddition
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>dst
    <tr><td>QASYMM8<td>QASYMM8<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED<td>QASYMM8_SIGNED
    <tr><td>QSYMM16<td>QSYMM16<td>QASYMM16
    <tr><td>QSYMM16<td>QSYMM16<td>S32
    <tr><td>U8<td>U8<td>U8
    <tr><td>S16<td>S16<td>S16
    <tr><td>S32<td>S32<td>S32
    <tr><td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32
    </table>
<tr>
  <td rowspan="1">ArithmeticSubtraction
  <td rowspan="1" style="width:200px;"> Function to substract 2 tensors.
  <td rowspan="1">
      <ul>
       <li>ANEURALNETWORKS_SUB
      </ul>
  <td>NEArithmeticSubtraction
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>dst
    <tr><td>QASYMM8<td>QASYMM8<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED<td>QASYMM8_SIGNED
    <tr><td>QSYMM16<td>QSYMM16<td>QASYMM16
    <tr><td>QSYMM16<td>QSYMM16<td>S32
    <tr><td>U8<td>U8<td>U8
    <tr><td>S16<td>S16<td>S16
    <tr><td>S32<td>S32<td>S32
    <tr><td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32
    </table>
<tr>
  <td rowspan="2">BatchNormalizationLayer
  <td rowspan="2" style="width:200px;"> Function to perform batch normalization.
  <td rowspan="2">
      <ul>
       <li>n/a
      </ul>
  <td>NEBatchNormalizationLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>F32<td>F32
    <tr><td>F16<td>F16
    </table>
<tr>
  <td>CLBatchNormalizationLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>F32<td>F32
    <tr><td>F16<td>F16
    </table>
<tr>
  <td rowspan="2">BatchToSpaceLayer
  <td rowspan="2" style="width:200px;"> Batch to space transformation.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_BATCH_TO_SPACE_ND
      </ul>
  <td>NEBatchToSpaceLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>dst
    <tr><td>All<td>s32<td>All
    </table>
<tr>
  <td>CLBatchToSpaceLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>dst
    <tr><td>All<td>s32<td>All
    </table>
<tr>
  <td rowspan="2">BitwiseAnd
  <td rowspan="2" style="width:200px;"> Function to perform bitwise AND between 2 tensors.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_LOGICAL_AND
      </ul>
  <td>NEBitwiseAnd
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>U8<td>U8
    </table>
<tr>
  <td>CLBitwiseAnd
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>U8<td>U8
    </table>
<tr>
  <td rowspan="2">BitwiseNot
  <td rowspan="2" style="width:200px;"> Function to perform bitwise NOT.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_LOGICAL_NOT
      </ul>
  <td>NEBitwiseNot
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>U8<td>U8
    </table>
<tr>
  <td>CLBitwiseNot
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>U8<td>U8
    </table>
<tr>
  <td rowspan="2">BitwiseOr
  <td rowspan="2" style="width:200px;"> Function to perform bitwise OR between 2 tensors.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_LOGICAL_OR
      </ul>
  <td>NEBitwiseOr
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>U8<td>U8
    </table>
<tr>
  <td>CLBitwiseOr
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>U8<td>U8
    </table>
<tr>
  <td rowspan="2">BitwiseXor
  <td rowspan="2" style="width:200px;"> Function to perform bitwise XOR between 2 tensors.
  <td rowspan="2">
      <ul>
       <li>n/a
      </ul>
  <td>NEBitwiseXor
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>U8<td>U8
    </table>
<tr>
  <td>CLBitwiseXor
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>U8<td>U8
    </table>
<tr>
  <td rowspan="2">BoundingBoxTransform
  <td rowspan="2" style="width:200px;"> Transform proposal bounding boxes to target bounding box using bounding box deltas.
  <td rowspan="2">
      <ul>
       <li>n/a
      </ul>
  <td>NEBoundingBoxTransform
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>dst
    <tr><td>QASYMM16<td>QASYMM8<td>QASYMM16
    <tr><td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32
    </table>
<tr>
  <td>CLBoundingBoxTransform
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>dst
    <tr><td>QASYMM16<td>QASYMM8<td>QASYMM16
    <tr><td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32
    </table>
<tr>
  <td rowspan="2">Cast
  <td rowspan="2" style="width:200px;"> Function to cast a tensor.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_CAST
      </ul>
  <td>NECast
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>QASYMM8_SIGNED<td>S16, S32, F32, F16
    <tr><td>QASYMM8<td>U16, S16, S32, F32, F16
    <tr><td>U8<td>U16, S16, S32, F32, F16
    <tr><td>U16<td>U8, U32
    <tr><td>S16<td>QASYMM8_SIGNED, U8, S32
    <tr><td>F16<td>QASYMM8_SIGNED, QASYMM8, F32, S32, U8
    <tr><td>S32<td>QASYMM8_SIGNED, QASYMM8, F16, F32, U8
    <tr><td>F32<td>QASYMM8_SIGNED, QASYMM8, BFLOAT16, F16, S32, U8
    </table>
<tr>
  <td>CLCast
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>U8<td>S8, U16, S16, U32, S32, F16, F32
    <tr><td>U16<td>U8, S8, S16, U32, S32, F16, F32
    <tr><td>S16<td>U8, S8, U16, U32, S32, F16, F32
    <tr><td>U32<td>U8, S8, U16, S16, S32, F16, F32
    <tr><td>S32<td>U8, S8, U16, S16, U32, F16, F32
    <tr><td>F16<td>U8, S8, U16, S16, U32, F32
    <tr><td>F32<td>U8, S8, U16, S16, U32, F16
    </table>
<tr>
  <td rowspan="2">ChannelShuffleLayer
  <td rowspan="2" style="width:200px;"> Function to shuffle the channels of the input tensor.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_CHANNEL_SHUFFLE
      </ul>
  <td>NEChannelShuffleLayer
  <td>
      <ul>
       <li>NCHW
       <li>NHWC
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td>CLChannelShuffleLayer
  <td>
      <ul>
       <li>NCHW
       <li>NHWC
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td rowspan="1">Comparison
  <td rowspan="1" style="width:200px;"> Function to compare 2 tensors.
  <td rowspan="1">
      <ul>
       <li>ANEURALNETWORKS_EQUAL
       <li>ANEURALNETWORKS_GREATER
       <li>ANEURALNETWORKS_GREATER_EQUAL
       <li>ANEURALNETWORKS_LESS
       <li>ANEURALNETWORKS_LESS_EQUAL
       <li>ANEURALNETWORKS_NOT_EQUAL
      </ul>
  <td>CLComparison
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>dst
    <tr><td>All<td>All<td>U8
    </table>
<tr>
  <td rowspan="2">ConcatenateLayer
  <td rowspan="2" style="width:200px;"> Function to concatenate tensors along a given axis.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_CONCATENATION
      </ul>
  <td>NEConcatenateLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>QASYMM8<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED
    <tr><td>F16<td>F16
    <tr><td>F32<td>F32
    </table>
<tr>
  <td>CLConcatenateLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>QASYMM8<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED
    <tr><td>F16<td>F16
    <tr><td>F32<td>F32
    </table>
<tr>
  <td rowspan="2">ConvertFullyConnectedWeights
  <td rowspan="2" style="width:200px;"> Function to transpose the weights for the fully connected layer.
  <td rowspan="2">
      <ul>
       <li>n/a
      </ul>
  <td>NEConvertFullyConnectedWeights
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td>CLConvertFullyConnectedWeights
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td rowspan="2">ConvolutionLayer
  <td rowspan="2" style="width:200px;"> Function to compute a convolution layer.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_CONV_2D
      </ul>
  <td>NEConvolutionLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>src2<th>dst
    <tr><td>F16<td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32<td>F32
    <tr><td>QASYMM8<td>QASYMM8<td>S32<td>QASYMM8
    <tr><td>QASYMM8<td>QSYMM8_PER_CHANNEL<td>S32<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED<td>S32<td>QASYMM8_SIGNED
    <tr><td>QASYMM8_SIGNED<td>QSYMM8_PER_CHANNEL<td>S32<td>QASYMM8_SIGNED
    </table>
<tr>
  <td>CLConvolutionLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>src2<th>dst
    <tr><td>F16<td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32<td>F32
    <tr><td>QASYMM8<td>QASYMM8<td>S32<td>QASYMM8
    <tr><td>QASYMM8<td>QSYMM8_PER_CHANNEL<td>S32<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED<td>S32<td>QASYMM8_SIGNED
    <tr><td>QASYMM8_SIGNED<td>QSYMM8_PER_CHANNEL<td>S32<td>QASYMM8_SIGNED
    </table>
<tr>
  <td rowspan="2">Conv3D
  <td rowspan="2" style="width:200px;"> Function to compute a 3d convolution layer.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_CONV_3D
      </ul>
  <td>NEConv3D
  <td>
      <ul>
       <li>NDHWC
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>src2<th>dst
    <tr><td>F16<td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32<td>F32
    <tr><td>QASYMM8<td>QASYMM8<td>S32<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED<td>S32<td>QASYMM8_SIGNED
    </table>
<tr>
  <td>CLConv3D
  <td>
      <ul>
       <li>NDHWC
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>src2<th>dst
    <tr><td>F16<td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32<td>F32
    <tr><td>QASYMM8<td>QASYMM8<td>S32<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED<td>S32<td>QASYMM8_SIGNED
    </table>
<tr>
  <td rowspan="2">Copy
  <td rowspan="2" style="width:200px;"> Function to copy a tensor.
  <td rowspan="2">
      <ul>
       <li>n/a
      </ul>
  <td>NECopy
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td>CLCopy
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td rowspan="1">Crop
  <td rowspan="1" style="width:200px;"> Performs a copy of input tensor to the output tensor.
  <td rowspan="1">
      <ul>
       <li>n/a
      </ul>
  <td>CLCrop
  <td>
      <ul>
       <li>NHWC
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>F32
    </table>
<tr>
  <td rowspan="2">CropResize
  <td rowspan="2" style="width:200px;"> Function to perform cropping and resizing.
  <td rowspan="2">
      <ul>
       <li>n/a
      </ul>
  <td>NECropResize
  <td>
      <ul>
       <li>NHWC
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>src2<th>dst
    <tr><td>All<td>F32<td>F32<td>F32
    </table>
<tr>
  <td>CLCropResize
  <td>
      <ul>
       <li>NHWC
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>src2<th>dst
    <tr><td>All<td>F32<td>F32<td>F32
    </table>
<tr>
  <td rowspan="2">DeconvolutionLayer
  <td rowspan="2" style="width:200px;"> Function to compute a deconvolution or transpose convolution.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_TRANSPOSE_CONV_2D
      </ul>
  <td>NEDeconvolutionLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>src2<th>dst
    <tr><td>F16<td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32<td>F32
    <tr><td>QASYMM8<td>QASYMM8<td>S32<td>QASYMM8
    <tr><td>QASYMM8<td>QSYMM8_PER_CHANNEL<td>S32<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED<td>S32<td>QASYMM8_SIGNED
    <tr><td>QASYMM8_SIGNED<td>QSYMM8_PER_CHANNEL<td>S32<td>QASYMM8_SIGNED
    </table>
<tr>
  <td>CLDeconvolutionLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>src2<th>dst
    <tr><td>F16<td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32<td>F32
    <tr><td>QASYMM8<td>QASYMM8<td>S32<td>QASYMM8
    <tr><td>QASYMM8<td>QSYMM8_PER_CHANNEL<td>S32<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED<td>S32<td>QASYMM8_SIGNED
    <tr><td>QASYMM8_SIGNED<td>QSYMM8_PER_CHANNEL<td>S32<td>QASYMM8_SIGNED
    </table>
<tr>
  <td rowspan="1">DeconvolutionLayerUpsample
  <td rowspan="1" style="width:200px;"> Function to execute deconvolution upsample on OpenCL.
  <td rowspan="1">
      <ul>
       <li>ANEURALNETWORKS_TRANSPOSE_CONV_2D
      </ul>
  <td>CLDeconvolutionLayerUpsample
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td rowspan="2">DepthConvertLayer
  <td rowspan="2" style="width:200px;"> Performs a down-scaling depth conversion.
  <td rowspan="2">
      <ul>
       <li>n/a
      </ul>
  <td>NEDepthConvertLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>QASYMM8<td>F16, F32
    <tr><td>U8<td>U16, S16, S32
    <tr><td>U16<td>U8, U32
    <tr><td>S16<td>U8, S32
    <tr><td>BFLOAT16<td>F32
    <tr><td>F16<td>QASYMM8, F32
    <tr><td>F32<td>QASYMM8, F16, BFLOAT16
    </table>
<tr>
  <td>CLDepthConvertLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>U8<td>S8, U16, S16, U32, S32, F16, F32
    <tr><td>U16<td>U8, S8, S16, U32, S32, F16, F32
    <tr><td>S16<td>U8, S8, U16, U32, S32, F16, F32
    <tr><td>U32<td>U8, S8, U16, S16, S32, F16, F32
    <tr><td>S32<td>U8, S8, U16, S16, U32, F16, F32
    <tr><td>F16<td>U8, S8, U16, S16, U32, F32
    <tr><td>F32<td>U8, S8, U16, S16, U32, F16
    </table>
<tr>
  <td rowspan="2">DepthToSpaceLayer
  <td rowspan="2" style="width:200px;"> Depth to Space transformation.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_DEPTH_TO_SPACE
      </ul>
  <td>NEDepthToSpaceLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td>CLDepthToSpaceLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td rowspan="2">DepthwiseConvolutionLayer
  <td rowspan="2" style="width:200px;"> Function to perform depthwise separable convolution.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_DEPTHWISE_CONV_2D
      </ul>
  <td>NEDepthwiseConvolutionLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>src2<th>dst
    <tr><td>F16<td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32<td>F32
    <tr><td>QASYMM8<td>QASYMM8<td>S32<td>QASYMM8
    <tr><td>QASYMM8<td>QSYMM8_PER_CHANNEL<td>S32<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED<td>S32<td>QASYMM8_SIGNED
    <tr><td>QASYMM8_SIGNED<td>QSYMM8_PER_CHANNEL<td>S32<td>QASYMM8_SIGNED
    </table>
<tr>
  <td>CLDepthwiseConvolutionLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>src2<th>dst
    <tr><td>F16<td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32<td>F32
    <tr><td>QASYMM8<td>QASYMM8<td>S32<td>QASYMM8
    <tr><td>QASYMM8<td>QSYMM8_PER_CHANNEL<td>S32<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED<td>S32<td>QASYMM8_SIGNED
    <tr><td>QASYMM8_SIGNED<td>QSYMM8_PER_CHANNEL<td>S32<td>QASYMM8_SIGNED
    </table>
<tr>
  <td rowspan="2">DequantizationLayer
  <td rowspan="2" style="width:200px;"> Function to dequantize the values in a tensor.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_DEQUANTIZE
      </ul>
  <td>NEDequantizationLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>QASYMM8<td>F16, F32
    <tr><td>QASYMM8_SIGNED<td>F16, F32
    <tr><td>QSYMM8_PER_CHANNEL<td>F16, F32
    <tr><td>QSYMM8<td>F16, F32
    <tr><td>QSYMM16<td>F16, F32
    </table>
<tr>
  <td>CLDequantizationLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>QASYMM8<td>F16, F32
    <tr><td>QASYMM8_SIGNED<td>F16, F32
    <tr><td>QSYMM8_PER_CHANNEL<td>F16, F32
    <tr><td>QSYMM8<td>F16, F32
    <tr><td>QSYMM16<td>F16, F32
    </table>
<tr>
  <td rowspan="1">DetectionPostProcessLayer
  <td rowspan="1" style="width:200px;"> Function to generate the detection output based on center size encoded boxes, class prediction and anchors by doing non maximum suppression (NMS).
  <td rowspan="1">
      <ul>
       <li>ANEURALNETWORKS_DETECTION_POSTPROCESSING
      </ul>
  <td>NEDetectionPostProcessLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0 - src2<th>dst0 - dst3
    <tr><td>QASYMM8<td>F32
    <tr><td>QASYMM8_SIGNED<td>F32
    <tr><td>F32<td>F32
    </table>
<tr>
  <td rowspan="2">DirectConvolutionLayer
  <td rowspan="2" style="width:200px;"> Function to compute direct convolution.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_CONV_2D
      </ul>
  <td>NEDirectConvolutionLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>src2<th>dst
    <tr><td>F16<td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32<td>F32
    </table>
<tr>
  <td>CLDirectConvolutionLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>src2<th>dst
    <tr><td>F16<td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32<td>F32
    <tr><td>QASYMM8<td>QASYMM8<td>S32<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED<td>S32<td>QASYMM8_SIGNED
    </table>
<tr>
  <td rowspan="1">DirectDeconvolutionLayer
  <td rowspan="1" style="width:200px;"> Function to run the deconvolution layer.
  <td rowspan="1">
      <ul>
       <li>ANEURALNETWORKS_TRANSPOSE_CONV_2D
      </ul>
  <td>CLDirectDeconvolutionLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>src2<th>dst
    <tr><td>F16<td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32<td>F32
    <tr><td>QASYMM8<td>QASYMM8<td>S32<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED<td>S32<td>QASYMM8_SIGNED
    <tr><td>QASYMM8<td>QSYMM8_PER_CHANNEL<td>S32<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QSYMM8_PER_CHANNEL<td>S32<td>QASYMM8_SIGNED
    </table>
<tr>
  <td rowspan="13">ElementwiseOperations
  <td rowspan="13" style="width:200px;"> Function to perform in Cpu: - Div - Max - Min - Pow - SquaredDiff - Comparisons (Equal, greater, greater_equal, less, less_equal, not_equal) Function to perform in CL: - Add - Sub - Div - Max - Min - Pow - SquaredDiff
  <td rowspan="13">
      <ul>
       <li>ANEURALNETWORKS_MAXIMUM
       <li>ANEURALNETWORKS_MINIMUM
       <li>ANEURALNETWORKS_POW
       <li>ANEURALNETWORKS_DIV
       <li>ANEURALNETWORKS_ADD
       <li>ANEURALNETWORKS_SUB
       <li>ANEURALNETWORKS_EQUAL
       <li>ANEURALNETWORKS_GREATER
       <li>ANEURALNETWORKS_GREATER_EQUAL
       <li>ANEURALNETWORKS_LESS
       <li>ANEURALNETWORKS_LESS_EQUAL
       <li>ANEURALNETWORKS_NOT_EQUAL
      </ul>
  <td>NEElementwiseMax
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>dst
    <tr><td>QASYMM8<td>QASYMM8<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED<td>QASYMM8_SIGNED
    <tr><td>S32<td>S32<td>S32
    <tr><td>S16<td>S16<td>S16
    <tr><td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32
    </table>
<tr>
  <td>NEElementwiseMin
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>dst
    <tr><td>QASYMM8<td>QASYMM8<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED<td>QASYMM8_SIGNED
    <tr><td>S32<td>S32<td>S32
    <tr><td>S16<td>S16<td>S16
    <tr><td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32
    </table>
<tr>
  <td>NEElementwiseSquaredDiff
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>dst
    <tr><td>QASYMM8<td>QASYMM8<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED<td>QASYMM8_SIGNED
    <tr><td>S32<td>S32<td>S32
    <tr><td>S16<td>S16<td>S16
    <tr><td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32
    </table>
<tr>
  <td>NEElementwiseDivision
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>dst
    <tr><td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32
    </table>
<tr>
  <td>NEElementwisePower
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>dst
    <tr><td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32
    </table>
<tr>
  <td>NEElementwiseComparison
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>dst
    <tr><td>QASYMM8<td>QASYMM8<td>U8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED<td>U8
    <tr><td>S32<td>S32<td>U8
    <tr><td>U8<td>U8<td>U8
    <tr><td>S16<td>S16<td>U8
    <tr><td>F16<td>F16<td>U8
    <tr><td>F32<td>F32<td>U8
    </table>
<tr>
  <td>CLArithmeticAddition
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>dst
    <tr><td>QASYMM8<td>QASYMM8<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED<td>QASYMM8_SIGNED
    <tr><td>QSYMM16<td>QSYMM16<td>QASYMM16
    <tr><td>U8<td>U8<td>U8
    <tr><td>U8<td>U8<td>S16
    <tr><td>U8<td>S16<td>S16
    <tr><td>S16<td>U8<td>S16
    <tr><td>S16<td>S16<td>S16
    <tr><td>S32<td>S32<td>S32
    <tr><td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32
    </table>
<tr>
  <td>CLArithmeticSubtraction
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>dst
    <tr><td>QASYMM8<td>QASYMM8<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED<td>QASYMM8_SIGNED
    <tr><td>QSYMM16<td>QSYMM16<td>QASYMM16
    <tr><td>U8<td>U8<td>U8
    <tr><td>U8<td>U8<td>S16
    <tr><td>U8<td>S16<td>S16
    <tr><td>S16<td>U8<td>S16
    <tr><td>S16<td>S16<td>S16
    <tr><td>S32<td>S32<td>S32
    <tr><td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32
    </table>
<tr>
  <td>CLArithmeticDivision
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>dst
    <tr><td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32
    </table>
<tr>
  <td>CLElementwiseMax
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>dst
    <tr><td>QASYMM8<td>QASYMM8<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED<td>QASYMM8_SIGNED
    <tr><td>QSYMM16<td>QSYMM16<td>QASYMM16
    <tr><td>U8<td>U8<td>U8
    <tr><td>S16<td>S16<td>S16
    <tr><td>S32<td>S32<td>S32
    <tr><td>U32<td>U32<td>U32
    <tr><td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32
    </table>
<tr>
  <td>CLElementwiseMin
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>dst
    <tr><td>QASYMM8<td>QASYMM8<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED<td>QASYMM8_SIGNED
    <tr><td>QSYMM16<td>QSYMM16<td>QASYMM16
    <tr><td>U8<td>U8<td>U8
    <tr><td>S16<td>S16<td>S16
    <tr><td>S32<td>S32<td>S32
    <tr><td>U32<td>U32<td>U32
    <tr><td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32
    </table>
<tr>
  <td>CLElementwiseSquaredDiff
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>dst
    <tr><td>QASYMM8<td>QASYMM8<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED<td>QASYMM8_SIGNED
    <tr><td>QSYMM16<td>QSYMM16<td>QASYMM16
    <tr><td>U8<td>U8<td>U8
    <tr><td>S16<td>S16<td>S16
    <tr><td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32
    </table>
<tr>
  <td>CLElementwisePower
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>dst
    <tr><td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32
    </table>
<tr>
  <td rowspan="8">ElementwiseUnaryLayer
  <td rowspan="8" style="width:200px;"> Function to perform: - Rsqrt - Exp - Neg - Log - Abs - Round - Sin
  <td rowspan="8">
      <ul>
       <li>ANEURALNETWORKS_ABS
       <li>ANEURALNETWORKS_EXP
       <li>ANEURALNETWORKS_LOG
       <li>ANEURALNETWORKS_NEG
       <li>ANEURALNETWORKS_RSQRT
       <li>ANEURALNETWORKS_SIN
      </ul>
  <td>NEElementwiseUnaryLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>F16<td>F16
    <tr><td>F32<td>F32
    <tr><td>S32<td>S32
    </table>
<tr>
  <td>CLRsqrtLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>F16<td>F16
    <tr><td>F32<td>F32
    </table>
<tr>
  <td>CLExpLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>F16<td>F16
    <tr><td>F32<td>F32
    </table>
<tr>
  <td>CLNegLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>F16<td>F16
    <tr><td>F32<td>F32
    <tr><td>S32<td>S32
    </table>
<tr>
  <td>CLSinLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>F16<td>F16
    <tr><td>F32<td>F32
    </table>
<tr>
  <td>CLLogLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>F16<td>F16
    <tr><td>F32<td>F32
    </table>
<tr>
  <td>CLAbsLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>F16<td>F16
    <tr><td>F32<td>F32
    </table>
<tr>
  <td>CLRoundLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>F16<td>F16
    <tr><td>F32<td>F32
    </table>
<tr>
  <td rowspan="2">FFT1D
  <td rowspan="2" style="width:200px;"> Fast Fourier Transform 1D.
  <td rowspan="2">
      <ul>
       <li>n/a
      </ul>
  <td>NEFFT1D
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>F32<td>F32
    </table>
<tr>
  <td>CLFFT1D
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>F32<td>F32
    <tr><td>F16<td>F16
    </table>
<tr>
  <td rowspan="2">FFT2D
  <td rowspan="2" style="width:200px;"> Fast Fourier Transform 2D.
  <td rowspan="2">
      <ul>
       <li>n/a
      </ul>
  <td>NEFFT2D
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>F32<td>F32
    </table>
<tr>
  <td>CLFFT2D
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>F32<td>F32
    <tr><td>F16<td>F16
    </table>
<tr>
  <td rowspan="2">FFTConvolutionLayer
  <td rowspan="2" style="width:200px;"> Fast Fourier Transform Convolution.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_CONV_2D
      </ul>
  <td>NEFFTConvolutionLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>F32<td>F32
    </table>
<tr>
  <td>CLFFTConvolutionLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>F32<td>F32
    <tr><td>F16<td>F16
    </table>
<tr>
  <td rowspan="2">Fill
  <td rowspan="2" style="width:200px;"> Set the values of a tensor with a given value.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_FILL
      </ul>
  <td>NEFill
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td>CLFill
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td rowspan="1">FillBorder
  <td rowspan="1" style="width:200px;"> Function to fill the borders within the XY-planes.
  <td rowspan="1">
      <ul>
       <li>n/a
      </ul>
  <td>NEFillBorder
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td rowspan="2">FlattenLayer
  <td rowspan="2" style="width:200px;"> Reshape a tensor to be 1D
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_RESHAPE
      </ul>
  <td>NEFlattenLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td>CLFlattenLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td rowspan="2">Floor
  <td rowspan="2" style="width:200px;"> Round the value to the lowest number.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_FLOOR
      </ul>
  <td>NEFloor
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>F32<td>F32
    <tr><td>F16<td>F16
    </table>
<tr>
  <td>CLFloor
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>F32<td>F32
    <tr><td>F16<td>F16
    </table>
<tr>
  <td rowspan="2">FullyConnectedLayer
  <td rowspan="2" style="width:200px;"> Function to perform a fully connected / dense layer.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_FULLY_CONNECTED
      </ul>
  <td>NEFullyConnectedLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>src2<th>dst
    <tr><td>F16<td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32<td>F32
    <tr><td>QASYMM8<td>QASYMM8<td>S32<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED<td>S32<td>QASYMM8_SIGNED
    </table>
<tr>
  <td>CLFullyConnectedLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>src2<th>dst
    <tr><td>F16<td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32<td>F32
    <tr><td>QASYMM8<td>QASYMM8<td>S32<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED<td>S32<td>QASYMM8_SIGNED
    </table>
<tr>
  <td rowspan="2">FuseBatchNormalization
  <td rowspan="2" style="width:200px;"> Function to fuse the batch normalization node to a preceding convolution node.
  <td rowspan="2">
      <ul>
       <li>n/a
      </ul>
  <td>NEFuseBatchNormalization
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>F32<td>F32
    <tr><td>F16<td>F16
    </table>
<tr>
  <td>CLFuseBatchNormalization
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>F32<td>F32
    <tr><td>F16<td>F16
    </table>
<tr>
  <td rowspan="2">Gather
  <td rowspan="2" style="width:200px;"> Performs the Gather operation along the chosen axis.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_GATHER
      </ul>
  <td>NEGather
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td>CLGather
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td rowspan="2">GEMM
  <td rowspan="2" style="width:200px;"> General Matrix Multiplication.
  <td rowspan="2">
      <ul>
       <li>n/a
      </ul>
  <td>NEGEMM
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>src2<th>dst
    <tr><td>F32<td>F32<td>F32<td>F32
    <tr><td>F16<td>F16<td>F16<td>F16
    <tr><td>BFLOAT16<td>BFLOAT16<td>BFLOAT16<td>BFLOAT16
    </table>
<tr>
  <td>CLGEMM
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>src2<th>dst
    <tr><td>F32<td>F32<td>F32<td>F32
    <tr><td>F16<td>F16<td>F16<td>F16
    </table>
<tr>
  <td rowspan="1">GEMMConv2d
  <td rowspan="1" style="width:200px;"> General Matrix Multiplication.
  <td rowspan="1">
      <ul>
       <li>ANEURALNETWORKS_CONV_2D
      </ul>
  <td>NEGEMMConv2d
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>src2<th>dst
    <tr><td>QASYMM8<td>QASYMM8<td>S32<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED<td>S32<td>QASYMM8_SIGNED
    <tr><td>F16<td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32<td>F32
    <tr><td>BFLOAT16<td>BFLOAT16<td>BFLOAT16<td>BFLOAT16
    </table>
<tr>
  <td rowspan="2">GEMMConvolutionLayer
  <td rowspan="2" style="width:200px;"> General Matrix Multiplication.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_CONV_2D
      </ul>
  <td>NEGEMMConvolutionLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>src2<th>dst
    <tr><td>F16<td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32<td>F32
    <tr><td>BFLOAT16<td>BFLOAT16<td>BFLOAT16<td>BFLOAT16
    <tr><td>QASYMM8<td>QASYMM8<td>S32<td>QASYMM8
    <tr><td>QASYMM8<td>QSYMM8_PER_CHANNEL<td>S32<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED<td>S32<td>QASYMM8_SIGNED
    <tr><td>QASYMM8_SIGNED<td>QSYMM8_PER_CHANNEL<td>S32<td>QASYMM8_SIGNED
    </table>
<tr>
  <td>CLGEMMConvolutionLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>src2<th>dst
    <tr><td>F16<td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32<td>F32
    <tr><td>QASYMM8<td>QASYMM8<td>S32<td>QASYMM8
    <tr><td>QASYMM8<td>QSYMM8_PER_CHANNEL<td>S32<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED<td>S32<td>QASYMM8_SIGNED
    <tr><td>QASYMM8_SIGNED<td>QSYMM8_PER_CHANNEL<td>S32<td>QASYMM8_SIGNED
    </table>
<tr>
  <td rowspan="1">GEMMDeconvolutionLayer
  <td rowspan="1" style="width:200px;"> General Matrix Multiplication.
  <td rowspan="1">
      <ul>
       <li>ANEURALNETWORKS_TRANSPOSE_CONV_2D
      </ul>
  <td>CLGEMMDeconvolutionLayer
  <td>
      <ul>
       <li>NHWC
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>src2<th>dst
    <tr><td>F16<td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32<td>F32
    <tr><td>QASYMM8<td>QASYMM8<td>S32<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED<td>S32<td>QASYMM8_SIGNED
    </table>
<tr>
  <td rowspan="2">GEMMLowpMatrixMultiplyCore
  <td rowspan="2" style="width:200px;"> General Matrix Multiplication.
  <td rowspan="2">
      <ul>
       <li>n/a
      </ul>
  <td>NEGEMMLowpMatrixMultiplyCore
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>src2<th>dst
    <tr><td>QASYMM8<td>QASYMM8<td>S32<td>QASYMM8
    <tr><td>QASYMM8<td>QSYMM8_PER_CHANNEL<td>S32<td>QASYMM8
    <tr><td>QASYMM8<td>QSYMM8<td>S32<td>QASYMM8
    <tr><td>QASYMM8<td>QASYMM8<td>S32<td>S32
    <tr><td>QASYMM8<td>QSYMM8_PER_CHANNEL<td>S32<td>S32
    <tr><td>QASYMM8<td>QSYMM8<td>S32<td>S32
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED<td>S32<td>QASYMM8_SIGNED
    <tr><td>QASYMM8_SIGNED<td>QSYMM8_PER_CHANNEL<td>S32<td>QASYMM8_SIGNED
    <tr><td>QASYMM8_SIGNED<td>QSYMM8<td>S32<td>QASYMM8_SIGNED
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED<td>S32<td>S32
    <tr><td>QASYMM8_SIGNED<td>QSYMM8_PER_CHANNEL<td>S32<td>S32
    <tr><td>QASYMM8_SIGNED<td>QSYMM8<td>S32<td>S32
    </table>
<tr>
  <td>CLGEMMLowpMatrixMultiplyCore
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>src2<th>dst
    <tr><td>QASYMM8<td>QASYMM8<td>S32<td>QASYMM8
    <tr><td>QASYMM8<td>QSYMM8_PER_CHANNEL<td>S32<td>QASYMM8
    <tr><td>QASYMM8<td>QSYMM8<td>S32<td>QASYMM8
    <tr><td>QASYMM8<td>QASYMM8<td>S32<td>S32
    <tr><td>QASYMM8<td>QSYMM8_PER_CHANNEL<td>S32<td>S32
    <tr><td>QASYMM8<td>QSYMM8<td>S32<td>S32
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED<td>S32<td>QASYMM8_SIGNED
    <tr><td>QASYMM8_SIGNED<td>QSYMM8_PER_CHANNEL<td>S32<td>QASYMM8_SIGNED
    <tr><td>QASYMM8_SIGNED<td>QSYMM8<td>S32<td>QASYMM8_SIGNED
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED<td>S32<td>S32
    <tr><td>QASYMM8_SIGNED<td>QSYMM8_PER_CHANNEL<td>S32<td>S32
    <tr><td>QASYMM8_SIGNED<td>QSYMM8<td>S32<td>S32
    </table>
<tr>
  <td rowspan="2">GEMMLowpOutputStage
  <td rowspan="2" style="width:200px;"> General Matrix Multiplication.
  <td rowspan="2">
      <ul>
       <li>n/a
      </ul>
  <td>NEGEMMLowpOutputStage
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>dst
    <tr><td>S32<td>S32<td>QASYMM8
    <tr><td>S32<td>S32<td>QASYMM8_SIGNED
    <tr><td>S32<td>S32<td>QSYMM16
    </table>
<tr>
  <td>CLGEMMLowpOutputStage
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>dst
    <tr><td>S32<td>S32<td>QASYMM8
    <tr><td>S32<td>S32<td>QASYMM8_SIGNED
    <tr><td>S32<td>S32<td>QSYMM16
    </table>
<tr>
  <td rowspan="2">GenerateProposalsLayer
  <td rowspan="2" style="width:200px;"> Function to generate proposals for a RPN (Region Proposal Network).
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_GENERATE_PROPOSALS
      </ul>
  <td>NEGenerateProposalsLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>src2<th>dst
    <tr><td>F16<td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32<td>F32
    <tr><td>QASYMM8<td>QSYMM8<td>QSYMM16<td>QASYMM8
    </table>
<tr>
  <td>CLGenerateProposalsLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>src2<th>dst
    <tr><td>F16<td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32<td>F32
    <tr><td>QASYMM8<td>QSYMM8<td>QSYMM16<td>QASYMM8
    </table>
<tr>
  <td rowspan="2">InstanceNormalizationLayer
  <td rowspan="2" style="width:200px;"> Function to perform a Instance normalization on a given axis.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_INSTANCE_NORMALIZATION
      </ul>
  <td>NEInstanceNormalizationLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>F16<td>F16
    <tr><td>F32<td>F32
    </table>
<tr>
  <td>CLInstanceNormalizationLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>F16<td>F16
    <tr><td>F32<td>F32
    </table>
<tr>
  <td rowspan="2">L2NormalizeLayer
  <td rowspan="2" style="width:200px;"> Function to perform a L2 normalization on a given axis.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_L2_NORMALIZATION
      </ul>
  <td>NEL2NormalizeLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>F16<td>F16
    <tr><td>F32<td>F32
    </table>
<tr>
  <td>CLL2NormalizeLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>F16<td>F16
    <tr><td>F32<td>F32
    </table>
<tr>
  <td rowspan="3">Logical
  <td rowspan="3" style="width:200px;"> Function to perform: - Logical AND - Logical OR - Logical NOT
  <td rowspan="3">
      <ul>
       <li>n/a
      </ul>
  <td>NELogicalAnd
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>dst
    <tr><td>U8<td>U8<td>U8
    </table>
<tr>
  <td>NELogicalOr
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>dst
    <tr><td>U8<td>U8<td>U8
    </table>
<tr>
  <td>NELogicalNot
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>U8<td>U8
    </table>
<tr>
  <td rowspan="1">LogicalAnd
  <td rowspan="1" style="width:200px;"> Function to perform Logical AND.
  <td rowspan="1">
      <ul>
       <li>n/a
      </ul>
  <td>CLLogicalAnd
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>dst
    <tr><td>U8<td>U8<td>U8
    </table>
<tr>
  <td rowspan="1">LogicalOr
  <td rowspan="1" style="width:200px;"> Function to perform Logical OR.
  <td rowspan="1">
      <ul>
       <li>n/a
      </ul>
  <td>CLLogicalOr
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>dst
    <tr><td>U8<td>U8<td>U8
    </table>
<tr>
  <td rowspan="1">LogicalNot
  <td rowspan="1" style="width:200px;"> Function to perform Logical NOT.
  <td rowspan="1">
      <ul>
       <li>n/a
      </ul>
  <td>CLLogicalNot
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>U8<td>U8
    </table>
<tr>
  <td rowspan="2">LSTMLayer
  <td rowspan="2" style="width:200px;"> Function to perform a single time step in a Long Short-Term Memory (LSTM) layer.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_LSTM
      </ul>
  <td>NELSTMLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0 - src13<th>dst0 - dst3
    <tr><td>F16<td>F16
    <tr><td>F32<td>F32
    </table>
<tr>
  <td>CLLSTMLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0 - src13<th>dst0 - dst3
    <tr><td>F16<td>F16
    <tr><td>F32<td>F32
    </table>
<tr>
  <td rowspan="2">LSTMLayerQuantized
  <td rowspan="2" style="width:200px;"> Function to perform quantized LSTM (Long Short-Term Memory)
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_QUANTIZED_LSTM
       <li>ANEURALNETWORKS_QUANTIZED_16BIT_LSTM
      </ul>
  <td>NELSTMLayerQuantized
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0 - src8<th>src9 - src12<th>src13<th>src14<th>dst0<th>dst1
    <tr><td>QASYMM8<td>S32<td>QSYMM16<td>QASYMM8<td>QSYMM16<td>QASYMM8
    </table>
<tr>
  <td>CLLSTMLayerQuantized
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0 - src8<th>src9 - src12<th>src13<th>src14<th>dst0<th>dst1
    <tr><td>QASYMM8<td>S32<td>QSYMM16<td>QASYMM8<td>QSYMM16<td>QASYMM8
    </table>
<tr>
  <td rowspan="2">MaxUnpoolingLayer
  <td rowspan="2" style="width:200px;"> Function to perform MaxUnpooling.
  <td rowspan="2">
      <ul>
       <li>n/a
      </ul>
  <td>NEMaxUnpoolingLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>QASYMM8<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED
    <tr><td>F16<td>F16
    <tr><td>F32<td>F32
    </table>
<tr>
  <td>CLMaxUnpoolingLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>QASYMM8<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED
    <tr><td>F16<td>F16
    <tr><td>F32<td>F32
    </table>
<tr>
  <td rowspan="2">MeanStdDevNormalizationLayer
  <td rowspan="2" style="width:200px;"> Function to execute mean and standard deviation normalization.
  <td rowspan="2">
      <ul>
       <li>n/a
      </ul>
  <td>NEMeanStdDevNormalizationLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>F32<td>F32
    <tr><td>F16<td>F16
    </table>
<tr>
  <td>CLMeanStdDevNormalizationLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>F32<td>F32
    <tr><td>F16<td>F16
    </table>
<tr>
  <td rowspan="2">NormalizationLayer
  <td rowspan="2" style="width:200px;"> Function to compute normalization layer.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_LOCAL_RESPONSE_NORMALIZATION
      </ul>
  <td>NENormalizationLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>F32<td>F32
    <tr><td>F16<td>F16
    </table>
<tr>
  <td>CLNormalizationLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>F32<td>F32
    <tr><td>F16<td>F16
    </table>
<tr>
  <td rowspan="2">PadLayer
  <td rowspan="2" style="width:200px;"> Function to pad a tensor.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_PAD
       <li>ANEURALNETWORKS_PAD_V2
      </ul>
  <td>NEPadLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td>CLPadLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td rowspan="2">Permute
  <td rowspan="2" style="width:200px;"> Function to transpose an ND tensor.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_TRANSPOSE
      </ul>
  <td>NEPermute
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td>CLPermute
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td rowspan="2">PixelWiseMultiplication
  <td rowspan="2" style="width:200px;"> Function to perform a multiplication.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_MUL
      </ul>
  <td>NEPixelWiseMultiplication
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>dst
    <tr><td>QASYMM8<td>QASYMM8<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED<td>QASYMM8_SIGNED
    <tr><td>QSYMM16<td>QSYMM16<td>QASYMM16
    <tr><td>QSYMM16<td>QSYMM16<td>S32
    <tr><td>U8<td>U8<td>U8
    <tr><td>U8<td>U8<td>S16
    <tr><td>U8<td>S16<td>S16
    <tr><td>S16<td>U8<td>S16
    <tr><td>S16<td>S16<td>S16
    <tr><td>F16<td>F16<td>F16
    <tr><td>F32<td>S32<td>F32
    </table>
<tr>
  <td>CLPixelWiseMultiplication
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>dst
    <tr><td>QASYMM8<td>QASYMM8<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED<td>QASYMM8_SIGNED
    <tr><td>QSYMM16<td>QSYMM16<td>QASYMM16
    <tr><td>QSYMM16<td>QSYMM16<td>S32
    <tr><td>U8<td>U8<td>U8
    <tr><td>U8<td>U8<td>S16
    <tr><td>U8<td>S16<td>S16
    <tr><td>S16<td>U8<td>S16
    <tr><td>S16<td>S16<td>S16
    <tr><td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32
    <tr><td>S32<td>S32<td>S32
    </table>
<tr>
  <td rowspan="2">PoolingLayer
  <td rowspan="2" style="width:200px;"> Function to perform pooling with the specified pooling operation.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_AVERAGE_POOL_2D
       <li>ANEURALNETWORKS_L2_POOL_2D
       <li>ANEURALNETWORKS_MAX_POOL_2D
      </ul>
  <td>NEPoolingLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>QASYMM8<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED
    <tr><td>F16<td>F16
    <tr><td>F32<td>F32
    </table>
<tr>
  <td>CLPoolingLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>QASYMM8<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED
    <tr><td>F16<td>F16
    <tr><td>F32<td>F32
    </table>
<tr>
  <td rowspan="2">Pooling3dLayer
  <td rowspan="2" style="width:200px;"> Function to perform pooling 3D with the specified pooling operation.
  <td rowspan="2">
      <ul>
       <li>N/A
      </ul>
  <td>NEPooling3dLayer
  <td>
      <ul>
       <li>NDHWC
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>F16<td>F16
    <tr><td>F32<td>F32
    <tr><td>QASYMM8<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED
    </table>
<tr>
  <td>CLPooling3dLayer
  <td>
      <ul>
       <li>NDHWC
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>F16<td>F16
    <tr><td>F32<td>F32
    <tr><td>QASYMM8<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED
    </table>
<tr>
  <td rowspan="2">PReluLayer
  <td rowspan="2" style="width:200px;"> Function to compute the activation layer with the PRELU activation function.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_PRELU
      </ul>
  <td>NEPReluLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>QASYMM8<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED
    <tr><td>F16<td>F16
    <tr><td>F32<td>F32
    </table>
<tr>
  <td>CLPReluLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>QASYMM8<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED
    <tr><td>F16<td>F16
    <tr><td>F32<td>F32
    </table>
<tr>
  <td rowspan="2">PriorBoxLayer
  <td rowspan="2" style="width:200px;"> Function to compute prior boxes and clip.
  <td rowspan="2">
      <ul>
       <li>n/a
      </ul>
  <td>NEPriorBoxLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>dst
    <tr><td>F32<td>F32<td>F32
    </table>
<tr>
  <td>CLPriorBoxLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>dst
    <tr><td>F32<td>F32<td>F32
    </table>
<tr>
  <td rowspan="2">QLSTMLayer
  <td rowspan="2" style="width:200px;"> Function to perform quantized LSTM (Long Short-Term Memory).
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_QUANTIZED_LSTM
       <li>ANEURALNETWORKS_QUANTIZED_16BIT_LSTM
      </ul>
  <td>NEQLSTMLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1 - src6<th>src7 -src9<th>src10<th>src11<th>dst0<th>dst1 - dst2
    <tr><td>QASYMM8_SIGNED<td>QASYMM8<td>S32<td>QSYMM16<td>QASYMM8_SIGNED<td>QSYMM16<td>QASYMM8_SIGNED
    </table>
<tr>
  <td>CLQLSTMLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1 - src6<th>src7 -src9<th>src10<th>src11<th>dst0<th>dst1 - dst2
    <tr><td>QASYMM8_SIGNED<td>QASYMM8<td>S32<td>QSYMM16<td>QASYMM8_SIGNED<td>QSYMM16<td>QASYMM8_SIGNED
    </table>
<tr>
  <td rowspan="2">QuantizationLayer
  <td rowspan="2" style="width:200px;"> Function to perform quantization layer
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_QUANTIZE
      </ul>
  <td>NEQuantizationLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>QASYMM8<td>QASYMM8, QASYMM8_SIGNED, QASYMM16
    <tr><td>QASYMM8_SIGNED<td>QASYMM8, QASYMM8_SIGNED, QASYMM16
    <tr><td>F16<td>QASYMM8, QASYMM8_SIGNED, QASYMM16
    <tr><td>F32<td>QASYMM8, QASYMM8_SIGNED, QASYMM16
    </table>
<tr>
  <td>CLQuantizationLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>QASYMM8<td>QASYMM8, QASYMM8_SIGNED, QASYMM16
    <tr><td>QASYMM8_SIGNED<td>QASYMM8, QASYMM8_SIGNED, QASYMM16
    <tr><td>F16<td>QASYMM8, QASYMM8_SIGNED, QASYMM16
    <tr><td>F32<td>QASYMM8, QASYMM8_SIGNED, QASYMM16
    </table>
<tr>
  <td rowspan="2">Range
  <td rowspan="2" style="width:200px;"> Function to generates a sequence of numbers starting from START and extends by increments of 'STEP' up to but not including 'END'.
  <td rowspan="2">
      <ul>
       <li>n/a
      </ul>
  <td>NERange
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>dst
    <tr><td>U8
    <tr><td>S8
    <tr><td>U16
    <tr><td>S16
    <tr><td>U32
    <tr><td>S32
    <tr><td>F16
    <tr><td>F32
    </table>
<tr>
  <td>CLRange
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>dst
    <tr><td>U8
    <tr><td>S8
    <tr><td>QASYMM8
    <tr><td>U16
    <tr><td>S16
    <tr><td>U32
    <tr><td>S32
    <tr><td>F16
    <tr><td>F32
    </table>
<tr>
  <td rowspan="2">ReduceMean
  <td rowspan="2" style="width:200px;"> Function to perform reduce mean operation.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_MEAN
      </ul>
  <td>NEReduceMean
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>QASYMM8<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED
    <tr><td>F16<td>F16
    <tr><td>F32<td>F32
    </table>
<tr>
  <td>CLReduceMean
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>QASYMM8<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED
    <tr><td>F16<td>F16
    <tr><td>F32<td>F32
    </table>
<tr>
  <td rowspan="2">ReductionOperation
  <td rowspan="2" style="width:200px;"> Function to perform reduce with the following operations - ARG_IDX_MAX: Index of the max value - ARG_IDX_MIN: Index of the min value - MEAN_SUM:    Mean of sum - PROD:        Product - SUM_SQUARE:  Sum of squares - SUM:         Sum - MIN:         Min - MAX:         Max
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_REDUCE_ALL
       <li>ANEURALNETWORKS_REDUCE_ANY
       <li>ANEURALNETWORKS_REDUCE_MAX
       <li>ANEURALNETWORKS_REDUCE_MIN
       <li>ANEURALNETWORKS_REDUCE_PROD
       <li>ANEURALNETWORKS_REDUCE_SUM
      </ul>
  <td>NEReductionOperation
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>QASYMM8<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED
    <tr><td>F16<td>F16
    <tr><td>F32<td>F32
    <tr><td>S32<td>S32
    </table>
<tr>
  <td>CLReductionOperation
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>QASYMM8<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED
    <tr><td>F16<td>F16
    <tr><td>F32<td>F32
    <tr><td>S32<td>S32
    </table>
<tr>
  <td rowspan="2">ReorgLayer
  <td rowspan="2" style="width:200px;"> Performs a reorganization layer of input tensor to the output tensor.
  <td rowspan="2">
      <ul>
       <li>n/a
      </ul>
  <td>NEReorgLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td>CLReorgLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td rowspan="2">ReshapeLayer
  <td rowspan="2" style="width:200px;"> Function to reshape a tensor.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_RESHAPE
       <li>ANEURALNETWORKS_SQUEEZE
      </ul>
  <td>NEReshapeLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td>CLReshapeLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td rowspan="2">Reverse
  <td rowspan="2" style="width:200px;"> Function to reverse tensor according to axis.
  <td rowspan="2">
      <ul>
       <li>n/a
      </ul>
  <td>NEReverse
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>dst
    <tr><td>All<td>U32<td>All
    </table>
<tr>
  <td>CLReverse
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>dst
    <tr><td>All<td>U32<td>All
    </table>
<tr>
  <td rowspan="2">RNNLayer
  <td rowspan="2" style="width:200px;"> Function to perform recurrent neural network layer.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_RNN
      </ul>
  <td>NERNNLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>src2<th>src3<th>dst0<th>dst1
    <tr><td>F16<td>F16<td>F16<td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32<td>F32<td>F32<td>F32
    </table>
<tr>
  <td>CLRNNLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>src2<th>src3<th>dst0<th>dst1
    <tr><td>F16<td>F16<td>F16<td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32<td>F32<td>F32<td>F32
    </table>
<tr>
  <td rowspan="2">ROIAlignLayer
  <td rowspan="2" style="width:200px;"> Function to perform ROI alignment.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_ROI_ALIGN
      </ul>
  <td>NEROIAlignLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>dst
    <tr><td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32
    <tr><td>QASYMM8<td>QASYMM16<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM16<td>QASYMM8_SIGNED
    </table>
<tr>
  <td>CLROIAlignLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>dst
    <tr><td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32
    <tr><td>QASYMM8<td>QASYMM16<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM16<td>QASYMM8_SIGNED
    </table>
<tr>
  <td rowspan="2">ROIPoolingLayer
  <td rowspan="2" style="width:200px;"> Function to perform ROI pooling.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_ROI_POOLING
      </ul>
  <td>NEROIPoolingLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>dst
    <tr><td>F32<td>U16<td>F32
    <tr><td>QASYMM8<td>U16<td>QASYMM8
    </table>
<tr>
  <td>CLROIPoolingLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>dst
    <tr><td>F16<td>U16<td>F16
    <tr><td>F32<td>U16<td>F32
    <tr><td>QASYMM8<td>U16<td>QASYMM8
    </table>
<tr>
  <td rowspan="2">Scale
  <td rowspan="2" style="width:200px;"> Function to perform resize a tensor using to interpolate: - Bilinear - Nearest neighbor
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_RESIZE_BILINEAR
       <li>ANEURALNETWORKS_RESIZE_NEAREST_NEIGHBOR
      </ul>
  <td>NEScale
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>QASYMM8<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED
    <tr><td>F16<td>F16
    <tr><td>F32<td>F32
    <tr><td>U8<td>U8
    <tr><td>S16<td>S16
    </table>
<tr>
  <td>CLScale
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>QASYMM8<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED
    <tr><td>F16<td>F16
    <tr><td>F32<td>F32
    <tr><td>U8<td>U8
    <tr><td>S16<td>S16
    </table>
<tr>
  <td rowspan="2">Select
  <td rowspan="2" style="width:200px;"> Function to select values from 2 tensors depending on an input tensor of booleans.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_SELECT
      </ul>
  <td>NESelect
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>src2<th>dst
    <tr><td>U8<td>All<td>All<td>All
    </table>
<tr>
  <td>CLSelect
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>src2<th>dst
    <tr><td>U8<td>All<td>All<td>All
    </table>
<tr>
  <td rowspan="2">Slice
  <td rowspan="2" style="width:200px;"> Function to perform tensor slicing.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_SLICE
      </ul>
  <td>NESlice
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td>CLSlice
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td rowspan="2">SoftmaxLayer
  <td rowspan="2" style="width:200px;"> Function to compute a SoftmaxLayer and a Log SoftmaxLayer.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_LOG_SOFTMAX
       <li>ANEURALNETWORKS_SOFTMAX
      </ul>
  <td>NESoftmaxLayerGeneric
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>QASYMM8<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED
    <tr><td>F16<td>F16
    <tr><td>F32<td>F32
    </table>
<tr>
  <td>CLSoftmaxLayerGeneric
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>QASYMM8<td>QASYMM8
    <tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED
    <tr><td>F16<td>F16
    <tr><td>F32<td>F32
    </table>
<tr>
  <td rowspan="2">SpaceToBatchLayer
  <td rowspan="2" style="width:200px;"> Function to divide a tensor spatially.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_SPACE_TO_BATCH_ND
      </ul>
  <td>NESpaceToBatchLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>src2<th>dst
    <tr><td>All<td>S32<td>S32<td>All
    </table>
<tr>
  <td>CLSpaceToBatchLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>src2<th>dst
    <tr><td>All<td>S32<td>S32<td>All
    </table>
<tr>
  <td rowspan="2">SpaceToDepthLayer
  <td rowspan="2" style="width:200px;"> Function to rearrange blocks of spatial data into depth.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_SPACE_TO_DEPTH
      </ul>
  <td>NESpaceToDepthLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td>CLSpaceToDepthLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td rowspan="2">Split
  <td rowspan="2" style="width:200px;"> Function to split a tensor along a given axis.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_SPLIT
      </ul>
  <td>NESplit
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td>CLSplit
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td rowspan="2">StackLayer
  <td rowspan="2" style="width:200px;"> Function to stack tensors along an axis.
  <td rowspan="2">
      <ul>
       <li>n/a
      </ul>
  <td>NEStackLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td>CLStackLayer
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td rowspan="2">StridedSlice
  <td rowspan="2" style="width:200px;"> Function to extract a strided slice of a tensor.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_STRIDED_SLICE
      </ul>
  <td>NEStridedSlice
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td>CLStridedSlice
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td rowspan="2">Tile
  <td rowspan="2" style="width:200px;"> Function to construct a tensor by tiling a given tensor.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_TILE
      </ul>
  <td>NETile
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td>CLTile
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td rowspan="2">Transpose
  <td rowspan="2" style="width:200px;"> Function to transpose a 2D tensor.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_TRANSPOSE
      </ul>
  <td>NETranspose
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td>CLTranspose
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td rowspan="2">Unstack
  <td rowspan="2" style="width:200px;"> Function to unpack a rank-R tensor into rank-(R-1) tensors.
  <td rowspan="2">
      <ul>
       <li>n/a
      </ul>
  <td>NEUnstack
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td>CLUnstack
  <td>
      <ul>
       <li>All
      </ul>
  <td>
    <table>
    <tr><th>src<th>dst
    <tr><td>All<td>All
    </table>
<tr>
  <td rowspan="2">WinogradConvolutionLayer
  <td rowspan="2" style="width:200px;"> Function to do Winograd Convolution.
  <td rowspan="2">
      <ul>
       <li>ANEURALNETWORKS_CONV_2D
      </ul>
  <td>NEWinogradConvolutionLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>src2<th>dst
    <tr><td>F16<td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32<td>F32
    </table>
<tr>
  <td>CLWinogradConvolutionLayer
  <td>
      <ul>
       <li>NHWC
       <li>NCHW
      </ul>
  <td>
    <table>
    <tr><th>src0<th>src1<th>src2<th>dst
    <tr><td>F16<td>F16<td>F16<td>F16
    <tr><td>F32<td>F32<td>F32<td>F32
    </table>
</table>

*/
} // namespace