Update operator list part3
Partially resolve: COMPMID-4199
Signed-off-by: Sheri Zhang <sheri.zhang@arm.com>
Change-Id: Id24702d258fb4e04ad948e7cf6c0efd98d2a5456
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5561
Reviewed-by: TeresaARM <teresa.charlinreyes@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
diff --git a/docs/09_operators_list.dox b/docs/09_operators_list.dox
index 244f292..fc41265 100644
--- a/docs/09_operators_list.dox
+++ b/docs/09_operators_list.dox
@@ -145,6 +145,62 @@
<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>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 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>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 rowspan="2">BatchNormalizationLayer
<td rowspan="2" style="width:200px;"> Function to perform batch normalization.
<td rowspan="2">
@@ -422,6 +478,28 @@
<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">
@@ -554,6 +632,23 @@
<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">
@@ -622,6 +717,24 @@
<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">
@@ -768,6 +881,25 @@
<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">
@@ -802,6 +934,364 @@
<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
+ </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">
@@ -1009,7 +1499,7 @@
<ul>
<li>ANEURALNETWORKS_FULLY_CONNECTED
</ul>
- <td>NEFullyConnectedLayerReshapeWeightsManaged
+ <td>NEFullyConnectedLayer
<td>
<ul>
<li>NHWC
@@ -1024,7 +1514,7 @@
<tr><td>QASYMM8_SIGNED<td>QASYMM8_SIGNED<td>S32<td>QASYMM8_SIGNED
</table>
<tr>
- <td>CLFullyConnectedLayerReshapeWeightsManaged
+ <td>CLFullyConnectedLayer
<td>
<ul>
<li>NHWC
@@ -1118,7 +1608,7 @@
<tr><td>BFLOAT16<td>BFLOAT16<td>BFLOAT16<td>BFLOAT16
</table>
<tr>
- <td>CLGEMMReshapeRHSMatrixKernelManaged
+ <td>CLGEMM
<td>
<ul>
<li>All
@@ -1130,13 +1620,34 @@
<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>NEConvolutionLayerReshapeWeights
+ <td>NEGEMMConvolutionLayer
<td>
<ul>
<li>NHWC
@@ -1154,7 +1665,7 @@
<tr><td>QASYMM8_SIGNED<td>QSYMM8_PER_CHANNEL<td>S32<td>QASYMM8_SIGNED
</table>
<tr>
- <td>CLConvolutionLayerReshapeWeights
+ <td>CLGEMMConvolutionLayer
<td>
<ul>
<li>NHWC
@@ -1171,6 +1682,26 @@
<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">
@@ -1223,6 +1754,38 @@
<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">
@@ -1319,6 +1882,96 @@
<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">
@@ -1660,7 +2313,7 @@
</table>
<tr>
<td rowspan="2">PriorBoxLayer
- <td rowspan="2" style="width:200px;"> Function to .
+ <td rowspan="2" style="width:200px;"> Function to compute prior boxes and clip.
<td rowspan="2">
<ul>
<li>n/a
@@ -2151,6 +2804,41 @@
<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">
@@ -2410,6 +3098,25 @@
<tr><td>F16<td>F16<td>F16<td>F16
<tr><td>F32<td>F32<td>F32<td>F32
</table>
+<tr>
+ <td rowspan="1">WinogradInputTransform
+ <td rowspan="1" style="width:200px;"> Function to.
+ <td rowspan="1">
+ <ul>
+ <li>n/a
+ </ul>
+ <td>CLWinogradInputTransform
+ <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>
</table>
*/