COMPMID-908 - Merge Activation layer with Convolution Layer (NEON. CL, GLES)

Change-Id: Iab06d0768ecf805b841e601185608aae88cf9166
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/120874
Tested-by: Jenkins <bsgcomp@arm.com>
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>

11 files changed, 74 insertions, 26 deletions

diff --git a/arm_compute/core/GLES_COMPUTE/kernels/GCDirectConvolutionLayerKernel.h b/arm_compute/core/GLES_COMPUTE/kernels/GCDirectConvolutionLayerKernel.h
index f1c9c1933d..630253f64f 100644
--- a/arm_compute/core/GLES_COMPUTE/kernels/GCDirectConvolutionLayerKernel.h
+++ b/arm_compute/core/GLES_COMPUTE/kernels/GCDirectConvolutionLayerKernel.h
@@ -58,8 +58,10 @@ public:
      * @param[out] output    The output tensor. First 2 lower dimensions represent a transform of each 3D input,
      *                       while every dimension above represents a batch. Data types supported: Same as @p input
      * @param[in]  conv_info Contains padding and stride information described in @ref PadStrideInfo.
+     * @param[in]  act_info  (Optional) Activation layer information in case of a fused activation.
      */
-    void configure(const IGCTensor *input, const IGCTensor *weights, const IGCTensor *bias, IGCTensor *output, const PadStrideInfo &conv_info);
+    void configure(const IGCTensor *input, const IGCTensor *weights, const IGCTensor *bias, IGCTensor *output,
+                   const PadStrideInfo &conv_info, const ActivationLayerInfo &act_info = ActivationLayerInfo());
 
     // Inherited methods overridden:
     BorderSize border_size() const override;
diff --git a/arm_compute/runtime/CL/functions/CLConvolutionLayer.h b/arm_compute/runtime/CL/functions/CLConvolutionLayer.h
index 5f383b28ab..7df765988f 100644
--- a/arm_compute/runtime/CL/functions/CLConvolutionLayer.h
+++ b/arm_compute/runtime/CL/functions/CLConvolutionLayer.h
@@ -24,10 +24,9 @@
 #ifndef __ARM_COMPUTE_CLCONVOLUTIONLAYER_H__
 #define __ARM_COMPUTE_CLCONVOLUTIONLAYER_H__
 
-#include "arm_compute/runtime/IFunction.h"
-
 #include "arm_compute/runtime/CL/functions/CLDirectConvolutionLayer.h"
 #include "arm_compute/runtime/CL/functions/CLGEMMConvolutionLayer.h"
+#include "arm_compute/runtime/IFunction.h"
 #include "arm_compute/runtime/IMemoryManager.h"
 
 #include <memory>
@@ -57,9 +56,10 @@ public:
      * @param[in]  conv_info    Contains padding and stride information described in @ref PadStrideInfo.
      * @param[in]  weights_info Specifies if the weights tensor has been reshaped with CLWeightsReshapeKernel. Data type supported: Same as @p input.
      * @param[in]  dilation     (Optional) Dilation, in elements, across x and y. Defaults to (1, 1).
+     * @param[in]  act_info     (Optional) Activation layer information in case of a fused activation.
      */
     void configure(ICLTensor *input, const ICLTensor *weights, const ICLTensor *biases, ICLTensor *output, const PadStrideInfo &conv_info, const WeightsInfo &weights_info = WeightsInfo(),
-                   const Size2D &dilation = Size2D(1U, 1U));
+                   const Size2D &dilation = Size2D(1U, 1U), const ActivationLayerInfo &act_info = ActivationLayerInfo());
     /** Static function to check if given info will lead to a valid configuration of @ref CLConvolutionLayer
      *
      * @param[in] input        Source tensor. 3 lower dimensions represent a single input [width, height, IFM],
@@ -72,11 +72,12 @@ public:
      * @param[in] conv_info    Contains padding and stride information described in @ref PadStrideInfo.
      * @param[in] weights_info Specifies if the weights tensor has been reshaped with CLWeightsReshapeKernel. Data type supported: Same as @p input.
      * @param[in] dilation     (Optional) Dilation, in elements, across x and y. Defaults to (1, 1).
+     * @param[in] act_info     (Optional) Activation layer information in case of a fused activation.
      *
      * @return a status
      */
     static Status validate(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *biases, const ITensorInfo *output, const PadStrideInfo &conv_info,
-                           const WeightsInfo &weights_info = WeightsInfo(), const Size2D &dilation = Size2D(1U, 1U));
+                           const WeightsInfo &weights_info = WeightsInfo(), const Size2D &dilation = Size2D(1U, 1U), const ActivationLayerInfo &act_info = ActivationLayerInfo());
     /** Static function to check if given info will return the convolution called by @ref CLConvolutionLayer
      *
      * @param[in] input        Source tensor. 3 lower dimensions represent a single input [width, height, IFM],
@@ -88,14 +89,14 @@ public:
      *                         Data types supported: Same as @p input.
      * @param[in] conv_info    Contains padding and stride information described in @ref PadStrideInfo.
      * @param[in] weights_info Specifies if the weights tensor has been reshaped with CLWeightsReshapeKernel. Data type supported: Same as @p input.
+     * @param[in] act_info     (Optional) Activation layer information in case of a fused activation.
      * @param[in] gpu_target   Specifies the @p GPUTarget.
      * @param[in] dilation     (Optional) Dilation, in elements, across x and y. Defaults to (1, 1).
      *
      * @return a status
      */
     static ConvolutionMethod get_convolution_method(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *biases, const ITensorInfo *output, const PadStrideInfo &conv_info,
-                                                    const WeightsInfo &weights_info, const GPUTarget gpu_target, const Size2D &dilation = Size2D(1U, 1U));
-
+                                                    const WeightsInfo &weights_info, const ActivationLayerInfo &act_info, const GPUTarget gpu_target, const Size2D &dilation = Size2D(1U, 1U));
     // Inherited methods overridden:
     void run() override;
 
diff --git a/arm_compute/runtime/CL/functions/CLDirectConvolutionLayer.h b/arm_compute/runtime/CL/functions/CLDirectConvolutionLayer.h
index 205c90c478..4363224a08 100644
--- a/arm_compute/runtime/CL/functions/CLDirectConvolutionLayer.h
+++ b/arm_compute/runtime/CL/functions/CLDirectConvolutionLayer.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017, 2018 ARM Limited.
+ * Copyright (c) 2017-2018 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -27,6 +27,7 @@
 #include "arm_compute/core/CL/kernels/CLDirectConvolutionLayerKernel.h"
 #include "arm_compute/core/CL/kernels/CLFillBorderKernel.h"
 #include "arm_compute/core/Types.h"
+#include "arm_compute/runtime/CL/functions/CLActivationLayer.h"
 #include "arm_compute/runtime/IFunction.h"
 
 #include <memory>
@@ -53,8 +54,9 @@ public:
      * @param[out] output    Destination tensor. 3 lower dimensions represent a single output [width, height, OFM], while the rest represent batch of outputs.
      *                       Data types supported: Same as @p input.
      * @param[in]  conv_info Contains padding and stride information described in @ref PadStrideInfo.
+     * @param[in]  act_info  (Optional) Activation layer information in case of a fused activation.
      */
-    void configure(ICLTensor *input, const ICLTensor *weights, const ICLTensor *biases, ICLTensor *output, const PadStrideInfo &conv_info);
+    void configure(ICLTensor *input, const ICLTensor *weights, const ICLTensor *biases, ICLTensor *output, const PadStrideInfo &conv_info, const ActivationLayerInfo &act_info = ActivationLayerInfo());
     /** Static function to check if given info will lead to a valid configuration of @ref CLDirectConvolutionLayer
      *
      * @param[in] input     Source tensor. 3 lower dimensions represent a single input [width, height, IFM],
@@ -65,10 +67,12 @@ public:
      * @param[in] output    Destination tensor. 3 lower dimensions represent a single output [width, height, OFM], while the rest represent batch of outputs.
      *                      Data types supported: Same as @p input.
      * @param[in] conv_info Contains padding and stride information described in @ref PadStrideInfo.
+     * @param[in] act_info  (Optional) Activation layer information in case of a fused activation.
      *
      * @return a status
      */
-    static Status validate(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *biases, const ITensorInfo *output, const PadStrideInfo &conv_info);
+    static Status validate(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *biases, const ITensorInfo *output, const PadStrideInfo &conv_info,
+                           const ActivationLayerInfo &act_info = ActivationLayerInfo());
 
     // Inherited methods overridden:
     void run() override;
@@ -76,6 +80,9 @@ public:
 private:
     CLDirectConvolutionLayerKernel _direct_conv_kernel;
     CLFillBorderKernel             _input_border_handler;
+    CLActivationLayer              _activationlayer_function;
+
+    bool _is_activationlayer_enabled;
 };
 }
 #endif /* __ARM_COMPUTE_CLDIRECTCONVOLUTIONLAYER_H__ */
diff --git a/arm_compute/runtime/CL/functions/CLGEMMConvolutionLayer.h b/arm_compute/runtime/CL/functions/CLGEMMConvolutionLayer.h
index 91d35ca0b2..9bfbb536ef 100644
--- a/arm_compute/runtime/CL/functions/CLGEMMConvolutionLayer.h
+++ b/arm_compute/runtime/CL/functions/CLGEMMConvolutionLayer.h
@@ -36,6 +36,7 @@
 #include "arm_compute/core/Types.h"
 #include "arm_compute/runtime/CL/CLMemoryGroup.h"
 #include "arm_compute/runtime/CL/CLTensor.h"
+#include "arm_compute/runtime/CL/functions/CLActivationLayer.h"
 #include "arm_compute/runtime/CL/functions/CLGEMM.h"
 #include "arm_compute/runtime/CL/functions/CLGEMMLowpMatrixMultiplyCore.h"
 #include "arm_compute/runtime/CL/functions/CLGEMMLowpOutputStage.h"
@@ -125,9 +126,10 @@ public:
      * @param[in]  weights_info Specifies if the weights tensor has been reshaped with CLWeightsReshapeKernel. If this is not part of the fully connected layer the weights
      *                          tensor has also been transposed with CLGEMMTranspose1xWKernel. Data type supported: Same as @p input.
      * @param[in]  dilation     (Optional) Dilation, in elements, across x and y. Defaults to (1, 1).
+     * @param[in]  act_info     (Optional) Activation layer information in case of a fused activation.
      */
     void configure(const ICLTensor *input, const ICLTensor *weights, const ICLTensor *biases, ICLTensor *output, const PadStrideInfo &conv_info, const WeightsInfo &weights_info = WeightsInfo(),
-                   const Size2D &dilation = Size2D(1U, 1U));
+                   const Size2D &dilation = Size2D(1U, 1U), const ActivationLayerInfo &act_info = ActivationLayerInfo());
     /** Static function to check if given info will lead to a valid configuration of @ref CLGEMMConvolutionLayer.
      *
      * @param[in]  input        Source tensor. 3 lower dimensions represent a single input [width, height, IFM],
@@ -142,11 +144,12 @@ public:
      * @param[in]  weights_info Specifies if the weights tensor has been reshaped with CLWeightsReshapeKernel. If this is not part of the fully connected layer the weights
      *                          tensor has also been transposed with CLGEMMTranspose1xWKernel. Data type supported: Same as @p input.
      * @param[in]  dilation     (Optional) Dilation, in elements, across x and y. Defaults to (1, 1).
+     * @param[in]  act_info     (Optional) Activation layer information in case of a fused activation.
      *
      * @return a status
      */
     static Status validate(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *biases, const ITensorInfo *output, const PadStrideInfo &conv_info,
-                           const WeightsInfo &weights_info = WeightsInfo(), const Size2D &dilation = Size2D(1U, 1U));
+                           const WeightsInfo &weights_info = WeightsInfo(), const Size2D &dilation = Size2D(1U, 1U), const ActivationLayerInfo &act_info = ActivationLayerInfo());
 
     // Inherited methods overridden:
     void run() override;
@@ -179,6 +182,7 @@ private:
     CLGEMMLowpMatrixMultiplyCore                        _mm_gemmlowp;
     CLGEMMLowpQuantizeDownInt32ToUint8ScaleByFixedPoint _gemmlowp_output_stage;
     CLCol2ImKernel                                      _col2im_kernel;
+    CLActivationLayer                                   _activationlayer_function;
 
     const ICLTensor *_original_weights;
 
@@ -189,6 +193,7 @@ private:
 
     bool _is_quantized;
     bool _is_first_run;
+    bool _is_activationlayer_enabled;
 };
 }
 #endif /* __ARM_COMPUTE_CLGEMMCONVOLUTIONLAYER_H__ */
diff --git a/arm_compute/runtime/CL/functions/CLWinogradConvolutionLayer.h b/arm_compute/runtime/CL/functions/CLWinogradConvolutionLayer.h
index 14de169236..2cf1f77fb4 100644
--- a/arm_compute/runtime/CL/functions/CLWinogradConvolutionLayer.h
+++ b/arm_compute/runtime/CL/functions/CLWinogradConvolutionLayer.h
@@ -27,6 +27,7 @@
 #include "arm_compute/core/CL/kernels/CLWinogradFilterTransformKernel.h"
 #include "arm_compute/core/CL/kernels/CLWinogradOutputTransformKernel.h"
 #include "arm_compute/core/Types.h"
+#include "arm_compute/runtime/CL/functions/CLActivationLayer.h"
 #include "arm_compute/runtime/CL/functions/CLGEMM.h"
 #include "arm_compute/runtime/CL/functions/CLWinogradInputTransform.h"
 #include "arm_compute/runtime/IFunction.h"
@@ -60,8 +61,10 @@ public:
      * @param[out] output    Destination tensor. 3 lower dimensions represent a single output [width, height, OFM], while the rest represent batch of outputs.
      *                       Data types supported: Same as @p input.
      * @param[in]  conv_info Contains padding and stride information described in @ref PadStrideInfo.
+     * @param[in]  act_info  (Optional) Activation layer information in case of a fused activation.
      */
-    void configure(ICLTensor *input, const ICLTensor *weights, const ICLTensor *biases, ICLTensor *output, const PadStrideInfo &conv_info);
+    void configure(ICLTensor *input, const ICLTensor *weights, const ICLTensor *biases, ICLTensor *output, const PadStrideInfo &conv_info,
+                   const ActivationLayerInfo &act_info = ActivationLayerInfo());
     /** Static function to check if given info will lead to a valid configuration of @ref CLWinogradConvolutionLayer
      *
      * @note: This function only works with 3x3 kernels and unit strides
@@ -74,10 +77,12 @@ public:
      * @param[out] output    Destination tensor. 3 lower dimensions represent a single output [width, height, OFM], while the rest represent batch of outputs.
      *                       Data types supported: Same as @p input.
      * @param[in]  conv_info Contains padding and stride information described in @ref PadStrideInfo.
+     * @param[in]  act_info  (Optional) Activation layer information in case of a fused activation.
      *
      * @return a status
      */
-    static Status validate(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *biases, const ITensorInfo *output, const PadStrideInfo &conv_info);
+    static Status validate(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *biases, const ITensorInfo *output, const PadStrideInfo &conv_info,
+                           const ActivationLayerInfo &act_info = ActivationLayerInfo());
 
     // Inherited methods overridden:
     void run() override;
@@ -88,10 +93,12 @@ private:
     CLWinogradInputTransform        _input_transform;
     CLWinogradFilterTransformKernel _filter_transform;
     CLWinogradOutputTransformKernel _output_transform;
+    CLActivationLayer               _activationlayer_function;
     CLTensor                        _input0;
     CLTensor                        _input1;
     CLTensor                        _batched_mm_output;
     bool                            _is_first_run;
+    bool                            _is_activationlayer_enabled;
 };
 }
 #endif /* __ARM_COMPUTE_CLWINOGRADCONVOLUTIONLAYER_H__ */
diff --git a/arm_compute/runtime/GLES_COMPUTE/functions/GCConvolutionLayer.h b/arm_compute/runtime/GLES_COMPUTE/functions/GCConvolutionLayer.h
index 198fa7ba94..54b17b40bb 100644
--- a/arm_compute/runtime/GLES_COMPUTE/functions/GCConvolutionLayer.h
+++ b/arm_compute/runtime/GLES_COMPUTE/functions/GCConvolutionLayer.h
@@ -35,6 +35,7 @@
 #include "arm_compute/core/Types.h"
 #include "arm_compute/runtime/GLES_COMPUTE/GCMemoryGroup.h"
 #include "arm_compute/runtime/GLES_COMPUTE/GCTensor.h"
+#include "arm_compute/runtime/GLES_COMPUTE/functions/GCActivationLayer.h"
 #include "arm_compute/runtime/IFunction.h"
 
 #include <memory>
@@ -100,9 +101,10 @@ public:
      * @param[in]  weights_info Specifies if the weights tensor has been reshaped with GCWeightsReshapeKernel. If this is not part of the fully connected layer the weights
      *                          tensor has also been transposed with GCGEMMTranspose1xWKernel. Data type supported: Same as @p input.
      * @param[in]  dilation     (Optional) Dilation, in elements, across x and y. Defaults to (1, 1).
+     * @param[in]  act_info     (Optional) Activation layer information in case of a fused activation.
      */
-    void configure(const IGCTensor *input, const IGCTensor *weights, const IGCTensor *biases, IGCTensor *output, const PadStrideInfo &conv_info, const WeightsInfo &weights_info = WeightsInfo(),
-                   const Size2D &dilation = Size2D(1U, 1U));
+    void configure(const IGCTensor *input, const IGCTensor *weights, const IGCTensor *biases, IGCTensor *output, const PadStrideInfo &conv_info,
+                   const WeightsInfo &weights_info = WeightsInfo(), const Size2D &dilation = Size2D(1U, 1U), const ActivationLayerInfo &act_info = ActivationLayerInfo());
 
     // Inherited methods overridden:
     void run() override;
@@ -125,6 +127,7 @@ private:
     GCGEMMMatrixMultiplyKernel       _mm_kernel;
     GCCol2ImKernel                   _output_col2im_kernel;
     GCFillBorderKernel               _fill_border;
+    GCActivationLayer                _activationlayer_function;
 
     GCTensor _input_im2col_reshaped;
     GCTensor _input_interleaved_reshaped;
@@ -136,6 +139,7 @@ private:
     bool _append_bias;
     bool _is_fully_connected_convolution;
     bool _are_weights_reshaped;
+    bool _is_activationlayer_enabled;
 };
 }
 
diff --git a/arm_compute/runtime/GLES_COMPUTE/functions/GCDirectConvolutionLayer.h b/arm_compute/runtime/GLES_COMPUTE/functions/GCDirectConvolutionLayer.h
index 976aee7b83..9cfdec0f7e 100644
--- a/arm_compute/runtime/GLES_COMPUTE/functions/GCDirectConvolutionLayer.h
+++ b/arm_compute/runtime/GLES_COMPUTE/functions/GCDirectConvolutionLayer.h
@@ -29,6 +29,7 @@
 #include "arm_compute/core/GLES_COMPUTE/kernels/GCFillBorderKernel.h"
 #include "arm_compute/core/GLES_COMPUTE/kernels/GCTensorShiftKernel.h"
 #include "arm_compute/core/Types.h"
+#include "arm_compute/runtime/GLES_COMPUTE/functions/GCActivationLayer.h"
 #include "arm_compute/runtime/IFunction.h"
 
 #include <memory>
@@ -62,8 +63,10 @@ public:
      * @param[out]    output    Destination tensor. 3 lower dimensions represent a single output [width, height, OFM], while the rest represent batch of outputs.
      *                          Data types supported: Same as @p input.
      * @param[in]     conv_info Contains padding and stride information described in @ref PadStrideInfo.
+     * @param[in]     act_info  (Optional) Activation layer information in case of a fused activation.
      */
-    void configure(IGCTensor *input, const IGCTensor *weights, const IGCTensor *biases, IGCTensor *output, const PadStrideInfo &conv_info);
+    void configure(IGCTensor *input, const IGCTensor *weights, const IGCTensor *biases, IGCTensor *output, const PadStrideInfo &conv_info,
+                   const ActivationLayerInfo &act_info = ActivationLayerInfo());
 
     // Inherited methods overridden:
     void run() override final;
diff --git a/arm_compute/runtime/NEON/functions/NEConvolutionLayer.h b/arm_compute/runtime/NEON/functions/NEConvolutionLayer.h
index 3e6e5abd28..c67951a7ee 100644
--- a/arm_compute/runtime/NEON/functions/NEConvolutionLayer.h
+++ b/arm_compute/runtime/NEON/functions/NEConvolutionLayer.h
@@ -62,9 +62,10 @@ public:
      * @param[in]  weights_info Specifies if the weights tensor has been reshaped with NEWeightsReshapeKernel. If this is not part of the fully connected layer the weights
      *                          tensor has also been transposed with NEGEMMTranspose1xWKernel. Data type supported: Same as @p input.
      * @param[in]  dilation     (Optional) Dilation, in elements, across x and y. Defaults to (1, 1).
+     * @param[in]  act_info     (Optional) Activation layer information in case of a fused activation. Only RELU, BOUNDED_RELU and LU_BOUNDED_RELU supported.
      */
     void configure(ITensor *input, const ITensor *weights, const ITensor *biases, ITensor *output, const PadStrideInfo &conv_info, const WeightsInfo &weights_info = WeightsInfo(),
-                   const Size2D &dilation = Size2D(1U, 1U));
+                   const Size2D &dilation = Size2D(1U, 1U), const ActivationLayerInfo &act_info = ActivationLayerInfo());
     /** Static function to check if given info will lead to a valid configuration of @ref NEConvolutionLayer
      *
      * @param[in] input        Source tensor. 3 lower dimensions represent a single input [width, height, IFM],
@@ -79,11 +80,12 @@ public:
      * @param[in] weights_info Specifies if the weights tensor has been reshaped with NEWeightsReshapeKernel. If this is not part of the fully connected layer the weights
      *                         tensor has also been transposed with NEGEMMTranspose1xWKernel. Data type supported: Same as @p input.
      * @param[in] dilation     (Optional) Dilation, in elements, across x and y. Defaults to (1, 1).
+     * @param[in] act_info     (Optional) Activation layer information in case of a fused activation.
      *
      * @return a status
      */
     static Status validate(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *biases, const ITensorInfo *output, const PadStrideInfo &conv_info,
-                           const WeightsInfo &weights_info = WeightsInfo(), const Size2D &dilation = Size2D(1U, 1U));
+                           const WeightsInfo &weights_info = WeightsInfo(), const Size2D &dilation = Size2D(1U, 1U), const ActivationLayerInfo &act_info = ActivationLayerInfo());
     /** Static function to check if given info will return the convolution called by @ref NEConvolutionLayer
      *
      * @param[in] input        Source tensor. 3 lower dimensions represent a single input [width, height, IFM],
@@ -98,11 +100,12 @@ public:
      * @param[in] weights_info Specifies if the weights tensor has been reshaped with NEWeightsReshapeKernel. If this is not part of the fully connected layer the weights
      *                         tensor has also been transposed with NEGEMMTranspose1xWKernel. Data type supported: Same as @p input.
      * @param[in] dilation     (Optional) Dilation, in elements, across x and y. Defaults to (1, 1).
+     * @param[in] act_info     (Optional) Activation layer information in case of a fused activation.
      *
      * @return the Convolution Method Hint
      */
     static ConvolutionMethod get_convolution_method(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *biases, const ITensorInfo *output, const PadStrideInfo &conv_info,
-                                                    const WeightsInfo &weights_info = WeightsInfo(), const Size2D &dilation = Size2D(1U, 1U));
+                                                    const WeightsInfo &weights_info = WeightsInfo(), const Size2D &dilation = Size2D(1U, 1U), const ActivationLayerInfo &act_info = ActivationLayerInfo());
 
     // Inherited methods overridden:
     void run() override;
diff --git a/arm_compute/runtime/NEON/functions/NEDirectConvolutionLayer.h b/arm_compute/runtime/NEON/functions/NEDirectConvolutionLayer.h
index e1aa839802..1eaad5cda6 100644
--- a/arm_compute/runtime/NEON/functions/NEDirectConvolutionLayer.h
+++ b/arm_compute/runtime/NEON/functions/NEDirectConvolutionLayer.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017 ARM Limited.
+ * Copyright (c) 2017-2018 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -31,6 +31,7 @@
 #include "arm_compute/runtime/IFunction.h"
 #include "arm_compute/runtime/IMemoryManager.h"
 #include "arm_compute/runtime/MemoryGroup.h"
+#include "arm_compute/runtime/NEON/functions/NEActivationLayer.h"
 #include "arm_compute/runtime/Tensor.h"
 
 #include <memory>
@@ -66,8 +67,9 @@ public:
      * @param[out]     output    Output tensor.
      *                           The 3rd dimensions must be equal to the 4th dimension of the @p kernels tensor. Data types supported: Same as @p input.
      * @param[in]      conv_info Contains padding and stride information described in @ref PadStrideInfo.
+     * @param[in]      act_info  (Optional) Activation layer information in case of a fused activation.
      */
-    void configure(ITensor *input, const ITensor *weights, const ITensor *bias, ITensor *output, const PadStrideInfo &conv_info);
+    void configure(ITensor *input, const ITensor *weights, const ITensor *bias, ITensor *output, const PadStrideInfo &conv_info, const ActivationLayerInfo &act_info = ActivationLayerInfo());
     /** Static function to check if given info will lead to a valid configuration of @ref NEDirectConvolutionLayer
      *
      * @note: DirectConvolution only works in the following configurations:
@@ -84,10 +86,12 @@ public:
      * @param[in] output    Output tensor.
      *                      The 3rd dimensions must be equal to the 4th dimension of the @p kernels tensor. Data types supported: Same as @p input.
      * @param[in] conv_info Contains padding and stride information described in @ref PadStrideInfo.
+     * @param[in] act_info  (Optional) Activation layer information in case of a fused activation.
      *
      * @return a status
      */
-    static Status validate(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *bias, const ITensorInfo *output, const PadStrideInfo &conv_info);
+    static Status validate(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *bias, const ITensorInfo *output, const PadStrideInfo &conv_info,
+                           const ActivationLayerInfo &act_info = ActivationLayerInfo());
 
     // Inherited methods overridden:
     void run() override;
@@ -97,9 +101,11 @@ private:
     NEDirectConvolutionLayerOutputStageKernel _output_stage_kernel;
     NEDirectConvolutionLayerKernel            _conv_kernel;
     NEFillBorderKernel                        _input_border_handler;
+    NEActivationLayer                         _activationlayer_function;
     Tensor                                    _accumulator;
     bool                                      _has_bias;
     bool                                      _is_fixed_point;
+    bool                                      _is_activationlayer_enabled;
 };
 }
 #endif /* __ARM_COMPUTE_NEDIRECTCONVOLUTIONLAYER_H__ */
diff --git a/arm_compute/runtime/NEON/functions/NEGEMMConvolutionLayer.h b/arm_compute/runtime/NEON/functions/NEGEMMConvolutionLayer.h
index e733fec4b6..24e23f133a 100644
--- a/arm_compute/runtime/NEON/functions/NEGEMMConvolutionLayer.h
+++ b/arm_compute/runtime/NEON/functions/NEGEMMConvolutionLayer.h
@@ -37,6 +37,7 @@
 #include "arm_compute/core/Types.h"
 #include "arm_compute/runtime/MemoryGroup.h"
 #include "arm_compute/runtime/NEON/AssemblyHelper.h"
+#include "arm_compute/runtime/NEON/functions/NEActivationLayer.h"
 #include "arm_compute/runtime/NEON/functions/NEGEMMLowpMatrixMultiplyCore.h"
 #include "arm_compute/runtime/NEON/functions/NEGEMMLowpOutputStage.h"
 #include "arm_compute/runtime/Tensor.h"
@@ -95,6 +96,7 @@ private:
  * -# @ref NEGEMMMatrixMultiplyKernel or @ref NEGEMMLowpMatrixMultiplyCore (if quantized asymmetric)
  * -# @ref NEGEMMLowpQuantizeDownInt32ToUint8Scale (if quantized asymmetric)
  * -# @ref NECol2ImKernel
+ * -# @ref NEActivationLayer (executed only if the activation layer is enabled)
  */
 class NEGEMMConvolutionLayer : public IFunction
 {
@@ -123,9 +125,10 @@ public:
      * @param[in]  weights_info Specifies if the weights tensor has been reshaped with NEWeightsReshapeKernel. If this is not part of the fully connected layer the weights
      *                          tensor has also been transposed with NEGEMMTranspose1xWKernel. Data type supported: Same as @p input.
      * @param[in]  dilation     (Optional) Dilation, in elements, across x and y. Defaults to (1, 1).
+     * @param[in]  act_info     (Optional) Activation layer information in case of a fused activation. Only RELU, BOUNDED_RELU and LU_BOUNDED_RELU supported.
      */
     void configure(const ITensor *input, const ITensor *weights, const ITensor *biases, ITensor *output, const PadStrideInfo &conv_info, const WeightsInfo &weights_info = WeightsInfo(),
-                   const Size2D &dilation = Size2D(1U, 1U));
+                   const Size2D &dilation = Size2D(1U, 1U), const ActivationLayerInfo &act_info = ActivationLayerInfo());
     /** Static function to check if given info will lead to a valid configuration of @ref NEGEMMConvolutionLayer
      *
      * @param[in] input        Source tensor. 3 lower dimensions represent a single input [width, height, IFM],
@@ -140,11 +143,12 @@ public:
      * @param[in] weights_info Specifies if the weights tensor has been reshaped with NEWeightsReshapeKernel. If this is not part of the fully connected layer the weights
      *                         tensor has also been transposed with NEGEMMTranspose1xWKernel. Data type supported: Same as @p input.
      * @param[in] dilation     (Optional) Dilation, in elements, across x and y. Defaults to (1, 1).
+     * @param[in] act_info     (Optional) Activation layer information in case of a fused activation. Only RELU, BOUNDED_RELU and LU_BOUNDED_RELU supported.
      *
      * @return a status
      */
     static Status validate(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *biases, const ITensorInfo *output, const PadStrideInfo &conv_info,
-                           const WeightsInfo &weights_info = WeightsInfo(), const Size2D &dilation = Size2D(1U, 1U));
+                           const WeightsInfo &weights_info = WeightsInfo(), const Size2D &dilation = Size2D(1U, 1U), const ActivationLayerInfo &act_info = ActivationLayerInfo());
 
     // Inherited methods overridden:
     void run() override;
@@ -171,6 +175,7 @@ private:
     NEGEMMLowpMatrixMultiplyCore                        _mm_gemmlowp;
     NEGEMMLowpQuantizeDownInt32ToUint8ScaleByFixedPoint _gemmlowp_output_stage;
     NECol2ImKernel                                      _output_col2im_kernel;
+    NEActivationLayer                                   _activationlayer_function;
 
     const ITensor *_original_weights;
 
@@ -186,6 +191,7 @@ private:
     bool _are_weights_reshaped;
     bool _is_quantized;
     bool _is_interleaved;
+    bool _is_activationlayer_enabled;
 };
 }
 #endif /* __ARM_COMPUTE_NECONVOLUTIONGEMMLAYER_H__ */
diff --git a/arm_compute/runtime/NEON/functions/NEWinogradLayer.h b/arm_compute/runtime/NEON/functions/NEWinogradLayer.h
index a939f82854..61a4caae3a 100644
--- a/arm_compute/runtime/NEON/functions/NEWinogradLayer.h
+++ b/arm_compute/runtime/NEON/functions/NEWinogradLayer.h
@@ -30,6 +30,7 @@
 #include "arm_compute/core/Types.h"
 #include "arm_compute/runtime/CPP/functions/CPPPermute.h"
 #include "arm_compute/runtime/MemoryGroup.h"
+#include "arm_compute/runtime/NEON/functions/NEActivationLayer.h"
 #include "arm_compute/runtime/Tensor.h"
 
 #include <memory>
@@ -61,8 +62,9 @@ public:
      * @param[out] output    Destination tensor. 3 lower dimensions represent a single output [width, height, OFM], while the rest represent batch of outputs.
      *                       Data types supported: Same as @p input.
      * @param[in]  conv_info Contains padding and stride information described in @ref PadStrideInfo. Currently only unit strides are supported.
+     * @param[in]  act_info  (Optional) Activation layer information in case of a fused activation.
      */
-    void configure(const ITensor *input, const ITensor *weights, const ITensor *biases, ITensor *output, const PadStrideInfo &conv_info);
+    void configure(const ITensor *input, const ITensor *weights, const ITensor *biases, ITensor *output, const PadStrideInfo &conv_info, const ActivationLayerInfo &act_info = ActivationLayerInfo());
 
     // Inherited methods overridden:
     void run() override;
@@ -94,6 +96,7 @@ private:
     std::unique_ptr<INEKernel> _transform_input_kernel;
     std::unique_ptr<INEKernel> _transform_output_kernel;
     std::unique_ptr<INEKernel> _transform_weights_kernel;
+    NEActivationLayer          _activationlayer_function;
 
     CPPPermute     _permute_input;
     CPPPermute     _permute_weights;
@@ -108,6 +111,7 @@ private:
     const ITensor *_weights;
     ITensor       *_output;
     bool           _reshaped_kernel;
+    bool           _is_activationlayer_enabled;
 };
 }
 #endif /* __ARM_COMPUTE_NEWINOGRADLAYER_H__ */