Make CLFloor and CLActivation kernels and functions state-less

Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com>
Change-Id: Icbe4e6a7c6732a59bdda0136af44c4852452dfd1
Signed-off-by: Michele Di Giorgio <michele.digiorgio@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/4891
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>

17 files changed, 509 insertions, 307 deletions

diff --git a/Android.bp b/Android.bp
index 8d6182f820..30ed0acf17 100644
--- a/Android.bp
+++ b/Android.bp
@@ -80,7 +80,6 @@ cc_library_static {
         "src/core/CL/gemm/reshaped_only_rhs/CLGEMMDefaultConfigReshapedRHSOnlyValhall.cpp",
         "src/core/CL/kernels/CLAbsoluteDifferenceKernel.cpp",
         "src/core/CL/kernels/CLAccumulateKernel.cpp",
-        "src/core/CL/kernels/CLActivationLayerKernel.cpp",
         "src/core/CL/kernels/CLArgMinMaxLayerKernel.cpp",
         "src/core/CL/kernels/CLBatchNormalizationLayerKernel.cpp",
         "src/core/CL/kernels/CLBatchToSpaceLayerKernel.cpp",
@@ -118,7 +117,6 @@ cc_library_static {
         "src/core/CL/kernels/CLFFTScaleKernel.cpp",
         "src/core/CL/kernels/CLFastCornersKernel.cpp",
         "src/core/CL/kernels/CLFillBorderKernel.cpp",
-        "src/core/CL/kernels/CLFloorKernel.cpp",
         "src/core/CL/kernels/CLFuseBatchNormalizationKernel.cpp",
         "src/core/CL/kernels/CLGEMMLowpMatrixMultiplyNativeKernel.cpp",
         "src/core/CL/kernels/CLGEMMLowpMatrixMultiplyReshapedKernel.cpp",
@@ -443,8 +441,10 @@ cc_library_static {
         "src/core/cpu/kernels/sub/neon/qasymm8.cpp",
         "src/core/cpu/kernels/sub/neon/qasymm8_signed.cpp",
         "src/core/cpu/kernels/sub/neon/qsymm16.cpp",
+        "src/core/gpu/cl/kernels/ClActivationKernel.cpp",
         "src/core/gpu/cl/kernels/ClBatchConcatenateKernel.cpp",
         "src/core/gpu/cl/kernels/ClDepthConcatenateKernel.cpp",
+        "src/core/gpu/cl/kernels/ClFloorKernel.cpp",
         "src/core/gpu/cl/kernels/ClHeightConcatenateKernel.cpp",
         "src/core/gpu/cl/kernels/ClWidthConcatenate2TensorsKernel.cpp",
         "src/core/gpu/cl/kernels/ClWidthConcatenate4TensorsKernel.cpp",
@@ -795,7 +795,9 @@ cc_library_static {
         "src/runtime/cpu/operators/CpuPermute.cpp",
         "src/runtime/cpu/operators/CpuReshape.cpp",
         "src/runtime/cpu/operators/CpuSub.cpp",
+        "src/runtime/gpu/cl/operators/ClActivation.cpp",
         "src/runtime/gpu/cl/operators/ClConcatenate.cpp",
+        "src/runtime/gpu/cl/operators/ClFloor.cpp",
         "utils/CommonGraphOptions.cpp",
         "utils/GraphUtils.cpp",
         "utils/Utils.cpp",
diff --git a/arm_compute/runtime/CL/functions/CLActivationLayer.h b/arm_compute/runtime/CL/functions/CLActivationLayer.h
index dc2cb62b71..aa62183909 100644
--- a/arm_compute/runtime/CL/functions/CLActivationLayer.h
+++ b/arm_compute/runtime/CL/functions/CLActivationLayer.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-2020 Arm Limited.
+ * Copyright (c) 2016-2021 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -24,10 +24,10 @@
 #ifndef ARM_COMPUTE_CLACTIVATIONLAYER_H
 #define ARM_COMPUTE_CLACTIVATIONLAYER_H
 
-#include "arm_compute/runtime/CL/ICLOperator.h"
-#include "arm_compute/runtime/CL/ICLSimpleFunction.h"
+#include "arm_compute/runtime/IFunction.h"
 
 #include "arm_compute/core/Types.h"
+#include "arm_compute/runtime/CL/CLRuntimeContext.h"
 
 namespace arm_compute
 {
@@ -35,7 +35,7 @@ class CLCompileContext;
 class ICLTensor;
 class ITensorInfo;
 
-/** Basic function to run @ref CLActivationLayerKernel
+/** Basic function to run @ref opencl::kernels::ClActivationKernel
  *
  * @note The function simulates an activation layer with the specified activation function.
  */
@@ -96,33 +96,5 @@ private:
     struct Impl;
     std::unique_ptr<Impl> _impl;
 };
-
-namespace experimental
-{
-/** Basic function to run @ref CLActivationLayerKernel */
-class CLActivation : public ICLOperator
-{
-public:
-    /** Set the input and output tensor.
-     *
-     * @param[in]      compile_context The compile context to be used.
-     * @param[in, out] input           Source tensor info. In case of @p output tensor = nullptr, this tensor will store the result
-     *                                 of the activation function. Data types supported: QASYMM8/QASYMM8_SIGNED/QSYMM16/F16/F32.
-     * @param[out]     output          Destination tensor info. Data type supported: same as @p input
-     * @param[in]      act_info        Activation layer parameters.
-     */
-    void configure(const CLCompileContext &compile_context, ITensorInfo *input, ITensorInfo *output, ActivationLayerInfo act_info);
-    /** Static function to check if given info will lead to a valid configuration of @ref CLActivationLayer
-     *
-     * @param[in] input    Source tensor info. In case of @p output tensor info = nullptr, this tensor will store the result
-     *                     of the activation function. Data types supported: QASYMM8/QASYMM8_SIGNED/QSYMM16/F16/F32.
-     * @param[in] output   Destination tensor info. Data type supported: same as @p input
-     * @param[in] act_info Activation layer information.
-     *
-     * @return a status
-     */
-    static Status validate(const ITensorInfo *input, const ITensorInfo *output, const ActivationLayerInfo &act_info);
-};
-} // namespace experimental
 } // namespace arm_compute
 #endif /* ARM_COMPUTE_CLACTIVATIONLAYER_H */
diff --git a/arm_compute/runtime/CL/functions/CLFloor.h b/arm_compute/runtime/CL/functions/CLFloor.h
index 85d7071194..7c7607002d 100644
--- a/arm_compute/runtime/CL/functions/CLFloor.h
+++ b/arm_compute/runtime/CL/functions/CLFloor.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017-2020 Arm Limited.
+ * Copyright (c) 2017-2021 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -24,20 +24,34 @@
 #ifndef ARM_COMPUTE_CLFLOOR_H
 #define ARM_COMPUTE_CLFLOOR_H
 
-#include "arm_compute/runtime/CL/ICLSimpleFunction.h"
+#include "arm_compute/runtime/IFunction.h"
 
 #include "arm_compute/core/Types.h"
 
+#include <memory>
+
 namespace arm_compute
 {
 class CLCompileContext;
 class ICLTensor;
 class ITensorInfo;
 
-/** Basic function to run @ref CLFloorKernel */
-class CLFloor : public ICLSimpleFunction
+/** Basic function to run @ref opencl::kernels::ClFloorKernel */
+class CLFloor : public IFunction
 {
 public:
+    /** Constructor */
+    CLFloor();
+    /** Destructor */
+    ~CLFloor();
+    /** Prevent instances of this class from being copied (As this class contains pointers) */
+    CLFloor(const CLFloor &) = delete;
+    /** Default move constructor */
+    CLFloor(CLFloor &&);
+    /** Prevent instances of this class from being copied (As this class contains pointers) */
+    CLFloor &operator=(const CLFloor &) = delete;
+    /** Default move assignment operator */
+    CLFloor &operator=(CLFloor &&);
     /** Set the source, destination of the kernel
      *
      * @param[in]  input  Source tensor. Data type supported: F16/F32.
@@ -59,6 +73,13 @@ public:
      * @return a status
      */
     static Status validate(const ITensorInfo *input, const ITensorInfo *output);
+
+    // Inherited methods overridden:
+    void run() override;
+
+private:
+    struct Impl;
+    std::unique_ptr<Impl> _impl;
 };
-}
+} // namespace arm_compute
 #endif /* ARM_COMPUTE_CLFLOOR_H */
diff --git a/docs/00_introduction.dox b/docs/00_introduction.dox
index af78a70abc..cbdc983159 100644
--- a/docs/00_introduction.dox
+++ b/docs/00_introduction.dox
@@ -182,7 +182,7 @@ v20.11 Public major release
    - @ref CLDepthConvertLayerKernel
    - @ref CLCopyKernel
    - @ref CLDepthwiseConvolutionLayer3x3NHWCKernel
-   - @ref CLActivationLayerKernel
+   - CLActivationLayerKernel
    - @ref CLWinogradFilterTransformKernel
    - CLWidthConcatenateLayerKernel
    - CLWidthConcatenate4TensorsKernel
@@ -535,7 +535,7 @@ v20.05 Public major release
      - @ref NEQLSTMLayer
      - @ref NEQLSTMLayerNormalizationKernel
  - Added HARD_SWISH support in:
-     - @ref CLActivationLayerKernel
+     - CLActivationLayerKernel
      - NEActivationLayerKernel
  - Deprecated OpenCL kernels / functions:
      - CLGEMMLowpQuantizeDownInt32ToUint8Scale
@@ -1200,7 +1200,7 @@ v17.09 Public major release
     - @ref CLDequantizationLayerKernel / @ref CLDequantizationLayer
     - @ref CLDirectConvolutionLayerKernel / @ref CLDirectConvolutionLayer
     - CLFlattenLayer
-    - @ref CLFloorKernel / @ref CLFloor
+    - CLFloorKernel / @ref CLFloor
     - CLGEMMTranspose1xW
     - CLGEMMMatrixVectorMultiplyKernel
     - @ref CLL2NormalizeLayerKernel / @ref CLL2NormalizeLayer
@@ -1306,7 +1306,7 @@ v17.02.1 Sources preview
 
 v17.02 Sources preview
  - New OpenCL kernels / functions:
-   - @ref CLActivationLayerKernel / @ref CLActivationLayer
+   - CLActivationLayerKernel / @ref CLActivationLayer
    - @ref CLChannelCombineKernel / @ref CLChannelCombine
    - @ref CLDerivativeKernel / @ref CLChannelExtract
    - @ref CLFastCornersKernel / @ref CLFastCorners
diff --git a/src/core/CL/CLKernels.h b/src/core/CL/CLKernels.h
index 11f1d2d7cf..76bc5271b4 100644
--- a/src/core/CL/CLKernels.h
+++ b/src/core/CL/CLKernels.h
@@ -27,7 +27,6 @@
 /* Header regrouping all the CL kernels */
 #include "src/core/CL/kernels/CLAbsoluteDifferenceKernel.h"
 #include "src/core/CL/kernels/CLAccumulateKernel.h"
-#include "src/core/CL/kernels/CLActivationLayerKernel.h"
 #include "src/core/CL/kernels/CLArgMinMaxLayerKernel.h"
 #include "src/core/CL/kernels/CLBatchNormalizationLayerKernel.h"
 #include "src/core/CL/kernels/CLBatchToSpaceLayerKernel.h"
@@ -65,7 +64,6 @@
 #include "src/core/CL/kernels/CLFFTScaleKernel.h"
 #include "src/core/CL/kernels/CLFastCornersKernel.h"
 #include "src/core/CL/kernels/CLFillBorderKernel.h"
-#include "src/core/CL/kernels/CLFloorKernel.h"
 #include "src/core/CL/kernels/CLFuseBatchNormalizationKernel.h"
 #include "src/core/CL/kernels/CLGEMMLowpMatrixMultiplyNativeKernel.h"
 #include "src/core/CL/kernels/CLGEMMLowpMatrixMultiplyReshapedKernel.h"
diff --git a/src/core/CL/kernels/CLActivationLayerKernel.h b/src/core/CL/kernels/CLActivationLayerKernel.h
deleted file mode 100644
index 821418f835..0000000000
--- a/src/core/CL/kernels/CLActivationLayerKernel.h
+++ /dev/null
@@ -1,77 +0,0 @@
-/*
- * Copyright (c) 2016-2020 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.
- */
-#ifndef ARM_COMPUTE_CLACTIVATIONLAYERKERNEL_H
-#define ARM_COMPUTE_CLACTIVATIONLAYERKERNEL_H
-
-#include "src/core/CL/ICLKernel.h"
-
-namespace arm_compute
-{
-class ICLTensor;
-/** Interface for the activation layer kernel. */
-class CLActivationLayerKernel : public ICLKernel
-{
-public:
-    /** Default constructor */
-    CLActivationLayerKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLActivationLayerKernel(const CLActivationLayerKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLActivationLayerKernel &operator=(const CLActivationLayerKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLActivationLayerKernel(CLActivationLayerKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLActivationLayerKernel &operator=(CLActivationLayerKernel &&) = default;
-    /** Default destructor */
-    ~CLActivationLayerKernel() = default;
-    /** Set the input and output tensor.
-     *
-     * @note If the output tensor is a nullptr, the activation function will be performed in-place
-     *
-     * @param[in]      compile_context The compile context to be used.
-     * @param[in, out] input           Source tensor. In case of @p output tensor = nullptr, this tensor will store the result
-     *                                 of the activation function. Data types supported: QASYMM8/QASYMM8_SIGNED/QSYMM16/F16/F32.
-     * @param[out]     output          Destination tensor. Data type supported: same as @p input
-     * @param[in]      act_info        Activation layer information.
-     */
-    void configure(const CLCompileContext &compile_context, ITensorInfo *input, ITensorInfo *output, ActivationLayerInfo act_info);
-    /** Static function to check if given info will lead to a valid configuration of @ref CLActivationLayerKernel
-     *
-     * @param[in] input    Source tensor info. In case of @p output tensor info = nullptr, this tensor will store the result
-     *                     of the activation function. Data types supported: QASYMM8/QASYMM8_SIGNED/QSYMM16/F16/F32.
-     * @param[in] output   Destination tensor info. Data type supported: same as @p input
-     * @param[in] act_info Activation layer information.
-     *
-     * @return a status
-     */
-    static Status validate(const ITensorInfo *input, const ITensorInfo *output, const ActivationLayerInfo &act_info);
-
-    // Inherited methods overridden:
-    void run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue) override;
-
-private:
-    bool _run_in_place;
-};
-} // namespace arm_compute
-#endif /*ARM_COMPUTE_CLACTIVATIONLAYERKERNEL_H */
diff --git a/src/core/CL/kernels/CLFloorKernel.h b/src/core/CL/kernels/CLFloorKernel.h
deleted file mode 100644
index f5635141e4..0000000000
--- a/src/core/CL/kernels/CLFloorKernel.h
+++ /dev/null
@@ -1,81 +0,0 @@
-/*
- * Copyright (c) 2017-2020 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.
- */
-#ifndef ARM_COMPUTE_CLFLOORKERNEL_H
-#define ARM_COMPUTE_CLFLOORKERNEL_H
-
-#include "src/core/CL/ICLKernel.h"
-
-namespace arm_compute
-{
-class ICLTensor;
-
-/** OpenCL kernel to perform a floor operation */
-class CLFloorKernel : public ICLKernel
-{
-public:
-    /** Default constructor */
-    CLFloorKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLFloorKernel(const CLFloorKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLFloorKernel &operator=(const CLFloorKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLFloorKernel(CLFloorKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLFloorKernel &operator=(CLFloorKernel &&) = default;
-    /** Default destructor */
-    ~CLFloorKernel() = default;
-    /** Set the source, destination of the kernel
-     *
-     * @param[in]  input  Source tensor. Data type supported: F16/F32.
-     * @param[out] output Destination tensor. Same as @p input
-     */
-    void configure(const ICLTensor *input, ICLTensor *output);
-
-    /** Set the source, destination of the kernel
-     *
-     * @param[in]  compile_context The compile context to be used.
-     * @param[in]  input           Source tensor. Data type supported: F16/F32.
-     * @param[out] output          Destination tensor. Same as @p input
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output);
-
-    /** Static function to check if given info will lead to a valid configuration of @ref CLFloorKernel
-     *
-     * @param[in] input  Source tensor info. Data type supported: F16/F32.
-     * @param[in] output Destination tensor info. Same as @p input
-     *
-     * @return a status
-     */
-    static Status validate(const ITensorInfo *input, const ITensorInfo *output);
-
-    // Inherited methods overridden:
-    void run(const Window &window, cl::CommandQueue &queue) override;
-
-private:
-    const ICLTensor *_input;
-    ICLTensor       *_output;
-};
-} // namespace arm_compute
-#endif /*ARM_COMPUTE_CLFLOORKERNEL_H */
diff --git a/src/core/CL/kernels/CLActivationLayerKernel.cpp b/src/core/gpu/cl/kernels/ClActivationKernel.cpp
index 00b6829e91..f48425f7cd 100644
--- a/src/core/CL/kernels/CLActivationLayerKernel.cpp
+++ b/src/core/gpu/cl/kernels/ClActivationKernel.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-2020 Arm Limited.
+ * Copyright (c) 2016-2021 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -21,7 +21,7 @@
  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
  */
-#include "src/core/CL/kernels/CLActivationLayerKernel.h"
+#include "src/core/gpu/cl/kernels/ClActivationKernel.h"
 
 #include "arm_compute/core/CL/CLCoreRuntimeContext.h"
 #include "arm_compute/core/CL/CLHelpers.h"
@@ -39,12 +39,16 @@
 
 namespace arm_compute
 {
+namespace opencl
+{
+namespace kernels
+{
 namespace
 {
-Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output, const ActivationLayerInfo &act_info)
+Status validate_arguments(const ITensorInfo *src, const ITensorInfo *dst, const ActivationLayerInfo &act_info)
 {
-    ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(input);
-    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::QASYMM8, DataType::QASYMM8_SIGNED, DataType::QSYMM16, DataType::F16, DataType::F32);
+    ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(src);
+    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(src, 1, DataType::QASYMM8, DataType::QASYMM8_SIGNED, DataType::QSYMM16, DataType::F16, DataType::F32);
 
     static std::set<ActivationLayerInfo::ActivationFunction> quantized_supported_activations =
     {
@@ -56,8 +60,8 @@ Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output, c
         ActivationLayerInfo::ActivationFunction::HARD_SWISH,
         ActivationLayerInfo::ActivationFunction::LEAKY_RELU,
     };
-    const DataType                                data_type = input->data_type();
-    const QuantizationInfo                       &oq_info   = (output != nullptr) ? output->quantization_info() : input->quantization_info();
+    const DataType                                data_type = src->data_type();
+    const QuantizationInfo                       &oq_info   = (dst != nullptr) ? dst->quantization_info() : src->quantization_info();
     const ActivationLayerInfo::ActivationFunction f_act     = act_info.activation();
 
     ARM_COMPUTE_RETURN_ERROR_ON_MSG(is_data_type_quantized(data_type) && (quantized_supported_activations.count(f_act) == 0),
@@ -72,41 +76,41 @@ Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output, c
     ARM_COMPUTE_RETURN_ERROR_ON(data_type == DataType::QASYMM8_SIGNED && (f_act == ActivationLayerInfo::ActivationFunction::TANH) && (oq_info != QuantizationInfo(1.f / 128.f, 0)));
     ARM_COMPUTE_RETURN_ERROR_ON(data_type == DataType::QASYMM8_SIGNED && (f_act == ActivationLayerInfo::ActivationFunction::LOGISTIC) && (oq_info != QuantizationInfo(1.f / 256.f, -128)));
 
-    // Checks performed when output is configured
-    if((output != nullptr) && (output->total_size() != 0))
+    // Checks performed when destination is configured
+    if((dst != nullptr) && (dst->total_size() != 0))
     {
-        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(input, output);
-        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
+        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(src, dst);
+        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(src, dst);
     }
 
     return Status{};
 }
 } // namespace
 
-CLActivationLayerKernel::CLActivationLayerKernel()
+ClActivationKernel::ClActivationKernel()
     : _run_in_place(false)
 {
 }
 
-void CLActivationLayerKernel::configure(const CLCompileContext &compile_context, ITensorInfo *input, ITensorInfo *output, ActivationLayerInfo act_info)
+void ClActivationKernel::configure(const ClCompileContext &compile_context, ITensorInfo *src, ITensorInfo *dst, ActivationLayerInfo act_info)
 {
-    ARM_COMPUTE_ERROR_ON_NULLPTR(input);
+    ARM_COMPUTE_ERROR_ON_NULLPTR(src);
 
-    auto padding_info = get_padding_info({ input, output });
+    auto padding_info = get_padding_info({ src, dst });
 
-    _run_in_place = (output == nullptr) || (output == input);
+    _run_in_place = (dst == nullptr) || (dst == src);
 
-    if(output != nullptr)
+    if(dst != nullptr)
     {
-        // Output auto inizialitation if not yet initialized
-        auto_init_if_empty(*output, *input->clone());
+        // Destination auto inizialitation if not yet initialized
+        auto_init_if_empty(*dst, *src->clone());
     }
 
-    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input, (output != nullptr) ? output : nullptr, act_info));
+    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(src, (dst != nullptr) ? dst : nullptr, act_info));
 
-    const unsigned int num_elems_processed_per_iteration = adjust_vec_size(16 / input->element_size(), input->dimension(0));
+    const unsigned int num_elems_processed_per_iteration = adjust_vec_size(16 / src->element_size(), src->dimension(0));
 
-    const DataType dt      = input->data_type();
+    const DataType dt      = src->data_type();
     float          a_const = act_info.a();
     float          b_const = act_info.b();
 
@@ -125,14 +129,14 @@ void CLActivationLayerKernel::configure(const CLCompileContext &compile_context,
     build_opts.add_option("-DACT=" + lower_string(string_from_activation_func(f_act)));
     build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(dt));
     build_opts.add_option("-DVEC_SIZE=" + support::cpp11::to_string(num_elems_processed_per_iteration));
-    build_opts.add_option("-DVEC_SIZE_LEFTOVER=" + support::cpp11::to_string(input->dimension(0) % num_elems_processed_per_iteration));
+    build_opts.add_option("-DVEC_SIZE_LEFTOVER=" + support::cpp11::to_string(src->dimension(0) % num_elems_processed_per_iteration));
 
     std::string kernel_name = std::string("activation_layer");
 
     // Set quantization info build options
     if(is_quantized)
     {
-        const UniformQuantizationInfo iq_info = input->quantization_info().uniform();
+        const UniformQuantizationInfo iq_info = src->quantization_info().uniform();
 
         if(!perform_activation_in_float)
         {
@@ -180,10 +184,10 @@ void CLActivationLayerKernel::configure(const CLCompileContext &compile_context,
         // Set correct kernel name
         kernel_name += perform_activation_in_float ? std::string("_quant_f32") : std::string("_quant");
 
-        // Set scale and offset of the input and output if they have different quantization info
-        if(output != nullptr)
+        // Set scale and offset of the source and destination if they have different quantization info
+        if(dst != nullptr)
         {
-            const UniformQuantizationInfo oq_info = output->quantization_info().uniform();
+            const UniformQuantizationInfo oq_info = dst->quantization_info().uniform();
 
             if(iq_info != oq_info)
             {
@@ -203,27 +207,27 @@ void CLActivationLayerKernel::configure(const CLCompileContext &compile_context,
     _kernel = create_kernel(compile_context, kernel_name, build_opts.options());
 
     // Configure kernel window
-    Window win = calculate_max_window(*input, Steps(num_elems_processed_per_iteration));
+    Window win = calculate_max_window(*src, Steps(num_elems_processed_per_iteration));
     ICLKernel::configure_internal(win);
 
     // Set config_id for enabling LWS tuning
     _config_id = "activation_layer_";
     _config_id += lower_string(string_from_data_type(dt));
     _config_id += "_";
-    _config_id += support::cpp11::to_string(input->dimension(0));
+    _config_id += support::cpp11::to_string(src->dimension(0));
     _config_id += "_";
-    _config_id += support::cpp11::to_string(input->dimension(1));
+    _config_id += support::cpp11::to_string(src->dimension(1));
 
     ARM_COMPUTE_ERROR_ON(has_padding_changed(padding_info));
 }
 
-Status CLActivationLayerKernel::validate(const ITensorInfo *input, const ITensorInfo *output, const ActivationLayerInfo &act_info)
+Status ClActivationKernel::validate(const ITensorInfo *src, const ITensorInfo *dst, const ActivationLayerInfo &act_info)
 {
-    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, output, act_info));
+    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(src, dst, act_info));
     return Status{};
 }
 
-void CLActivationLayerKernel::run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue)
+void ClActivationKernel::run_op(ITensorPack &tensors, const Window &window, ::cl::CommandQueue &queue)
 {
     ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
     ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
@@ -247,4 +251,6 @@ void CLActivationLayerKernel::run_op(ITensorPack &tensors, const Window &window,
     }
     while(collapsed.slide_window_slice_3D(slice));
 }
+} // namespace kernels
+} // namespace opencl
 } // namespace arm_compute
diff --git a/src/core/gpu/cl/kernels/ClActivationKernel.h b/src/core/gpu/cl/kernels/ClActivationKernel.h
new file mode 100644
index 0000000000..30adc552b3
--- /dev/null
+++ b/src/core/gpu/cl/kernels/ClActivationKernel.h
@@ -0,0 +1,74 @@
+/*
+ * Copyright (c) 2016-2021 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.
+ */
+#ifndef ARM_COMPUTE_CL_ACTIVATION_KERNEL_H
+#define ARM_COMPUTE_CL_ACTIVATION_KERNEL_H
+
+#include "src/core/common/Macros.h"
+#include "src/core/gpu/cl/ClCompileContext.h"
+#include "src/core/gpu/cl/IClKernel.h"
+
+namespace arm_compute
+{
+namespace opencl
+{
+namespace kernels
+{
+/** Interface for the activation kernel. */
+class ClActivationKernel : public IClKernel
+{
+public:
+    ClActivationKernel();
+    ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE(ClActivationKernel);
+    /** Configure kernel for a given list of arguments
+     *
+     * @note If the output tensor is a nullptr, the activation function will be performed in-place
+     *
+     * @param[in]      compile_context The compile context to be used.
+     * @param[in, out] src             Source tensor. In case of @p dst tensor = nullptr, this tensor will store the result
+     *                                 of the activation function. Data types supported: QASYMM8/QASYMM8_SIGNED/QSYMM16/F16/F32.
+     * @param[out]     dst             Destination tensor. Data type supported: same as @p src
+     * @param[in]      act_info        Activation layer information.
+     */
+    void configure(const ClCompileContext &compile_context, ITensorInfo *src, ITensorInfo *dst, ActivationLayerInfo act_info);
+    /** Static function to check if given info will lead to a valid configuration of @ref ClActivationKernel
+     *
+     * @param[in] src      Source tensor info. In case of @p dst tensor info = nullptr, this tensor will store the result
+     *                     of the activation function. Data types supported: QASYMM8/QASYMM8_SIGNED/QSYMM16/F16/F32.
+     * @param[in] dst      Destination tensor info. Data type supported: same as @p src
+     * @param[in] act_info Activation layer information.
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *src, const ITensorInfo *dst, const ActivationLayerInfo &act_info);
+
+    // Inherited methods overridden:
+    void run_op(ITensorPack &tensors, const Window &window, ::cl::CommandQueue &queue) override;
+
+private:
+    bool _run_in_place;
+};
+} // namespace kernels
+} // namespace opencl
+} // namespace arm_compute
+#endif /* ARM_COMPUTE_CL_ACTIVATION_KERNEL_H */
diff --git a/src/core/CL/kernels/CLFloorKernel.cpp b/src/core/gpu/cl/kernels/ClFloorKernel.cpp
index 6af1822656..7296d40eaf 100644
--- a/src/core/CL/kernels/CLFloorKernel.cpp
+++ b/src/core/gpu/cl/kernels/ClFloorKernel.cpp
@@ -21,7 +21,7 @@
  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
  */
-#include "src/core/CL/kernels/CLFloorKernel.h"
+#include "src/core/gpu/cl/kernels/ClFloorKernel.h"
 
 #include "arm_compute/core/CL/CLHelpers.h"
 #include "arm_compute/core/CL/CLKernelLibrary.h"
@@ -33,52 +33,49 @@
 #include "src/core/CL/CLValidate.h"
 #include "src/core/helpers/AutoConfiguration.h"
 #include "src/core/helpers/WindowHelpers.h"
+#include "support/Cast.h"
 #include "support/StringSupport.h"
 
 namespace arm_compute
 {
+namespace opencl
+{
+namespace kernels
+{
 namespace
 {
-Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output)
+Status validate_arguments(const ITensorInfo *src, const ITensorInfo *dst)
 {
-    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output);
-    ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(input);
-    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::F16, DataType::F32);
+    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(src, dst);
+    ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(src);
+    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(src, 1, DataType::F16, DataType::F32);
 
     // Validate in case of configured output
-    if(output->total_size() > 0)
+    if(dst->total_size() > 0)
     {
-        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
-        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(input, output);
+        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(src, dst);
+        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(src, dst);
     }
 
     return Status{};
 }
 } // namespace
 
-CLFloorKernel::CLFloorKernel()
-    : _input(nullptr), _output(nullptr)
-{
-}
-
-void CLFloorKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output)
+void ClFloorKernel::configure(const ClCompileContext &compile_context, const ITensorInfo *src, ITensorInfo *dst)
 {
-    ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
+    ARM_COMPUTE_ERROR_ON_NULLPTR(src, dst);
 
     // Auto initialize output
-    auto_init_if_empty(*output->info(), input->info()->tensor_shape(), 1, input->info()->data_type());
+    auto_init_if_empty(*dst, src->tensor_shape(), 1, src->data_type());
 
     // Validate
-    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(), output->info()));
-    auto padding_info = get_padding_info({ input, output });
+    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(src, dst));
+    auto padding_info = get_padding_info({ src, dst });
 
-    _input  = input;
-    _output = output;
-
-    const unsigned int vec_size_x           = adjust_vec_size(max_cl_vector_width / input->info()->element_size(), input->info()->dimension(0));
-    const int          vec_size_x_leftovers = input->info()->dimension(0) % vec_size_x;
+    const unsigned int vec_size_x           = adjust_vec_size(max_cl_vector_width / src->element_size(), src->dimension(0));
+    const int          vec_size_x_leftovers = src->dimension(0) % vec_size_x;
     CLBuildOptions     build_opts;
-    build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(input->info()->data_type()));
+    build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(src->data_type()));
     build_opts.add_option("-DVEC_SIZE=" + support::cpp11::to_string(vec_size_x));
     build_opts.add_option("-DVEC_SIZE_LEFTOVER=" + support::cpp11::to_string(vec_size_x_leftovers));
 
@@ -86,37 +83,37 @@ void CLFloorKernel::configure(const CLCompileContext &compile_context, const ICL
     _kernel = create_kernel(compile_context, "floor_layer", build_opts.options());
 
     // Configure kernel window
-    Window win = calculate_max_window(*input->info(), Steps(vec_size_x));
-    ICLKernel::configure_internal(win);
+    Window win = calculate_max_window(*src, Steps(vec_size_x));
+    IClKernel::configure_internal(win);
     ARM_COMPUTE_ERROR_ON(has_padding_changed(padding_info));
 }
 
-void CLFloorKernel::configure(const ICLTensor *input, ICLTensor *output)
+Status ClFloorKernel::validate(const ITensorInfo *src, const ITensorInfo *dst)
 {
-    configure(CLKernelLibrary::get().get_compile_context(), input, output);
-}
-
-Status CLFloorKernel::validate(const ITensorInfo *input, const ITensorInfo *output)
-{
-    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, output));
+    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(src, dst));
     return Status{};
 }
 
-void CLFloorKernel::run(const Window &window, cl::CommandQueue &queue)
+void ClFloorKernel::run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue)
 {
     ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
+    ARM_COMPUTE_ERROR_ON_MISMATCHING_WINDOWS(IClKernel::window(), window);
+
+    const auto src = utils::cast::polymorphic_downcast<const ICLTensor *>(tensors.get_const_tensor(TensorType::ACL_SRC));
+    auto       dst = utils::cast::polymorphic_downcast<ICLTensor *>(tensors.get_tensor(TensorType::ACL_DST));
 
-    Window collapsed = window.collapse_if_possible(ICLKernel::window(), Window::DimZ);
+    Window collapsed = window.collapse_if_possible(IClKernel::window(), Window::DimZ);
     Window slice     = collapsed.first_slice_window_3D();
 
     do
     {
         unsigned int idx = 0;
-        add_3D_tensor_argument(idx, _input, slice);
-        add_3D_tensor_argument(idx, _output, slice);
+        add_3D_tensor_argument(idx, src, slice);
+        add_3D_tensor_argument(idx, dst, slice);
         enqueue(queue, *this, slice, lws_hint());
     }
     while(collapsed.slide_window_slice_3D(slice));
 }
+} // namespace kernels
+} // namespace opencl
 } // namespace arm_compute
diff --git a/src/core/gpu/cl/kernels/ClFloorKernel.h b/src/core/gpu/cl/kernels/ClFloorKernel.h
new file mode 100644
index 0000000000..09ab801f2d
--- /dev/null
+++ b/src/core/gpu/cl/kernels/ClFloorKernel.h
@@ -0,0 +1,66 @@
+/*
+ * Copyright (c) 2017-2021 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.
+ */
+#ifndef ARM_COMPUTE_CL_FLOOR_KERNEL_H
+#define ARM_COMPUTE_CL_FLOOR_KERNEL_H
+
+#include "src/core/common/Macros.h"
+#include "src/core/gpu/cl/ClCompileContext.h"
+#include "src/core/gpu/cl/IClKernel.h"
+
+namespace arm_compute
+{
+namespace opencl
+{
+namespace kernels
+{
+/** OpenCL kernel to perform a floor operation */
+class ClFloorKernel : public IClKernel
+{
+public:
+    ClFloorKernel() = default;
+    ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE(ClFloorKernel);
+    /** Configure kernel for a given list of arguments
+     *
+     * @param[in]  compile_context The compile context to be used.
+     * @param[in]  src             Source tensor. Data type supported: F16/F32.
+     * @param[out] dst             Destination tensor. Same as @p src
+     */
+    void configure(const ClCompileContext &compile_context, const ITensorInfo *src, ITensorInfo *dst);
+
+    /** Static function to check if given info will lead to a valid configuration of @ref ClFloorKernel
+     *
+     * @param[in] src Source tensor info. Data type supported: F16/F32.
+     * @param[in] dst Destination tensor info. Same as @p src
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *src, const ITensorInfo *dst);
+
+    // Inherited methods overridden:
+    void run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue) override;
+};
+} // namespace kernels
+} // namespace opencl
+} // namespace arm_compute
+#endif /* ARM_COMPUTE_CL_FLOOR_KERNEL_H */
diff --git a/src/runtime/CL/functions/CLActivationLayer.cpp b/src/runtime/CL/functions/CLActivationLayer.cpp
index 0070e43f8c..9c71b2aa7d 100644
--- a/src/runtime/CL/functions/CLActivationLayer.cpp
+++ b/src/runtime/CL/functions/CLActivationLayer.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-2020 Arm Limited.
+ * Copyright (c) 2016-2021 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -25,32 +25,19 @@
 
 #include "arm_compute/core/CL/ICLTensor.h"
 #include "arm_compute/core/Types.h"
+#include "arm_compute/core/Validate.h"
 #include "arm_compute/runtime/CL/CLRuntimeContext.h"
-#include "src/core/CL/kernels/CLActivationLayerKernel.h"
+#include "src/core/CL/ICLKernel.h"
+#include "src/runtime/gpu/cl/operators/ClActivation.h"
 
 namespace arm_compute
 {
-namespace experimental
-{
-void CLActivation::configure(const CLCompileContext &compile_context, ITensorInfo *input, ITensorInfo *output, ActivationLayerInfo act_info)
-{
-    auto k = std::make_unique<CLActivationLayerKernel>();
-    k->configure(compile_context, input, output, act_info);
-    _kernel = std::move(k);
-}
-
-Status CLActivation::validate(const ITensorInfo *input, const ITensorInfo *output, const ActivationLayerInfo &act_info)
-{
-    return CLActivationLayerKernel::validate(input, output, act_info);
-}
-} // namespace experimental
-
 struct CLActivationLayer::Impl
 {
-    const ICLTensor                            *src{ nullptr };
-    ICLTensor                                  *dst{ nullptr };
-    CLRuntimeContext                           *ctx{ nullptr };
-    std::unique_ptr<experimental::CLActivation> op{ nullptr };
+    const ICLTensor                      *src{ nullptr };
+    ICLTensor                            *dst{ nullptr };
+    CLRuntimeContext                     *ctx{ nullptr };
+    std::unique_ptr<opencl::ClActivation> op{ nullptr };
 };
 
 CLActivationLayer::CLActivationLayer(CLRuntimeContext *ctx)
@@ -58,12 +45,9 @@ CLActivationLayer::CLActivationLayer(CLRuntimeContext *ctx)
 {
     _impl->ctx = ctx;
 }
-
 CLActivationLayer::CLActivationLayer(CLActivationLayer &&) = default;
-
 CLActivationLayer &CLActivationLayer::operator=(CLActivationLayer &&) = default;
-
-CLActivationLayer::~CLActivationLayer() = default;
+CLActivationLayer::~CLActivationLayer()                               = default;
 
 void CLActivationLayer::configure(ICLTensor *input, ICLTensor *output, ActivationLayerInfo act_info)
 {
@@ -77,13 +61,13 @@ void CLActivationLayer::configure(const CLCompileContext &compile_context, ICLTe
     _impl->src = input;
     _impl->dst = output == nullptr ? input : output;
 
-    _impl->op = std::make_unique<experimental::CLActivation>();
+    _impl->op = std::make_unique<opencl::ClActivation>();
     _impl->op->configure(compile_context, _impl->src->info(), _impl->dst->info(), act_info);
 }
 
 Status CLActivationLayer::validate(const ITensorInfo *input, const ITensorInfo *output, const ActivationLayerInfo &act_info)
 {
-    return experimental::CLActivation::validate(input, output, act_info);
+    return opencl::ClActivation::validate(input, output, act_info);
 }
 
 void CLActivationLayer::run()
diff --git a/src/runtime/CL/functions/CLFloor.cpp b/src/runtime/CL/functions/CLFloor.cpp
index 5549d09b24..4c5e482b10 100644
--- a/src/runtime/CL/functions/CLFloor.cpp
+++ b/src/runtime/CL/functions/CLFloor.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017-2020 Arm Limited.
+ * Copyright (c) 2016-2021 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -23,10 +23,30 @@
  */
 #include "arm_compute/runtime/CL/functions/CLFloor.h"
 
-#include "src/core/CL/kernels/CLFloorKernel.h"
+#include "arm_compute/core/CL/CLKernelLibrary.h"
+#include "arm_compute/core/CL/ICLTensor.h"
+#include "arm_compute/core/Types.h"
+#include "arm_compute/core/Validate.h"
+#include "src/core/CL/ICLKernel.h"
+#include "src/runtime/gpu/cl/operators/ClFloor.h"
 
 namespace arm_compute
 {
+struct CLFloor::Impl
+{
+    const ICLTensor                 *src{ nullptr };
+    ICLTensor                       *dst{ nullptr };
+    std::unique_ptr<opencl::ClFloor> op{ nullptr };
+};
+
+CLFloor::CLFloor()
+    : _impl(std::make_unique<Impl>())
+{
+}
+CLFloor::CLFloor(CLFloor &&) = default;
+CLFloor &CLFloor::operator=(CLFloor &&) = default;
+CLFloor::~CLFloor()                     = default;
+
 void CLFloor::configure(const ICLTensor *input, ICLTensor *output)
 {
     configure(CLKernelLibrary::get().get_compile_context(), input, output);
@@ -34,13 +54,25 @@ void CLFloor::configure(const ICLTensor *input, ICLTensor *output)
 
 void CLFloor::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output)
 {
-    auto k = std::make_unique<CLFloorKernel>();
-    k->configure(compile_context, input, output);
-    _kernel = std::move(k);
+    ARM_COMPUTE_ERROR_ON_NULLPTR(input);
+
+    _impl->src = input;
+    _impl->dst = output;
+
+    _impl->op = std::make_unique<opencl::ClFloor>();
+    _impl->op->configure(compile_context, _impl->src->info(), _impl->dst->info());
 }
 
 Status CLFloor::validate(const ITensorInfo *input, const ITensorInfo *output)
 {
-    return CLFloorKernel::validate(input, output);
+    return opencl::ClFloor::validate(input, output);
+}
+
+void CLFloor::run()
+{
+    ITensorPack pack;
+    pack.add_tensor(TensorType::ACL_SRC, _impl->src);
+    pack.add_tensor(TensorType::ACL_DST, _impl->dst);
+    _impl->op->run(pack);
 }
 } // namespace arm_compute
diff --git a/src/runtime/gpu/cl/operators/ClActivation.cpp b/src/runtime/gpu/cl/operators/ClActivation.cpp
new file mode 100644
index 0000000000..71aa57bdbd
--- /dev/null
+++ b/src/runtime/gpu/cl/operators/ClActivation.cpp
@@ -0,0 +1,45 @@
+/*
+ * Copyright (c) 2016-2021 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.
+ */
+#include "src/runtime/gpu/cl/operators/ClActivation.h"
+
+#include "src/core/gpu/cl/ClCompileContext.h"
+#include "src/core/gpu/cl/kernels/ClActivationKernel.h"
+
+namespace arm_compute
+{
+namespace opencl
+{
+void ClActivation::configure(const ClCompileContext &compile_context, ITensorInfo *src, ITensorInfo *dst, const ActivationLayerInfo &act_info)
+{
+    auto k = std::make_unique<kernels::ClActivationKernel>();
+    k->configure(compile_context, src, dst, act_info);
+    _kernel = std::move(k);
+}
+
+Status ClActivation::validate(const ITensorInfo *src, const ITensorInfo *dst, const ActivationLayerInfo &act_info)
+{
+    return kernels::ClActivationKernel::validate(src, dst, act_info);
+}
+} // namespace opencl
+} // namespace arm_compute
diff --git a/src/runtime/gpu/cl/operators/ClActivation.h b/src/runtime/gpu/cl/operators/ClActivation.h
new file mode 100644
index 0000000000..235b826b87
--- /dev/null
+++ b/src/runtime/gpu/cl/operators/ClActivation.h
@@ -0,0 +1,60 @@
+/*
+ * Copyright (c) 2021 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.
+ */
+#ifndef ARM_COMPUTE_CL_ACTIVATION_H
+#define ARM_COMPUTE_CL_ACTIVATION_H
+
+#include "src/core/gpu/cl/ClCompileContext.h"
+#include "src/runtime/gpu/cl/IClOperator.h"
+
+namespace arm_compute
+{
+namespace opencl
+{
+/** Basic function to run @ref kernels::ClActivationKernel */
+class ClActivation : public IClOperator
+{
+public:
+    /** Constructor */
+    ClActivation() = default;
+    /** Configure operator for a given list of arguments
+     *
+     * @param[in]  compile_context The compile context to be used.
+     * @param[in]  src             Source tensor info. Data types supported: QASYMM8/QASYMM8_SIGNED/QSYMM16/F16/F32.
+     * @param[out] dst             Destination tensor info. Data type supported: same as @p src
+     * @param[in]  activation_info Activation layer parameters.
+     */
+    void configure(const ClCompileContext &compile_context, ITensorInfo *src, ITensorInfo *dst, const ActivationLayerInfo &activation_info);
+    /** Static function to check if given info will lead to a valid configuration of @ref ClActivation
+     *
+     * @param[in] src      Source tensor info. Data types supported: QASYMM8/QASYMM8_SIGNED/QSYMM16/F16/F32.
+     * @param[in] dst      Destination tensor info. Data type supported: same as @p src
+     * @param[in] act_info Activation layer information.
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *src, const ITensorInfo *dst, const ActivationLayerInfo &act_info);
+};
+} // namespace opencl
+} // namespace arm_compute
+#endif /* ARM_COMPUTE_CL_ACTIVATION_H */
diff --git a/src/runtime/gpu/cl/operators/ClFloor.cpp b/src/runtime/gpu/cl/operators/ClFloor.cpp
new file mode 100644
index 0000000000..94e77c0c54
--- /dev/null
+++ b/src/runtime/gpu/cl/operators/ClFloor.cpp
@@ -0,0 +1,45 @@
+/*
+ * Copyright (c) 2017-2021 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.
+ */
+#include "src/runtime/gpu/cl/operators/ClFloor.h"
+
+#include "src/core/gpu/cl/ClCompileContext.h"
+#include "src/core/gpu/cl/kernels/ClFloorKernel.h"
+
+namespace arm_compute
+{
+namespace opencl
+{
+void ClFloor::configure(const ClCompileContext &compile_context, const ITensorInfo *src, ITensorInfo *dst)
+{
+    auto k = std::make_unique<kernels::ClFloorKernel>();
+    k->configure(compile_context, src, dst);
+    _kernel = std::move(k);
+}
+
+Status ClFloor::validate(const ITensorInfo *src, const ITensorInfo *dst)
+{
+    return kernels::ClFloorKernel::validate(src, dst);
+}
+} // namespace opencl
+} // namespace arm_compute
diff --git a/src/runtime/gpu/cl/operators/ClFloor.h b/src/runtime/gpu/cl/operators/ClFloor.h
new file mode 100644
index 0000000000..f54eef9140
--- /dev/null
+++ b/src/runtime/gpu/cl/operators/ClFloor.h
@@ -0,0 +1,58 @@
+/*
+ * Copyright (c) 2021 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.
+ */
+#ifndef ARM_COMPUTE_CL_FLOOR_H
+#define ARM_COMPUTE_CL_FLOOR_H
+
+#include "src/core/gpu/cl/ClCompileContext.h"
+#include "src/runtime/gpu/cl/IClOperator.h"
+
+namespace arm_compute
+{
+namespace opencl
+{
+/** Basic function to run @ref kernels::ClFloorKernel */
+class ClFloor : public IClOperator
+{
+public:
+    /** Constructor */
+    ClFloor() = default;
+    /** Configure operator for a given list of arguments
+     *
+     * @param[in] compile_context The compile context to be used.
+     * @param[in] src             Source tensor info. Data types supported: F16/F32.
+     * @param[in] dst             Destination tensor info. Data type supported: same as @p src
+     */
+    void configure(const ClCompileContext &compile_context, const ITensorInfo *src, ITensorInfo *dst);
+    /** Static function to check if given info will lead to a valid configuration of @ref ClFloor
+     *
+     * @param[in] src Source tensor info. Data types supported: F16/F32.
+     * @param[in] dst Destination tensor info. Data type supported: same as @p src
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *src, const ITensorInfo *dst);
+};
+} // namespace opencl
+} // namespace arm_compute
+#endif /* ARM_COMPUTE_CL_FLOOR_H */