COMPMID-3392: Collapse TensorMaps into a single TensorPack

Collapse InputTensorMap and OutputTensorMap to a single TensorPack
mechanism.

Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com>
Change-Id: Ie2fdfc6b07d84ad589169ec99ca64fcf45a00bec
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/c/VisualCompute/ComputeLibrary/+/253783
Tested-by: bsgcomp <bsgcomp@arm.com>
Reviewed-by: Michalis Spyrou <michalis.spyrou@arm.com>
Signed-off-by: Michalis Spyrou <michalis.spyrou@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/3641
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: SiCong Li <sicong.li@arm.com>
Reviewed-by: Sheri Zhang <sheri.zhang@arm.com>

45 files changed, 173 insertions, 103 deletions

diff --git a/arm_compute/core/CL/ICLKernel.h b/arm_compute/core/CL/ICLKernel.h
index 6e606ff857..d4990a1dee 100644
--- a/arm_compute/core/CL/ICLKernel.h
+++ b/arm_compute/core/CL/ICLKernel.h
@@ -225,14 +225,13 @@ public:
      *
      * @note The queue is *not* flushed by this method, and therefore the kernel will not have been executed by the time this method returns.
      *
-     * @param[in]     inputs  A vector containing the input tensors.
-     * @param[in]     outputs A vector containing the output tensors.
+     * @param[in]     tensors A vector containing the tensors to operato on.
      * @param[in]     window  Region on which to execute the kernel. (Must be a valid region of the window returned by window()).
      * @param[in,out] queue   Command queue on which to enqueue the kernel.
      */
-    virtual void run_op(const InputTensorMap &inputs, const OutputTensorMap &outputs, const Window &window, cl::CommandQueue &queue)
+    virtual void run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue)
     {
-        ARM_COMPUTE_UNUSED(inputs, outputs, window, queue);
+        ARM_COMPUTE_UNUSED(tensors, window, queue);
     }
     /** Add the passed parameters to the object's kernel's arguments starting from the index idx.
      *
diff --git a/arm_compute/core/CL/kernels/CLActivationLayerKernel.h b/arm_compute/core/CL/kernels/CLActivationLayerKernel.h
index cf62ca600b..81d4ccb065 100644
--- a/arm_compute/core/CL/kernels/CLActivationLayerKernel.h
+++ b/arm_compute/core/CL/kernels/CLActivationLayerKernel.h
@@ -68,8 +68,7 @@ public:
     static Status validate(const ITensorInfo *input, const ITensorInfo *output, const ActivationLayerInfo &act_info);
 
     // Inherited methods overridden:
-    void run_op(const InputTensorMap &inputs, const OutputTensorMap &outputs,
-                const Window &window, cl::CommandQueue &queue) override;
+    void run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue) override;
 
 private:
     bool _run_in_place;
diff --git a/arm_compute/core/CL/kernels/CLBatchConcatenateLayerKernel.h b/arm_compute/core/CL/kernels/CLBatchConcatenateLayerKernel.h
index 890fecf0c9..bb8968ca83 100644
--- a/arm_compute/core/CL/kernels/CLBatchConcatenateLayerKernel.h
+++ b/arm_compute/core/CL/kernels/CLBatchConcatenateLayerKernel.h
@@ -73,8 +73,7 @@ public:
     static Status validate(const ITensorInfo *input, unsigned int batch_offset, const ITensorInfo *output);
 
     // Inherited methods overridden:
-    void run_op(const InputTensorMap &inputs, const OutputTensorMap &outputs,
-                const Window &window, cl::CommandQueue &queue) override;
+    void run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue) override;
 
 private:
     unsigned int _batch_offset;
diff --git a/arm_compute/core/CL/kernels/CLDepthConcatenateLayerKernel.h b/arm_compute/core/CL/kernels/CLDepthConcatenateLayerKernel.h
index ad365d655b..d8493bc5d8 100644
--- a/arm_compute/core/CL/kernels/CLDepthConcatenateLayerKernel.h
+++ b/arm_compute/core/CL/kernels/CLDepthConcatenateLayerKernel.h
@@ -71,8 +71,7 @@ public:
     static Status validate(const ITensorInfo *input, unsigned int depth_offset, const ITensorInfo *output);
 
     // Inherited methods overridden:
-    void run_op(const InputTensorMap &inputs, const OutputTensorMap &outputs,
-                const Window &window, cl::CommandQueue &queue) override;
+    void run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue) override;
 
 private:
     unsigned int _depth_offset;
diff --git a/arm_compute/core/CL/kernels/CLElementWiseUnaryLayerKernel.h b/arm_compute/core/CL/kernels/CLElementWiseUnaryLayerKernel.h
index f8f158c5bc..82cd953b68 100644
--- a/arm_compute/core/CL/kernels/CLElementWiseUnaryLayerKernel.h
+++ b/arm_compute/core/CL/kernels/CLElementWiseUnaryLayerKernel.h
@@ -60,8 +60,7 @@ public:
     static Status validate(const ITensorInfo *input, const ITensorInfo *output, const ElementWiseUnary &op);
 
     // Inherited methods overridden:
-    void run_op(const InputTensorMap &inputs, const OutputTensorMap &outputs,
-                const Window &window, cl::CommandQueue &queue) override;
+    void run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue) override;
 };
 } // namespace arm_compute
 #endif /* ARM_COMPUTE_CLELEMENTWISEUNARYLAYERKERNEL_H */
diff --git a/arm_compute/core/CL/kernels/CLElementwiseOperationKernel.h b/arm_compute/core/CL/kernels/CLElementwiseOperationKernel.h
index 76bc879638..b459292161 100644
--- a/arm_compute/core/CL/kernels/CLElementwiseOperationKernel.h
+++ b/arm_compute/core/CL/kernels/CLElementwiseOperationKernel.h
@@ -54,8 +54,7 @@ public:
     ~CLElementwiseOperationKernel() = default;
 
     // Inherited methods overridden:
-    void run_op(const InputTensorMap &inputs, const OutputTensorMap &outputs, const Window &window, cl::CommandQueue &queue) override;
-
+    void run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue) override;
     BorderSize border_size() const override;
 
 protected:
diff --git a/arm_compute/core/CL/kernels/CLFillBorderKernel.h b/arm_compute/core/CL/kernels/CLFillBorderKernel.h
index 8cad68dc1a..5323af4c0e 100644
--- a/arm_compute/core/CL/kernels/CLFillBorderKernel.h
+++ b/arm_compute/core/CL/kernels/CLFillBorderKernel.h
@@ -85,7 +85,7 @@ public:
     void set_constant_border(unsigned int idx, const PixelValue &constant_border_value);
 
     // Inherited methods overridden:
-    void run_op(const InputTensorMap &inputs, const OutputTensorMap &outputs, const Window &window, cl::CommandQueue &queue) override;
+    void run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue) override;
     void run(const Window &window, cl::CommandQueue &queue) override;
     bool is_parallelisable() const override;
 
diff --git a/arm_compute/core/CL/kernels/CLHeightConcatenateLayerKernel.h b/arm_compute/core/CL/kernels/CLHeightConcatenateLayerKernel.h
index 0563fad414..4fa2b40881 100644
--- a/arm_compute/core/CL/kernels/CLHeightConcatenateLayerKernel.h
+++ b/arm_compute/core/CL/kernels/CLHeightConcatenateLayerKernel.h
@@ -68,8 +68,7 @@ public:
     static Status validate(const ITensorInfo *input, unsigned int height_offset, const ITensorInfo *output);
 
     // Inherited methods overridden:
-    void run_op(const InputTensorMap &inputs, const OutputTensorMap &outputs,
-                const Window &window, cl::CommandQueue &queue) override;
+    void run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue) override;
 
 private:
     unsigned int _height_offset;
diff --git a/arm_compute/core/CL/kernels/CLPixelWiseMultiplicationKernel.h b/arm_compute/core/CL/kernels/CLPixelWiseMultiplicationKernel.h
index 86159fc915..6b5bd11bde 100644
--- a/arm_compute/core/CL/kernels/CLPixelWiseMultiplicationKernel.h
+++ b/arm_compute/core/CL/kernels/CLPixelWiseMultiplicationKernel.h
@@ -132,7 +132,7 @@ public:
                            ConvertPolicy overflow_policy, RoundingPolicy rounding_policy, const ActivationLayerInfo &act_info = ActivationLayerInfo());
 
     // Inherited methods overridden:
-    void run_op(const InputTensorMap &inputs, const OutputTensorMap &outputs, const Window &window, cl::CommandQueue &queue) override;
+    void run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue) override;
     BorderSize border_size() const override;
 
 private:
@@ -184,7 +184,7 @@ public:
     static Status validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output, const ActivationLayerInfo &act_info = ActivationLayerInfo());
 
     // Inherited methods overridden:
-    void run_op(const InputTensorMap &inputs, const OutputTensorMap &outputs, const Window &window, cl::CommandQueue &queue) override;
+    void run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue) override;
     BorderSize border_size() const override;
 
 private:
diff --git a/arm_compute/core/CL/kernels/CLReshapeLayerKernel.h b/arm_compute/core/CL/kernels/CLReshapeLayerKernel.h
index 11c06d1cab..6e3f255c52 100644
--- a/arm_compute/core/CL/kernels/CLReshapeLayerKernel.h
+++ b/arm_compute/core/CL/kernels/CLReshapeLayerKernel.h
@@ -53,8 +53,7 @@ public:
     static Status validate(const ITensorInfo *input, const ITensorInfo *output);
 
     // Inherited methods overridden:
-    void run_op(const InputTensorMap &inputs, const OutputTensorMap &outputs,
-                const Window &window, cl::CommandQueue &queue) override;
+    void run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue) override;
 };
 } // namespace arm_compute
 #endif /*ARM_COMPUTE_CLRESHAPELAYERKERNEL_H */
diff --git a/arm_compute/core/CL/kernels/CLStridedSliceKernel.h b/arm_compute/core/CL/kernels/CLStridedSliceKernel.h
index 28a665b113..74311b71fa 100644
--- a/arm_compute/core/CL/kernels/CLStridedSliceKernel.h
+++ b/arm_compute/core/CL/kernels/CLStridedSliceKernel.h
@@ -73,8 +73,7 @@ public:
                            int32_t begin_mask, int32_t end_mask, int32_t shrink_axis_mask);
 
     // Inherited methods overridden:
-    void run_op(const InputTensorMap &inputs, const OutputTensorMap &outputs,
-                const Window &window, cl::CommandQueue &queue) override;
+    void run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue) override;
 };
 } // namespace arm_compute
 #endif /*ARM_COMPUTE_CL_STRIDED_SLICE_KERNEL_H */
diff --git a/arm_compute/core/CL/kernels/CLWidthConcatenate2TensorsKernel.h b/arm_compute/core/CL/kernels/CLWidthConcatenate2TensorsKernel.h
index aaedaa90ee..a379b5f0b8 100644
--- a/arm_compute/core/CL/kernels/CLWidthConcatenate2TensorsKernel.h
+++ b/arm_compute/core/CL/kernels/CLWidthConcatenate2TensorsKernel.h
@@ -67,8 +67,7 @@ public:
     static Status validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output);
 
     // Inherited methods overridden:
-    void run_op(const InputTensorMap &inputs, const OutputTensorMap &outputs,
-                const Window &window, cl::CommandQueue &queue) override;
+    void run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue) override;
 };
 } // namespace arm_compute
 #endif /* ARM_COMPUTE_CLWIDTHCONCATENATE_2TENSORS_KERNEL_H */
diff --git a/arm_compute/core/CL/kernels/CLWidthConcatenate4TensorsKernel.h b/arm_compute/core/CL/kernels/CLWidthConcatenate4TensorsKernel.h
index b1d6e890a0..6b0e8ee21d 100644
--- a/arm_compute/core/CL/kernels/CLWidthConcatenate4TensorsKernel.h
+++ b/arm_compute/core/CL/kernels/CLWidthConcatenate4TensorsKernel.h
@@ -71,8 +71,7 @@ public:
     static Status validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *input3, const ITensorInfo *input4, const ITensorInfo *output);
 
     // Inherited methods overridden:
-    void run_op(const InputTensorMap &inputs, const OutputTensorMap &outputs,
-                const Window &window, cl::CommandQueue &queue) override;
+    void run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue) override;
 };
 } // namespace arm_compute
 #endif /* ARM_COMPUTE_CLWIDTHCONCATENATE_4TENSORS_KERNEL_H */
diff --git a/arm_compute/core/CL/kernels/CLWidthConcatenateLayerKernel.h b/arm_compute/core/CL/kernels/CLWidthConcatenateLayerKernel.h
index cb8777d5fd..32e90af404 100644
--- a/arm_compute/core/CL/kernels/CLWidthConcatenateLayerKernel.h
+++ b/arm_compute/core/CL/kernels/CLWidthConcatenateLayerKernel.h
@@ -68,8 +68,7 @@ public:
     static Status validate(const ITensorInfo *input, unsigned int width_offset, const ITensorInfo *output);
 
     // Inherited methods overridden:
-    void run_op(const InputTensorMap &inputs, const OutputTensorMap &outputs,
-                const Window &window, cl::CommandQueue &queue) override;
+    void run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue) override;
 
 private:
     unsigned int _width_offset;
diff --git a/arm_compute/core/CPP/ICPPKernel.h b/arm_compute/core/CPP/ICPPKernel.h
index 3f784cf6d9..ab369ffe1d 100644
--- a/arm_compute/core/CPP/ICPPKernel.h
+++ b/arm_compute/core/CPP/ICPPKernel.h
@@ -79,14 +79,13 @@ public:
      *
      * @note The width of the window has to be a multiple of num_elems_processed_per_iteration().
      *
-     * @param[in] inputs  A vector containing the input tensors.
-     * @param[in] outputs A vector containing the output tensors.
+     * @param[in] tensors A vector containing the tensors to operate on.
      * @param[in] window  Region on which to execute the kernel. (Must be a region of the window returned by window())
      * @param[in] info    Info about executing thread and CPU.
      */
-    virtual void run_op(const InputTensorMap &inputs, const OutputTensorMap &outputs, const Window &window, const ThreadInfo &info)
+    virtual void run_op(ITensorPack &tensors, const Window &window, const ThreadInfo &info)
     {
-        ARM_COMPUTE_UNUSED(inputs, outputs, window, info);
+        ARM_COMPUTE_UNUSED(tensors, window, info);
     }
 
     /** Name of the kernel
diff --git a/arm_compute/core/ITensorPack.h b/arm_compute/core/ITensorPack.h
new file mode 100644
index 0000000000..36b6aea490
--- /dev/null
+++ b/arm_compute/core/ITensorPack.h
@@ -0,0 +1,100 @@
+/*
+ * Copyright (c) 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_ITENSORPACK_H
+#define ARM_COMPUTE_ITENSORPACK_H
+
+#include <cstdint>
+#include <map>
+
+namespace arm_compute
+{
+// Forward declaration
+class ITensor;
+
+/** Tensor packing service */
+class ITensorPack
+{
+private:
+    struct PackElement
+    {
+        PackElement() = default;
+        PackElement(ITensor *tensor)
+            : tensor(tensor), ctensor(nullptr)
+        {
+        }
+        PackElement(const ITensor *ctensor)
+            : tensor(nullptr), ctensor(ctensor)
+        {
+        }
+
+        ITensor       *tensor{ nullptr };
+        const ITensor *ctensor{ nullptr };
+    };
+
+public:
+    /** Default Constructor */
+    ITensorPack() = default;
+    /** Add tensor to the pack
+     *
+     * @param[in] id     ID/type of the tensor to add
+     * @param[in] tensor Tensor to add
+     */
+    void add_tensor(int id, ITensor *tensor);
+
+    /** Add const tensor to the pack
+     *
+     * @param[in] id     ID/type of the tensor to add
+     * @param[in] tensor Tensor to add
+     */
+    void add_tensor(int id, const ITensor *tensor);
+    /** Get tensor of a given id from the pac
+     *
+     * @param[in] id ID of tensor to extract
+     *
+     * @return The pointer to the tensor if exist and is non-const else nullptr
+     */
+    ITensor *get_tensor(int id);
+    /** Get constant tensor of a given id
+     *
+     * @param[in] id ID of tensor to extract
+     *
+     * @return The pointer to the tensor if exist and is const else nullptr
+     */
+    const ITensor *get_const_tensor(int id) const;
+    /** Pack size accessor
+     *
+     * @return Number of tensors registered to the pack
+     */
+    size_t size() const;
+    /** Checks if pack is empty
+     *
+     * @return True if empty else false
+     */
+    bool empty() const;
+
+private:
+    std::map<unsigned int, PackElement> _pack{}; /**< Container with the packed tensors */
+};
+} // namespace arm_compute
+#endif /*ARM_COMPUTE_ITENSORPACK_H */
diff --git a/arm_compute/core/NEON/kernels/NEActivationLayerKernel.h b/arm_compute/core/NEON/kernels/NEActivationLayerKernel.h
index 960e6fa3e5..325647bd66 100644
--- a/arm_compute/core/NEON/kernels/NEActivationLayerKernel.h
+++ b/arm_compute/core/NEON/kernels/NEActivationLayerKernel.h
@@ -76,8 +76,7 @@ public:
     static Status validate(const ITensorInfo *input, const ITensorInfo *output, const ActivationLayerInfo &act_info);
 
     // Inherited methods overridden:
-    void run_op(const InputTensorMap &inputs, const OutputTensorMap &outputs,
-                const Window &window, const ThreadInfo &info) override;
+    void run_op(ITensorPack &tensors, const Window &window, const ThreadInfo &info) override;
 
 private:
     using ActivationFunction = ActivationLayerInfo::ActivationFunction;
diff --git a/arm_compute/core/NEON/kernels/NEArithmeticAdditionKernel.h b/arm_compute/core/NEON/kernels/NEArithmeticAdditionKernel.h
index 764596d7dc..eece5708e8 100644
--- a/arm_compute/core/NEON/kernels/NEArithmeticAdditionKernel.h
+++ b/arm_compute/core/NEON/kernels/NEArithmeticAdditionKernel.h
@@ -86,7 +86,7 @@ public:
     static Status validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output, ConvertPolicy policy);
 
     // Inherited methods overridden:
-    void run_op(const InputTensorMap &inputs, const OutputTensorMap &outputs, const Window &window, const ThreadInfo &info) override;
+    void run_op(ITensorPack &tensors, const Window &window, const ThreadInfo &info) override;
 
 private:
     /** Common signature for all the specialised add functions
diff --git a/arm_compute/core/NEON/kernels/NEArithmeticSubtractionKernel.h b/arm_compute/core/NEON/kernels/NEArithmeticSubtractionKernel.h
index 86c3dd8e88..e3a41a2b1c 100644
--- a/arm_compute/core/NEON/kernels/NEArithmeticSubtractionKernel.h
+++ b/arm_compute/core/NEON/kernels/NEArithmeticSubtractionKernel.h
@@ -98,7 +98,7 @@ public:
     static Status validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output, ConvertPolicy policy);
 
     // Inherited methods overridden:
-    void run_op(const InputTensorMap &inputs, const OutputTensorMap &outputs, const Window &window, const ThreadInfo &info) override;
+    void run_op(ITensorPack &tensors, const Window &window, const ThreadInfo &info) override;
 
 private:
     /** Common signature for all the specialised sub functions
diff --git a/arm_compute/core/NEON/kernels/NEBatchConcatenateLayerKernel.h b/arm_compute/core/NEON/kernels/NEBatchConcatenateLayerKernel.h
index f397a29b48..478890925b 100644
--- a/arm_compute/core/NEON/kernels/NEBatchConcatenateLayerKernel.h
+++ b/arm_compute/core/NEON/kernels/NEBatchConcatenateLayerKernel.h
@@ -76,8 +76,7 @@ public:
     static Status validate(const ITensorInfo *input, unsigned int batch_offset, const ITensorInfo *output);
 
     // Inherited methods overridden:
-    void run_op(const InputTensorMap &inputs, const OutputTensorMap &outputs,
-                const Window &window, const ThreadInfo &info) override;
+    void run_op(ITensorPack &tensors, const Window &window, const ThreadInfo &info) override;
 
 private:
     using BatchConcatFunction = void(const ITensor *in, ITensor *out, unsigned int batch_offset, const Window &window);
diff --git a/arm_compute/core/NEON/kernels/NEDepthConcatenateLayerKernel.h b/arm_compute/core/NEON/kernels/NEDepthConcatenateLayerKernel.h
index e1aaa59f25..3b2b9a1b79 100644
--- a/arm_compute/core/NEON/kernels/NEDepthConcatenateLayerKernel.h
+++ b/arm_compute/core/NEON/kernels/NEDepthConcatenateLayerKernel.h
@@ -76,8 +76,7 @@ public:
     static Status validate(const ITensorInfo *input, unsigned int depth_offset, const ITensorInfo *output);
 
     // Inherited methods overridden:
-    void run_op(const InputTensorMap &inputs, const OutputTensorMap &outputs,
-                const Window &window, const ThreadInfo &info) override;
+    void run_op(ITensorPack &tensors, const Window &window, const ThreadInfo &info) override;
 
 private:
     using DepthConcatFunction = void(const ITensor *in, ITensor *out, unsigned int depth_offset, const Window &window);
diff --git a/arm_compute/core/NEON/kernels/NEElementwiseOperationKernel.h b/arm_compute/core/NEON/kernels/NEElementwiseOperationKernel.h
index cf43aca4b0..47b8c3b7c8 100644
--- a/arm_compute/core/NEON/kernels/NEElementwiseOperationKernel.h
+++ b/arm_compute/core/NEON/kernels/NEElementwiseOperationKernel.h
@@ -67,8 +67,7 @@ public:
     using ElementwiseFunction = void(const ITensor *input1, const ITensor *input2, ITensor *output, const Window &window);
 
     // Inherited methods overridden:
-    void run_op(const InputTensorMap &inputs, const OutputTensorMap &outputs,
-                const Window &window, const ThreadInfo &info) override;
+    void run_op(ITensorPack &tensors, const Window &window, const ThreadInfo &info) override;
 
 protected:
     /** Validate the argument passed to the kernel
diff --git a/arm_compute/core/NEON/kernels/NEHeightConcatenateLayerKernel.h b/arm_compute/core/NEON/kernels/NEHeightConcatenateLayerKernel.h
index d463b53e2c..8a5e86acc4 100644
--- a/arm_compute/core/NEON/kernels/NEHeightConcatenateLayerKernel.h
+++ b/arm_compute/core/NEON/kernels/NEHeightConcatenateLayerKernel.h
@@ -74,8 +74,7 @@ public:
     static Status validate(const ITensorInfo *input, unsigned int height_offset, const ITensorInfo *output);
 
     // Inherited methods overridden:
-    void run_op(const InputTensorMap &inputs, const OutputTensorMap &outputs,
-                const Window &window, const ThreadInfo &info) override;
+    void run_op(ITensorPack &tensors, const Window &window, const ThreadInfo &info) override;
 
 private:
     unsigned int _height_offset;
diff --git a/arm_compute/core/NEON/kernels/NEPixelWiseMultiplicationKernel.h b/arm_compute/core/NEON/kernels/NEPixelWiseMultiplicationKernel.h
index 8876d04c62..c530d78c42 100644
--- a/arm_compute/core/NEON/kernels/NEPixelWiseMultiplicationKernel.h
+++ b/arm_compute/core/NEON/kernels/NEPixelWiseMultiplicationKernel.h
@@ -111,7 +111,7 @@ public:
     static Status validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output, float scale, ConvertPolicy overflow_policy, RoundingPolicy rounding_policy);
 
     // Inherited methods overridden
-    void run_op(const InputTensorMap &inputs, const OutputTensorMap &outputs, const Window &window, const ThreadInfo &info) override;
+    void run_op(ITensorPack &tensors, const Window &window, const ThreadInfo &info) override;
 
 private:
     /** Common signature for all the specialised multiplication functions with integer scaling factor
@@ -178,7 +178,7 @@ public:
     static Status validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output);
 
     // Inherited methods overridden:
-    void run_op(const InputTensorMap &inputs, const OutputTensorMap &outputs, const Window &window, const ThreadInfo &info) override;
+    void run_op(ITensorPack &tensors, const Window &window, const ThreadInfo &info) override;
 };
 
 } // namespace arm_compute
diff --git a/arm_compute/core/NEON/kernels/NEReshapeLayerKernel.h b/arm_compute/core/NEON/kernels/NEReshapeLayerKernel.h
index 97e91ed03d..a4b8426e41 100644
--- a/arm_compute/core/NEON/kernels/NEReshapeLayerKernel.h
+++ b/arm_compute/core/NEON/kernels/NEReshapeLayerKernel.h
@@ -57,8 +57,7 @@ public:
     static Status validate(const ITensorInfo *input, const ITensorInfo *output);
 
     // Inherited methods overridden:
-    void run_op(const InputTensorMap &inputs, const OutputTensorMap &outputs,
-                const Window &window, const ThreadInfo &info) override;
+    void run_op(ITensorPack &tensors, const Window &window, const ThreadInfo &info) override;
 };
 } // namespace arm_compute
 #endif /*ARM_COMPUTE_NERESHAPELAYERKERNEL_H */
diff --git a/arm_compute/core/NEON/kernels/NEStridedSliceKernel.h b/arm_compute/core/NEON/kernels/NEStridedSliceKernel.h
index 59caeecb0d..be55fd75de 100644
--- a/arm_compute/core/NEON/kernels/NEStridedSliceKernel.h
+++ b/arm_compute/core/NEON/kernels/NEStridedSliceKernel.h
@@ -91,8 +91,7 @@ public:
                            int32_t begin_mask, int32_t end_mask, int32_t shrink_axis_mask);
 
     // Inherited methods overridden:
-    void run_op(const InputTensorMap &inputs, const OutputTensorMap &outputs,
-                const Window &window, const ThreadInfo &info) override;
+    void run_op(ITensorPack &tensors, const Window &window, const ThreadInfo &info) override;
 
 private:
     Coordinates _starts_abs;    /**< Absolute start coordinates */
diff --git a/arm_compute/core/NEON/kernels/NEWidthConcatenateLayerKernel.h b/arm_compute/core/NEON/kernels/NEWidthConcatenateLayerKernel.h
index b5336ad026..64d741deab 100644
--- a/arm_compute/core/NEON/kernels/NEWidthConcatenateLayerKernel.h
+++ b/arm_compute/core/NEON/kernels/NEWidthConcatenateLayerKernel.h
@@ -73,8 +73,7 @@ public:
     static Status validate(const ITensorInfo *input, unsigned int width_offset, const ITensorInfo *output);
 
     // Inherited methods overridden:
-    void run_op(const InputTensorMap &inputs, const OutputTensorMap &outputs,
-                const Window &window, const ThreadInfo &info) override;
+    void run_op(ITensorPack &tensors, const Window &window, const ThreadInfo &info) override;
 
 private:
     unsigned int _width_offset;
diff --git a/arm_compute/core/experimental/Types.h b/arm_compute/core/experimental/Types.h
index eca833e708..4dee5ff70d 100644
--- a/arm_compute/core/experimental/Types.h
+++ b/arm_compute/core/experimental/Types.h
@@ -24,13 +24,14 @@
 #ifndef ARM_COMPUTE_EXPERIMENTAL_TYPES_H
 #define ARM_COMPUTE_EXPERIMENTAL_TYPES_H
 
+#include "arm_compute/core/ITensorPack.h"
 #include "arm_compute/core/TensorShape.h"
 
-#include <map>
 #include <vector>
 
 namespace arm_compute
 {
+// Forward declaration
 class ITensor;
 
 /** Memory type */
@@ -51,10 +52,6 @@ enum TensorType : int32_t
     ACL_SRC_VEC = 256,
 };
 
-using InputTensorMap    = std::map<int32_t, const ITensor *>;
-using OutputTensorMap   = std::map<int32_t, ITensor *>;
-using OperatorTensorMap = OutputTensorMap;
-
 namespace experimental
 {
 struct MemoryInfo
diff --git a/arm_compute/runtime/CL/CLScheduler.h b/arm_compute/runtime/CL/CLScheduler.h
index 93595c65c7..8a22832792 100644
--- a/arm_compute/runtime/CL/CLScheduler.h
+++ b/arm_compute/runtime/CL/CLScheduler.h
@@ -76,11 +76,10 @@ public:
     /** Schedule the execution of the passed kernel if possible.
      *
      * @param[in] kernel  Kernel to execute.
-     * @param[in] inputs  Vector containing the input tensors.
-     * @param[in] outputs Vector containing the output tensors.
+     * @param[in] tensors Vector containing the tensors to operate on.
      * @param[in] flush   (Optional) Specifies if the command queue will be flushed after running the kernel.
      */
-    void enqueue_op(ICLKernel &kernel, const InputTensorMap &inputs, const OutputTensorMap &outputs, bool flush = true);
+    void enqueue_op(ICLKernel &kernel, ITensorPack &tensors, bool flush = true);
 
     /** Initialises the context and command queue to be used by the scheduler.
      *
@@ -152,7 +151,7 @@ public:
     bool is_initialised() const;
 
 private:
-    void enqueue_common(ICLKernel &kernel, const InputTensorMap &inputs, const OutputTensorMap &outputs, bool flush);
+    void enqueue_common(ICLKernel &kernel, ITensorPack &tensors, bool flush);
     /** Flag to ensure symbols initialisation is happening before Scheduler creation */
     static std::once_flag _initialize_symbols;
 
diff --git a/arm_compute/runtime/CL/CLTuner.h b/arm_compute/runtime/CL/CLTuner.h
index aa31181d2d..3b45a2177e 100644
--- a/arm_compute/runtime/CL/CLTuner.h
+++ b/arm_compute/runtime/CL/CLTuner.h
@@ -116,7 +116,7 @@ public:
     // Inherited methods overridden:
     void tune_kernel_static(ICLKernel &kernel) override;
     void tune_kernel_dynamic(ICLKernel &kernel) override;
-    void tune_kernel_dynamic(ICLKernel &kernel, const InputTensorMap &inputs, const OutputTensorMap &outputs) override;
+    void tune_kernel_dynamic(ICLKernel &kernel, ITensorPack &tensors) override;
 
     /** Is the kernel_event set ?
      *
@@ -127,11 +127,12 @@ public:
 private:
     /** Find optimal LWS using brute-force approach
      *
-     * @param[in] kernel OpenCL kernel to be tuned with LWS
+     * @param[in]     kernel  OpenCL kernel to be tuned with LWS
+     * @param[in,out] tensors Tensors for the kernel to operate on
      *
      * @return The optimal LWS to use
      */
-    cl::NDRange find_optimal_lws(ICLKernel &kernel, const InputTensorMap &inputs, const OutputTensorMap &outputs);
+    cl::NDRange find_optimal_lws(ICLKernel &kernel, ITensorPack &tensors);
 
     std::unordered_map<std::string, cl::NDRange> _lws_table;
     cl::Event   _kernel_event;
diff --git a/arm_compute/runtime/CL/ICLOperator.h b/arm_compute/runtime/CL/ICLOperator.h
index 2d6c96e815..526b7e93e9 100644
--- a/arm_compute/runtime/CL/ICLOperator.h
+++ b/arm_compute/runtime/CL/ICLOperator.h
@@ -54,8 +54,8 @@ public:
     ICLOperator &operator=(ICLOperator &&) = default;
 
     // Inherited methods overridden:
-    void run(InputTensorMap inputs, OutputTensorMap outputs, OperatorTensorMap workspace) override;
-    void prepare(OperatorTensorMap constants) override;
+    void run(ITensorPack &tensors) override;
+    void prepare(ITensorPack &constants) override;
     MemoryRequirements workspace() const override;
 
 protected:
diff --git a/arm_compute/runtime/CL/ICLTuner.h b/arm_compute/runtime/CL/ICLTuner.h
index 4bc8ddf632..0f951c384e 100644
--- a/arm_compute/runtime/CL/ICLTuner.h
+++ b/arm_compute/runtime/CL/ICLTuner.h
@@ -54,10 +54,9 @@ public:
     /** Tune OpenCL kernel dynamically
      *
      * @param[in]      kernel  Kernel to tune
-     * @param[in]      inputs  Inputs for the kernel to use
-     * @param[in, out] outputs Outputs for the kernel to use
+     * @param[in, out] tensors Tensors for the kernel to use
      */
-    virtual void tune_kernel_dynamic(ICLKernel &kernel, const InputTensorMap &inputs, const OutputTensorMap &outputs) = 0;
+    virtual void tune_kernel_dynamic(ICLKernel &kernel, ITensorPack &tensors) = 0;
 };
 } // namespace arm_compute
 #endif /*ARM_COMPUTE_ICLTUNER_H */
diff --git a/arm_compute/runtime/CL/functions/CLConcatenateLayer.h b/arm_compute/runtime/CL/functions/CLConcatenateLayer.h
index 99a2053a5a..f535c8ea97 100644
--- a/arm_compute/runtime/CL/functions/CLConcatenateLayer.h
+++ b/arm_compute/runtime/CL/functions/CLConcatenateLayer.h
@@ -144,7 +144,7 @@ public:
     static Status validate(const std::vector<const ITensorInfo *> &inputs_vector, const ITensorInfo *output, size_t axis);
 
     // Inherited methods overridden:
-    void run(InputTensorMap inputs, OutputTensorMap outputs, OperatorTensorMap workspace) override;
+    void run(ITensorPack &tensors) override;
 
 private:
     std::vector<std::unique_ptr<ICLKernel>> _concat_kernels;
diff --git a/arm_compute/runtime/CL/functions/CLElementwiseOperations.h b/arm_compute/runtime/CL/functions/CLElementwiseOperations.h
index 5af24c90ac..2d9d43863d 100644
--- a/arm_compute/runtime/CL/functions/CLElementwiseOperations.h
+++ b/arm_compute/runtime/CL/functions/CLElementwiseOperations.h
@@ -98,7 +98,7 @@ public:
     static Status validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output, ConvertPolicy policy, const ActivationLayerInfo &act_info = ActivationLayerInfo());
 
     // Inherited methods overridden:
-    void run(InputTensorMap inputs, OutputTensorMap outputs, OperatorTensorMap workspace) override;
+    void run(ITensorPack &tensors) override;
 
 private:
     CLFillBorderKernel _border_handler;
@@ -168,7 +168,7 @@ public:
     static Status validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output, ConvertPolicy policy, const ActivationLayerInfo &act_info = ActivationLayerInfo());
 
     // Inherited methods overridden:
-    void run(InputTensorMap inputs, OutputTensorMap outputs, OperatorTensorMap workspace) override;
+    void run(ITensorPack &tensors) override;
 
 private:
     CLFillBorderKernel _border_handler;
@@ -207,7 +207,7 @@ public:
     static Status validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output, const ActivationLayerInfo &act_info = ActivationLayerInfo());
 
     // Inherited methods overridden:
-    void run(InputTensorMap inputs, OutputTensorMap outputs, OperatorTensorMap workspace) override;
+    void run(ITensorPack &tensors) override;
 
 private:
     CLFillBorderKernel _border_handler;
@@ -246,7 +246,7 @@ public:
     static Status validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output, const ActivationLayerInfo &act_info = ActivationLayerInfo());
 
     // Inherited methods overridden:
-    void run(InputTensorMap inputs, OutputTensorMap outputs, OperatorTensorMap workspace) override;
+    void run(ITensorPack &tensors) override;
 
 private:
     CLFillBorderKernel _border_handler;
@@ -285,7 +285,7 @@ public:
     static Status validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output, const ActivationLayerInfo &act_info = ActivationLayerInfo());
 
     // Inherited methods overridden:
-    void run(InputTensorMap inputs, OutputTensorMap outputs, OperatorTensorMap workspace) override;
+    void run(ITensorPack &tensors) override;
 
 private:
     CLFillBorderKernel _border_handler;
@@ -324,7 +324,7 @@ public:
     static Status validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output, const ActivationLayerInfo &act_info = ActivationLayerInfo());
 
     // Inherited methods overridden:
-    void run(InputTensorMap inputs, OutputTensorMap outputs, OperatorTensorMap workspace) override;
+    void run(ITensorPack &tensors) override;
 
 private:
     CLFillBorderKernel _border_handler;
@@ -363,7 +363,7 @@ public:
     static Status validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output, const ActivationLayerInfo &act_info = ActivationLayerInfo());
 
     // Inherited methods overridden:
-    void run(InputTensorMap inputs, OutputTensorMap outputs, OperatorTensorMap workspace) override;
+    void run(ITensorPack &tensors) override;
 
 private:
     CLFillBorderKernel _border_handler;
diff --git a/arm_compute/runtime/CL/functions/CLPReluLayer.h b/arm_compute/runtime/CL/functions/CLPReluLayer.h
index 08567cccfb..84743508df 100644
--- a/arm_compute/runtime/CL/functions/CLPReluLayer.h
+++ b/arm_compute/runtime/CL/functions/CLPReluLayer.h
@@ -64,7 +64,7 @@ public:
     static Status validate(const ITensorInfo *input, const ITensorInfo *alpha, const ITensorInfo *output);
 
     // Inherited methods overridden:
-    void run(InputTensorMap inputs, OutputTensorMap outputs, OperatorTensorMap workspace) override;
+    void run(ITensorPack &tensors) override;
 
 private:
     CLFillBorderKernel _border_handler;
diff --git a/arm_compute/runtime/CL/functions/CLPixelWiseMultiplication.h b/arm_compute/runtime/CL/functions/CLPixelWiseMultiplication.h
index ca8d77e6b7..2066012306 100644
--- a/arm_compute/runtime/CL/functions/CLPixelWiseMultiplication.h
+++ b/arm_compute/runtime/CL/functions/CLPixelWiseMultiplication.h
@@ -103,7 +103,7 @@ public:
                            ConvertPolicy overflow_policy, RoundingPolicy rounding_policy, const ActivationLayerInfo &act_info = ActivationLayerInfo());
 
     // Inherited methods overridden:
-    void run(InputTensorMap inputs, OutputTensorMap outputs, OperatorTensorMap workspace) override;
+    void run(ITensorPack &tensors) override;
 
 private:
     CLFillBorderKernel _border_handler;
@@ -136,7 +136,7 @@ public:
     static Status validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output, const ActivationLayerInfo &act_info = ActivationLayerInfo());
 
     // Inherited methods overridden:
-    void run(InputTensorMap inputs, OutputTensorMap outputs, OperatorTensorMap workspace) override;
+    void run(ITensorPack &tensors) override;
 
 private:
     CLFillBorderKernel _border_handler;
diff --git a/arm_compute/runtime/CL/tuners/BifrostTuner.h b/arm_compute/runtime/CL/tuners/BifrostTuner.h
index 830f7d9067..237693fb88 100644
--- a/arm_compute/runtime/CL/tuners/BifrostTuner.h
+++ b/arm_compute/runtime/CL/tuners/BifrostTuner.h
@@ -37,7 +37,7 @@ public:
     // Inherited overriden methods
     void tune_kernel_static(ICLKernel &kernel) override;
     void tune_kernel_dynamic(ICLKernel &kernel) override;
-    void tune_kernel_dynamic(ICLKernel &kernel, const InputTensorMap &inputs, const OutputTensorMap &outputs) override;
+    void tune_kernel_dynamic(ICLKernel &kernel, ITensorPack &tensors) override;
 };
 } // namespace tuners
 } // namespace arm_compute
diff --git a/arm_compute/runtime/CL/tuners/MidgardTuner.h b/arm_compute/runtime/CL/tuners/MidgardTuner.h
index c702e7a2aa..86d46044c2 100644
--- a/arm_compute/runtime/CL/tuners/MidgardTuner.h
+++ b/arm_compute/runtime/CL/tuners/MidgardTuner.h
@@ -37,7 +37,7 @@ public:
     // Inherited overriden methods
     void tune_kernel_static(ICLKernel &kernel) override;
     void tune_kernel_dynamic(ICLKernel &kernel) override;
-    void tune_kernel_dynamic(ICLKernel &kernel, const InputTensorMap &inputs, const OutputTensorMap &outputs) override;
+    void tune_kernel_dynamic(ICLKernel &kernel, ITensorPack &tensors) override;
 };
 } // namespace tuners
 } // namespace arm_compute
diff --git a/arm_compute/runtime/CPP/CPPScheduler.h b/arm_compute/runtime/CPP/CPPScheduler.h
index 9d55ed448e..e8ad427eba 100644
--- a/arm_compute/runtime/CPP/CPPScheduler.h
+++ b/arm_compute/runtime/CPP/CPPScheduler.h
@@ -52,7 +52,7 @@ public:
     void set_num_threads_with_affinity(unsigned int num_threads, BindFunc func) override;
     unsigned int num_threads() const override;
     void schedule(ICPPKernel *kernel, const Hints &hints) override;
-    void schedule_op(ICPPKernel *kernel, const Hints &hints, const InputTensorMap &inputs, const OutputTensorMap &outputs) override;
+    void schedule_op(ICPPKernel *kernel, const Hints &hints, ITensorPack &tensors) override;
 
 protected:
     /** Will run the workloads in parallel using num_threads
@@ -62,7 +62,7 @@ protected:
     void run_workloads(std::vector<Workload> &workloads) override;
 
 private:
-    void schedule_common(ICPPKernel *kernel, const Hints &hints, const InputTensorMap &inputs, const OutputTensorMap &outputs);
+    void schedule_common(ICPPKernel *kernel, const Hints &hints, ITensorPack &tensors);
     struct Impl;
     std::unique_ptr<Impl> _impl;
 };
diff --git a/arm_compute/runtime/IOperator.h b/arm_compute/runtime/IOperator.h
index d72fca4399..e7952bb748 100644
--- a/arm_compute/runtime/IOperator.h
+++ b/arm_compute/runtime/IOperator.h
@@ -40,13 +40,10 @@ public:
     virtual ~IOperator() = default;
     /** Run the kernels contained in the function
      *
-     *
-     * @param[in] inputs    Vector that contains the input tensors.
-     * @param[in] outputs   Vector that contains the output tensors.
-     * @param[in] workspace Vector that contains the workspace tensors.
+     * @param[in] tensors Vector that contains the tensors to operate on.
      *
      */
-    virtual void run(InputTensorMap inputs, OutputTensorMap outputs, OperatorTensorMap workspace) = 0;
+    virtual void run(ITensorPack &tensors) = 0;
     /** Prepare the function for executing
      *
      * Any one off pre-processing step required by the function is handled here
@@ -55,7 +52,7 @@ public:
      *
      * @note Prepare stage might not need all the function's buffers' backing memory to be available in order to execute
      */
-    virtual void prepare(OperatorTensorMap constants) = 0;
+    virtual void prepare(ITensorPack &constants) = 0;
 
     /** Return the memory requirements required by the workspace
      */
diff --git a/arm_compute/runtime/IScheduler.h b/arm_compute/runtime/IScheduler.h
index fff77274bd..98627538e8 100644
--- a/arm_compute/runtime/IScheduler.h
+++ b/arm_compute/runtime/IScheduler.h
@@ -168,10 +168,9 @@ public:
      *
      * @param[in] kernel  Kernel to execute.
      * @param[in] hints   Hints for the scheduler.
-     * @param[in] inputs  Vector containing the input tensors.
-     * @param[in] outputs Vector containing the output tensors.
+     * @param[in] tensors Vector containing the tensors to operate on.
      */
-    virtual void schedule_op(ICPPKernel *kernel, const Hints &hints, const InputTensorMap &inputs, const OutputTensorMap &outputs) = 0;
+    virtual void schedule_op(ICPPKernel *kernel, const Hints &hints, ITensorPack &tensors) = 0;
 
     /** Execute all the passed workloads
      *
diff --git a/arm_compute/runtime/NEON/INEOperator.h b/arm_compute/runtime/NEON/INEOperator.h
index f91305543f..415e767eec 100644
--- a/arm_compute/runtime/NEON/INEOperator.h
+++ b/arm_compute/runtime/NEON/INEOperator.h
@@ -24,6 +24,7 @@
 #ifndef ARM_COMPUTE_INEOPERATOR_H
 #define ARM_COMPUTE_INEOPERATOR_H
 
+#include "../../core/ITensor.h"
 #include "arm_compute/core/NEON/INEKernel.h"
 #include "arm_compute/runtime/IOperator.h"
 #include "arm_compute/runtime/IRuntimeContext.h"
@@ -54,8 +55,8 @@ public:
     INEOperator &operator=(INEOperator &&) = default;
 
     // Inherited methods overridden:
-    void run(InputTensorMap inputs, OutputTensorMap outputs, OperatorTensorMap workspace) override;
-    void prepare(OperatorTensorMap constants) override;
+    void run(ITensorPack &tensors) override;
+    void prepare(ITensorPack &constants) override;
     MemoryRequirements workspace() const override;
 
 protected:
diff --git a/arm_compute/runtime/NEON/functions/NEConcatenateLayer.h b/arm_compute/runtime/NEON/functions/NEConcatenateLayer.h
index 73c62330c5..1d703ae729 100644
--- a/arm_compute/runtime/NEON/functions/NEConcatenateLayer.h
+++ b/arm_compute/runtime/NEON/functions/NEConcatenateLayer.h
@@ -132,7 +132,7 @@ public:
     static Status validate(const std::vector<const ITensorInfo *> &inputs_vector, const ITensorInfo *output, size_t axis);
 
     // Inherited methods overridden:
-    void run(InputTensorMap inputs, OutputTensorMap outputs, OperatorTensorMap workspace) override;
+    void run(ITensorPack &tensors) override;
 
 private:
     std::vector<std::unique_ptr<INEKernel>> _concat_kernels;
diff --git a/arm_compute/runtime/OMP/OMPScheduler.h b/arm_compute/runtime/OMP/OMPScheduler.h
index 56bd6baaa6..56a31cc076 100644
--- a/arm_compute/runtime/OMP/OMPScheduler.h
+++ b/arm_compute/runtime/OMP/OMPScheduler.h
@@ -63,10 +63,9 @@ public:
      *
      * @param[in] kernel  Kernel to execute.
      * @param[in] hints   Hints for the scheduler.
-     * @param[in] inputs  Vector containing the input tensors.
-     * @param[in] outputs Vector containing the output tensors.
+     * @param[in] tensors Vector containing the tensors to operate on.
      */
-    void schedule_op(ICPPKernel *kernel, const Hints &hints, const InputTensorMap &inputs, const OutputTensorMap &outputs) override;
+    void schedule_op(ICPPKernel *kernel, const Hints &hints, ITensorPack &tensors) override;
 
 protected:
     /** Execute all the passed workloads
diff --git a/arm_compute/runtime/SingleThreadScheduler.h b/arm_compute/runtime/SingleThreadScheduler.h
index 42fc742062..d45730e499 100644
--- a/arm_compute/runtime/SingleThreadScheduler.h
+++ b/arm_compute/runtime/SingleThreadScheduler.h
@@ -54,10 +54,9 @@ public:
      *
      * @param[in] kernel  Kernel to execute.
      * @param[in] hints   Hints for the scheduler.
-     * @param[in] inputs  Vector containing the input tensors.
-     * @param[in] outputs Vector containing the output tensors.
+     * @param[in] tensors Vector containing the tensors to operate on.
      */
-    void schedule_op(ICPPKernel *kernel, const Hints &hints, const InputTensorMap &inputs, const OutputTensorMap &outputs) override;
+    void schedule_op(ICPPKernel *kernel, const Hints &hints, ITensorPack &tensors) override;
 
 protected:
     /** Will run the workloads sequentially and in order.