diff --git a/src/runtime/CL/CLOperator.cpp b/src/runtime/CL/CLOperator.cpp
index c41454e..57a4d0e 100644
--- a/src/runtime/CL/CLOperator.cpp
+++ b/src/runtime/CL/CLOperator.cpp
@@ -33,19 +33,17 @@
 {
 }
 
-void ICLOperator::run(InputTensorMap inputs, OutputTensorMap outputs, OperatorTensorMap workspace)
+void ICLOperator::run(ITensorPack &tensors)
 {
-    ARM_COMPUTE_UNUSED(workspace);
-
-    if(inputs.empty())
+    if(tensors.empty())
     {
         ARM_COMPUTE_ERROR("No inputs provided");
     }
 
-    CLScheduler::get().enqueue_op(*_kernel.get(), inputs, outputs, false);
+    CLScheduler::get().enqueue_op(*_kernel.get(), tensors, false);
 }
 
-void ICLOperator::prepare(OperatorTensorMap constants)
+void ICLOperator::prepare(ITensorPack &constants)
 {
     ARM_COMPUTE_UNUSED(constants);
 }
diff --git a/src/runtime/CL/CLScheduler.cpp b/src/runtime/CL/CLScheduler.cpp
index 5ef66f4..ccef5cb 100644
--- a/src/runtime/CL/CLScheduler.cpp
+++ b/src/runtime/CL/CLScheduler.cpp
@@ -151,22 +151,22 @@
     _cl_tuner       = cl_tuner;
 }
 
-void CLScheduler::enqueue_common(ICLKernel &kernel, const InputTensorMap &inputs, const OutputTensorMap &outputs, bool flush)
+void CLScheduler::enqueue_common(ICLKernel &kernel, ITensorPack &tensors, bool flush)
 {
     ARM_COMPUTE_ERROR_ON_MSG(!_is_initialised,
                              "The CLScheduler is not initialised yet! Please call the CLScheduler::get().default_init(), \
                              or CLScheduler::get()::init() and CLKernelLibrary::get()::init() function before running functions!");
 
-    const bool inject_memory = !inputs.empty();
+    const bool inject_memory = !tensors.empty();
 
     // Tune the kernel if the CLTuner has been provided
     if(_cl_tuner != nullptr)
     {
-        inject_memory ? _cl_tuner->tune_kernel_dynamic(kernel, inputs, outputs) : _cl_tuner->tune_kernel_dynamic(kernel);
+        inject_memory ? _cl_tuner->tune_kernel_dynamic(kernel, tensors) : _cl_tuner->tune_kernel_dynamic(kernel);
     }
 
     // Run kernel
-    inject_memory ? kernel.run_op(inputs, outputs, kernel.window(), _queue) : kernel.run(kernel.window(), _queue);
+    inject_memory ? kernel.run_op(tensors, kernel.window(), _queue) : kernel.run(kernel.window(), _queue);
 
     if(flush)
     {
@@ -176,11 +176,12 @@
 
 void CLScheduler::enqueue(ICLKernel &kernel, bool flush)
 {
-    enqueue_common(kernel, {}, {}, flush);
+    ITensorPack pack;
+    enqueue_common(kernel, pack, flush);
 }
 
-void CLScheduler::enqueue_op(ICLKernel &kernel, const InputTensorMap &inputs, const OutputTensorMap &outputs, bool flush)
+void CLScheduler::enqueue_op(ICLKernel &kernel, ITensorPack &tensors, bool flush)
 {
-    enqueue_common(kernel, inputs, outputs, flush);
+    enqueue_common(kernel, tensors, flush);
 }
 } // namespace arm_compute
diff --git a/src/runtime/CL/CLTuner.cpp b/src/runtime/CL/CLTuner.cpp
index b2e3476..adfe67f 100644
--- a/src/runtime/CL/CLTuner.cpp
+++ b/src/runtime/CL/CLTuner.cpp
@@ -77,10 +77,11 @@
 
 void CLTuner::tune_kernel_dynamic(ICLKernel &kernel)
 {
-    tune_kernel_dynamic(kernel, {}, {});
+    ITensorPack pack;
+    tune_kernel_dynamic(kernel, pack);
 }
 
-void CLTuner::tune_kernel_dynamic(ICLKernel &kernel, const InputTensorMap &inputs, const OutputTensorMap &outputs)
+void CLTuner::tune_kernel_dynamic(ICLKernel &kernel, ITensorPack &tensors)
 {
     // Get the configuration ID from the kernel and append GPU target name and number of available compute units
     const std::string config_id = kernel.config_id() + "_" + string_from_target(kernel.get_target()) + "_MP" + support::cpp11::to_string(CLKernelLibrary::get().get_num_compute_units());
@@ -95,7 +96,7 @@
             if(_tune_new_kernels)
             {
                 // Find the optimal LWS for the kernel
-                cl::NDRange opt_lws = find_optimal_lws(kernel, inputs, outputs);
+                cl::NDRange opt_lws = find_optimal_lws(kernel, tensors);
 
                 // Insert the optimal LWS in the table
                 add_lws_to_table(config_id, opt_lws);
@@ -117,7 +118,7 @@
     _lws_table.emplace(kernel_id, optimal_lws);
 }
 
-cl::NDRange CLTuner::find_optimal_lws(ICLKernel &kernel, const InputTensorMap &inputs, const OutputTensorMap &outputs)
+cl::NDRange CLTuner::find_optimal_lws(ICLKernel &kernel, ITensorPack &tensors)
 {
     // Profiling queue
     cl::CommandQueue queue_profiler;
@@ -172,8 +173,8 @@
     cl::NDRange gws = ICLKernel::gws_from_window(kernel.window());
 
     // Run the kernel with default lws to be used as baseline
-    const bool inject_memory = !inputs.empty();
-    inject_memory ? kernel.run_op(inputs, outputs, kernel.window(), queue_profiler) : kernel.run(kernel.window(), queue_profiler);
+    const bool inject_memory = !tensors.empty();
+    inject_memory ? kernel.run_op(tensors, kernel.window(), queue_profiler) : kernel.run(kernel.window(), queue_profiler);
 
     queue_profiler.finish();
 
@@ -203,7 +204,7 @@
         kernel.set_lws_hint(lws_test);
 
         // Run the kernel
-        inject_memory ? kernel.run_op(inputs, outputs, kernel.window(), queue_profiler) : kernel.run(kernel.window(), queue_profiler);
+        inject_memory ? kernel.run_op(tensors, kernel.window(), queue_profiler) : kernel.run(kernel.window(), queue_profiler);
 
         queue_profiler.finish();
 
diff --git a/src/runtime/CL/functions/CLActivationLayer.cpp b/src/runtime/CL/functions/CLActivationLayer.cpp
index 784473d..5ddf227 100644
--- a/src/runtime/CL/functions/CLActivationLayer.cpp
+++ b/src/runtime/CL/functions/CLActivationLayer.cpp
@@ -89,9 +89,9 @@
 
 void CLActivationLayer::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC, _impl->src } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
-
-    _impl->op->run(src, dst, {});
+    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/CL/functions/CLConcatenateLayer.cpp b/src/runtime/CL/functions/CLConcatenateLayer.cpp
index 1ddda02..4214813 100644
--- a/src/runtime/CL/functions/CLConcatenateLayer.cpp
+++ b/src/runtime/CL/functions/CLConcatenateLayer.cpp
@@ -220,16 +220,14 @@
     return Status{};
 }
 
-void CLConcatenation::run(InputTensorMap inputs, OutputTensorMap outputs, OperatorTensorMap workspace)
+void CLConcatenation::run(ITensorPack &tensors)
 {
-    ARM_COMPUTE_UNUSED(workspace);
-
-    if(inputs.empty() || outputs.empty())
+    if(tensors.empty())
     {
         ARM_COMPUTE_ERROR("No inputs provided");
     }
 
-    if(inputs.size() != _num_inputs)
+    if(static_cast<int>(tensors.size()) - 1 != static_cast<int>(_num_inputs))
     {
         ARM_COMPUTE_ERROR("Configured with different number of inputs");
     }
@@ -237,15 +235,17 @@
     if(_axis == Window::DimX && (_num_inputs == 2 || _num_inputs == 4))
     {
         ARM_COMPUTE_ERROR_ON(_concat_kernels.empty());
-        CLScheduler::get().enqueue_op(*_concat_kernels.at(0), inputs, outputs, true);
+        CLScheduler::get().enqueue_op(*_concat_kernels.at(0), tensors, true);
     }
     else
     {
         int i = 0;
         for(auto &k : _concat_kernels)
         {
-            const InputTensorMap input = { { TensorType::ACL_SRC, inputs.at(ACL_SRC_VEC + i) } };
-            CLScheduler::get().enqueue_op(*k, input, outputs, true);
+            ITensorPack pack;
+            pack.add_tensor(TensorType::ACL_SRC, tensors.get_const_tensor(ACL_SRC_VEC + i));
+            pack.add_tensor(TensorType::ACL_DST, tensors.get_tensor(ACL_DST));
+            CLScheduler::get().enqueue_op(*k, pack, true);
             ++i;
         }
     }
@@ -303,13 +303,13 @@
 
 void CLConcatenateLayer::run()
 {
-    InputTensorMap srcs;
+    ITensorPack pack;
     for(unsigned i = 0; i < _impl->num_inputs; ++i)
     {
-        srcs.insert(std::make_pair(TensorType::ACL_SRC_VEC + i, _impl->srcs.at(i)));
+        pack.add_tensor(TensorType::ACL_SRC_VEC + i, _impl->srcs.at(i));
     }
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
+    pack.add_tensor(TensorType::ACL_DST, _impl->dst);
 
-    _impl->op->run(srcs, dst, {});
+    _impl->op->run(pack);
 }
 } // namespace arm_compute
diff --git a/src/runtime/CL/functions/CLElementWiseUnaryLayer.cpp b/src/runtime/CL/functions/CLElementWiseUnaryLayer.cpp
index f8e9694..de94255 100644
--- a/src/runtime/CL/functions/CLElementWiseUnaryLayer.cpp
+++ b/src/runtime/CL/functions/CLElementWiseUnaryLayer.cpp
@@ -153,10 +153,10 @@
 
 void CLRsqrtLayer::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC, _impl->src } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
-
-    _impl->op->run(src, dst, {});
+    ITensorPack pack;
+    pack.add_tensor(TensorType::ACL_SRC, _impl->src);
+    pack.add_tensor(TensorType::ACL_DST, _impl->dst);
+    _impl->op->run(pack);
 }
 
 struct CLExpLayer::Impl
@@ -195,10 +195,10 @@
 
 void CLExpLayer::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC, _impl->src } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
-
-    _impl->op->run(src, dst, {});
+    ITensorPack pack;
+    pack.add_tensor(TensorType::ACL_SRC, _impl->src);
+    pack.add_tensor(TensorType::ACL_DST, _impl->dst);
+    _impl->op->run(pack);
 }
 
 struct CLNegLayer::Impl
@@ -236,10 +236,10 @@
 
 void CLNegLayer::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC, _impl->src } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
-
-    _impl->op->run(src, dst, {});
+    ITensorPack pack;
+    pack.add_tensor(TensorType::ACL_SRC, _impl->src);
+    pack.add_tensor(TensorType::ACL_DST, _impl->dst);
+    _impl->op->run(pack);
 }
 
 struct CLSinLayer::Impl
@@ -277,10 +277,10 @@
 
 void CLSinLayer::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC, _impl->src } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
-
-    _impl->op->run(src, dst, {});
+    ITensorPack pack;
+    pack.add_tensor(TensorType::ACL_SRC, _impl->src);
+    pack.add_tensor(TensorType::ACL_DST, _impl->dst);
+    _impl->op->run(pack);
 }
 
 struct CLAbsLayer::Impl
@@ -318,10 +318,10 @@
 
 void CLAbsLayer::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC, _impl->src } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
-
-    _impl->op->run(src, dst, {});
+    ITensorPack pack;
+    pack.add_tensor(TensorType::ACL_SRC, _impl->src);
+    pack.add_tensor(TensorType::ACL_DST, _impl->dst);
+    _impl->op->run(pack);
 }
 
 struct CLLogLayer::Impl
@@ -359,10 +359,10 @@
 
 void CLLogLayer::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC, _impl->src } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
-
-    _impl->op->run(src, dst, {});
+    ITensorPack pack;
+    pack.add_tensor(TensorType::ACL_SRC, _impl->src);
+    pack.add_tensor(TensorType::ACL_DST, _impl->dst);
+    _impl->op->run(pack);
 }
 
 struct CLRoundLayer::Impl
@@ -400,9 +400,9 @@
 
 void CLRoundLayer::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC, _impl->src } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
-
-    _impl->op->run(src, dst, {});
+    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/CL/functions/CLElementwiseOperations.cpp b/src/runtime/CL/functions/CLElementwiseOperations.cpp
index e66e4bf..6f66472 100644
--- a/src/runtime/CL/functions/CLElementwiseOperations.cpp
+++ b/src/runtime/CL/functions/CLElementwiseOperations.cpp
@@ -47,19 +47,21 @@
     }
 }
 
-void select_border_input(InputTensorMap &tensor_map, InputTensorMap &inputs, OutputTensorMap &outputs)
+ITensorPack select_border_input(ITensorPack &tensors)
 {
-    if(outputs.at(TensorType::ACL_DST)->info()->dimension(0) > 1)
+    ITensorPack pack;
+    if(tensors.get_tensor(TensorType::ACL_DST)->info()->dimension(0) > 1)
     {
-        if(inputs.at(TensorType::ACL_SRC_1)->info()->dimension(0) == 1)
+        if(tensors.get_const_tensor(TensorType::ACL_SRC_1)->info()->dimension(0) == 1)
         {
-            tensor_map[TensorType::ACL_SRC] = inputs.at(TensorType::ACL_SRC_1);
+            pack.add_tensor(TensorType::ACL_SRC, tensors.get_const_tensor(TensorType::ACL_SRC_1));
         }
         else
         {
-            tensor_map[TensorType::ACL_SRC] = inputs.at(TensorType::ACL_SRC_0);
+            pack.add_tensor(TensorType::ACL_SRC, tensors.get_const_tensor(TensorType::ACL_SRC_0));
         }
     }
+    return pack;
 }
 } // namespace
 
@@ -83,12 +85,11 @@
     return CLSaturatedArithmeticOperationKernel::validate(ArithmeticOperation::ADD, input1, input2, output, policy, act_info);
 }
 
-void CLArithmeticAddition::run(InputTensorMap inputs, OutputTensorMap outputs, OperatorTensorMap workspace)
+void CLArithmeticAddition::run(ITensorPack &tensors)
 {
-    InputTensorMap src;
-    select_border_input(src, inputs, outputs);
-    CLScheduler::get().enqueue_op(_border_handler, src, {});
-    ICLOperator::run(inputs, outputs, workspace);
+    auto border_pack = select_border_input(tensors);
+    CLScheduler::get().enqueue_op(_border_handler, border_pack);
+    ICLOperator::run(tensors);
 }
 
 CLArithmeticSubtraction::CLArithmeticSubtraction()
@@ -110,12 +111,11 @@
     return CLSaturatedArithmeticOperationKernel::validate(ArithmeticOperation::SUB, input1, input2, output, policy, act_info);
 }
 
-void CLArithmeticSubtraction::run(InputTensorMap inputs, OutputTensorMap outputs, OperatorTensorMap workspace)
+void CLArithmeticSubtraction::run(ITensorPack &tensors)
 {
-    InputTensorMap src;
-    select_border_input(src, inputs, outputs);
-    CLScheduler::get().enqueue_op(_border_handler, src, {});
-    ICLOperator::run(inputs, outputs, workspace);
+    auto border_pack = select_border_input(tensors);
+    CLScheduler::get().enqueue_op(_border_handler, border_pack);
+    ICLOperator::run(tensors);
 }
 
 CLArithmeticDivision::CLArithmeticDivision()
@@ -136,12 +136,11 @@
     return CLArithmeticOperationKernel::validate(ArithmeticOperation::DIV, input1, input2, output, act_info);
 }
 
-void CLArithmeticDivision::run(InputTensorMap inputs, OutputTensorMap outputs, OperatorTensorMap workspace)
+void CLArithmeticDivision::run(ITensorPack &tensors)
 {
-    InputTensorMap src;
-    select_border_input(src, inputs, outputs);
-    CLScheduler::get().enqueue_op(_border_handler, src, {});
-    ICLOperator::run(inputs, outputs, workspace);
+    auto border_pack = select_border_input(tensors);
+    CLScheduler::get().enqueue_op(_border_handler, border_pack);
+    ICLOperator::run(tensors);
 }
 
 CLElementwiseMax::CLElementwiseMax()
@@ -162,12 +161,11 @@
     return CLArithmeticOperationKernel::validate(ArithmeticOperation::MAX, input1, input2, output, act_info);
 }
 
-void CLElementwiseMax::run(InputTensorMap inputs, OutputTensorMap outputs, OperatorTensorMap workspace)
+void CLElementwiseMax::run(ITensorPack &tensors)
 {
-    InputTensorMap src;
-    select_border_input(src, inputs, outputs);
-    CLScheduler::get().enqueue_op(_border_handler, src, {});
-    ICLOperator::run(inputs, outputs, workspace);
+    auto border_pack = select_border_input(tensors);
+    CLScheduler::get().enqueue_op(_border_handler, border_pack);
+    ICLOperator::run(tensors);
 }
 
 CLElementwiseMin::CLElementwiseMin()
@@ -188,12 +186,11 @@
     return CLArithmeticOperationKernel::validate(ArithmeticOperation::MIN, input1, input2, output, act_info);
 }
 
-void CLElementwiseMin::run(InputTensorMap inputs, OutputTensorMap outputs, OperatorTensorMap workspace)
+void CLElementwiseMin::run(ITensorPack &tensors)
 {
-    InputTensorMap src;
-    select_border_input(src, inputs, outputs);
-    CLScheduler::get().enqueue_op(_border_handler, src, {});
-    ICLOperator::run(inputs, outputs, workspace);
+    auto border_pack = select_border_input(tensors);
+    CLScheduler::get().enqueue_op(_border_handler, border_pack);
+    ICLOperator::run(tensors);
 }
 
 CLElementwiseSquaredDiff::CLElementwiseSquaredDiff()
@@ -214,12 +211,11 @@
     return CLArithmeticOperationKernel::validate(ArithmeticOperation::SQUARED_DIFF, input1, input2, output, act_info);
 }
 
-void CLElementwiseSquaredDiff::run(InputTensorMap inputs, OutputTensorMap outputs, OperatorTensorMap workspace)
+void CLElementwiseSquaredDiff::run(ITensorPack &tensors)
 {
-    InputTensorMap src;
-    select_border_input(src, inputs, outputs);
-    CLScheduler::get().enqueue_op(_border_handler, src, {});
-    ICLOperator::run(inputs, outputs, workspace);
+    auto border_pack = select_border_input(tensors);
+    CLScheduler::get().enqueue_op(_border_handler, border_pack);
+    ICLOperator::run(tensors);
 }
 
 CLElementwisePower::CLElementwisePower()
@@ -240,12 +236,11 @@
     return CLArithmeticOperationKernel::validate(ArithmeticOperation::POWER, input1, input2, output, act_info);
 }
 
-void CLElementwisePower::run(InputTensorMap inputs, OutputTensorMap outputs, OperatorTensorMap workspace)
+void CLElementwisePower::run(ITensorPack &tensors)
 {
-    InputTensorMap src;
-    select_border_input(src, inputs, outputs);
-    CLScheduler::get().enqueue_op(_border_handler, src, {});
-    ICLOperator::run(inputs, outputs, workspace);
+    auto border_pack = select_border_input(tensors);
+    CLScheduler::get().enqueue_op(_border_handler, border_pack);
+    ICLOperator::run(tensors);
 }
 } // namespace experimental
 
@@ -287,10 +282,12 @@
 
 void CLArithmeticAddition::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC_0, _impl->src_0 }, { TensorType::ACL_SRC_1, _impl->src_1 } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
+    ITensorPack pack;
+    pack.add_tensor(TensorType::ACL_SRC_0, _impl->src_0);
+    pack.add_tensor(TensorType::ACL_SRC_1, _impl->src_1);
+    pack.add_tensor(TensorType::ACL_DST, _impl->dst);
 
-    _impl->op->run(src, dst, {});
+    _impl->op->run(pack);
 }
 
 struct CLArithmeticSubtraction::Impl
@@ -331,10 +328,12 @@
 
 void CLArithmeticSubtraction::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC_0, _impl->src_0 }, { TensorType::ACL_SRC_1, _impl->src_1 } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
+    ITensorPack pack;
+    pack.add_tensor(TensorType::ACL_SRC_0, _impl->src_0);
+    pack.add_tensor(TensorType::ACL_SRC_1, _impl->src_1);
+    pack.add_tensor(TensorType::ACL_DST, _impl->dst);
 
-    _impl->op->run(src, dst, {});
+    _impl->op->run(pack);
 }
 
 struct CLArithmeticDivision::Impl
@@ -374,10 +373,12 @@
 
 void CLArithmeticDivision::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC_0, _impl->src_0 }, { TensorType::ACL_SRC_1, _impl->src_1 } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
+    ITensorPack pack;
+    pack.add_tensor(TensorType::ACL_SRC_0, _impl->src_0);
+    pack.add_tensor(TensorType::ACL_SRC_1, _impl->src_1);
+    pack.add_tensor(TensorType::ACL_DST, _impl->dst);
 
-    _impl->op->run(src, dst, {});
+    _impl->op->run(pack);
 }
 
 struct CLElementwiseMax::Impl
@@ -417,10 +418,12 @@
 
 void CLElementwiseMax::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC_0, _impl->src_0 }, { TensorType::ACL_SRC_1, _impl->src_1 } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
+    ITensorPack pack;
+    pack.add_tensor(TensorType::ACL_SRC_0, _impl->src_0);
+    pack.add_tensor(TensorType::ACL_SRC_1, _impl->src_1);
+    pack.add_tensor(TensorType::ACL_DST, _impl->dst);
 
-    _impl->op->run(src, dst, {});
+    _impl->op->run(pack);
 }
 
 struct CLElementwiseMin::Impl
@@ -460,10 +463,12 @@
 
 void CLElementwiseMin::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC_0, _impl->src_0 }, { TensorType::ACL_SRC_1, _impl->src_1 } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
+    ITensorPack pack;
+    pack.add_tensor(TensorType::ACL_SRC_0, _impl->src_0);
+    pack.add_tensor(TensorType::ACL_SRC_1, _impl->src_1);
+    pack.add_tensor(TensorType::ACL_DST, _impl->dst);
 
-    _impl->op->run(src, dst, {});
+    _impl->op->run(pack);
 }
 
 struct CLElementwiseSquaredDiff::Impl
@@ -504,10 +509,12 @@
 
 void CLElementwiseSquaredDiff::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC_0, _impl->src_0 }, { TensorType::ACL_SRC_1, _impl->src_1 } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
+    ITensorPack pack;
+    pack.add_tensor(TensorType::ACL_SRC_0, _impl->src_0);
+    pack.add_tensor(TensorType::ACL_SRC_1, _impl->src_1);
+    pack.add_tensor(TensorType::ACL_DST, _impl->dst);
 
-    _impl->op->run(src, dst, {});
+    _impl->op->run(pack);
 }
 
 struct CLElementwisePower::Impl
@@ -547,9 +554,11 @@
 
 void CLElementwisePower::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC_0, _impl->src_0 }, { TensorType::ACL_SRC_1, _impl->src_1 } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
+    ITensorPack pack;
+    pack.add_tensor(TensorType::ACL_SRC_0, _impl->src_0);
+    pack.add_tensor(TensorType::ACL_SRC_1, _impl->src_1);
+    pack.add_tensor(TensorType::ACL_DST, _impl->dst);
 
-    _impl->op->run(src, dst, {});
+    _impl->op->run(pack);
 }
 } // namespace arm_compute
diff --git a/src/runtime/CL/functions/CLPReluLayer.cpp b/src/runtime/CL/functions/CLPReluLayer.cpp
index fbb466a..e03bd13 100644
--- a/src/runtime/CL/functions/CLPReluLayer.cpp
+++ b/src/runtime/CL/functions/CLPReluLayer.cpp
@@ -44,19 +44,22 @@
         }
     }
 }
-void select_border_input(InputTensorMap &tensor_map, InputTensorMap &inputs, OutputTensorMap &outputs)
+
+ITensorPack select_border_input(ITensorPack &tensors)
 {
-    if(outputs.at(TensorType::ACL_DST)->info()->dimension(0) > 1)
+    ITensorPack pack;
+    if(tensors.get_tensor(TensorType::ACL_DST)->info()->dimension(0) > 1)
     {
-        if(inputs.at(TensorType::ACL_SRC_1)->info()->dimension(0) == 1)
+        if(tensors.get_const_tensor(TensorType::ACL_SRC_1)->info()->dimension(0) == 1)
         {
-            tensor_map[TensorType::ACL_SRC] = inputs.at(TensorType::ACL_SRC_1);
+            pack.add_tensor(TensorType::ACL_SRC, tensors.get_const_tensor(TensorType::ACL_SRC_1));
         }
         else
         {
-            tensor_map[TensorType::ACL_SRC] = inputs.at(TensorType::ACL_SRC_0);
+            pack.add_tensor(TensorType::ACL_SRC, tensors.get_const_tensor(TensorType::ACL_SRC_0));
         }
     }
+    return pack;
 }
 } // namespace
 
@@ -80,12 +83,11 @@
     return CLArithmeticOperationKernel::validate(ArithmeticOperation::PRELU, input, alpha, output);
 }
 
-void CLPReluLayer::run(InputTensorMap inputs, OutputTensorMap outputs, OperatorTensorMap workspace)
+void CLPReluLayer::run(ITensorPack &tensors)
 {
-    InputTensorMap src;
-    select_border_input(src, inputs, outputs);
-    CLScheduler::get().enqueue_op(_border_handler, src, {});
-    ICLOperator::run(inputs, outputs, workspace);
+    auto border_pack = select_border_input(tensors);
+    CLScheduler::get().enqueue_op(_border_handler, border_pack);
+    ICLOperator::run(tensors);
 }
 } // namespace experimental
 
@@ -126,9 +128,11 @@
 
 void CLPReluLayer::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC_0, _impl->src_0 }, { TensorType::ACL_SRC_1, _impl->src_1 } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
+    ITensorPack pack;
+    pack.add_tensor(TensorType::ACL_SRC_0, _impl->src_0);
+    pack.add_tensor(TensorType::ACL_SRC_1, _impl->src_1);
+    pack.add_tensor(TensorType::ACL_DST, _impl->dst);
 
-    _impl->op->run(src, dst, {});
+    _impl->op->run(pack);
 }
 } // namespace arm_compute
diff --git a/src/runtime/CL/functions/CLPixelWiseMultiplication.cpp b/src/runtime/CL/functions/CLPixelWiseMultiplication.cpp
index 34e06a3..883ce68 100644
--- a/src/runtime/CL/functions/CLPixelWiseMultiplication.cpp
+++ b/src/runtime/CL/functions/CLPixelWiseMultiplication.cpp
@@ -34,19 +34,21 @@
 {
 namespace
 {
-void select_border_input(InputTensorMap &tensor_map, InputTensorMap &inputs, OutputTensorMap &outputs)
+ITensorPack select_border_input(ITensorPack &tensors)
 {
-    if(outputs.at(TensorType::ACL_DST)->info()->dimension(0) > 1)
+    ITensorPack pack;
+    if(tensors.get_tensor(TensorType::ACL_DST)->info()->dimension(0) > 1)
     {
-        if(inputs.at(TensorType::ACL_SRC_1)->info()->dimension(0) == 1)
+        if(tensors.get_const_tensor(TensorType::ACL_SRC_1)->info()->dimension(0) == 1)
         {
-            tensor_map[TensorType::ACL_SRC] = inputs.at(TensorType::ACL_SRC_1);
+            pack.add_tensor(TensorType::ACL_SRC, tensors.get_const_tensor(TensorType::ACL_SRC_1));
         }
         else
         {
-            tensor_map[TensorType::ACL_SRC] = inputs.at(TensorType::ACL_SRC_0);
+            pack.add_tensor(TensorType::ACL_SRC, tensors.get_const_tensor(TensorType::ACL_SRC_0));
         }
     }
+    return pack;
 }
 } // namespace
 
@@ -81,12 +83,11 @@
     return CLPixelWiseMultiplicationKernel::validate(input1, input2, output, scale, overflow_policy, rounding_policy, act_info);
 }
 
-void CLPixelWiseMultiplication::run(InputTensorMap inputs, OutputTensorMap outputs, OperatorTensorMap workspace)
+void CLPixelWiseMultiplication::run(ITensorPack &tensors)
 {
-    InputTensorMap src;
-    select_border_input(src, inputs, outputs);
-    CLScheduler::get().enqueue_op(_border_handler, src, {});
-    ICLOperator::run(inputs, outputs, workspace);
+    auto border_pack = select_border_input(tensors);
+    CLScheduler::get().enqueue_op(_border_handler, border_pack);
+    ICLOperator::run(tensors);
 }
 
 CLComplexPixelWiseMultiplication::CLComplexPixelWiseMultiplication()
@@ -116,12 +117,11 @@
     return CLComplexPixelWiseMultiplicationKernel::validate(input1, input2, output, act_info);
 }
 
-void CLComplexPixelWiseMultiplication::run(InputTensorMap inputs, OutputTensorMap outputs, OperatorTensorMap workspace)
+void CLComplexPixelWiseMultiplication::run(ITensorPack &tensors)
 {
-    InputTensorMap src;
-    select_border_input(src, inputs, outputs);
-    CLScheduler::get().enqueue_op(_border_handler, src, {});
-    ICLOperator::run(inputs, outputs, workspace);
+    auto border_pack = select_border_input(tensors);
+    CLScheduler::get().enqueue_op(_border_handler, border_pack);
+    ICLOperator::run(tensors);
 }
 } // namespace experimental
 
@@ -165,10 +165,12 @@
 
 void CLPixelWiseMultiplication::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC_0, _impl->src_0 }, { TensorType::ACL_SRC_1, _impl->src_1 } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
+    ITensorPack pack;
+    pack.add_tensor(TensorType::ACL_SRC_0, _impl->src_0);
+    pack.add_tensor(TensorType::ACL_SRC_1, _impl->src_1);
+    pack.add_tensor(TensorType::ACL_DST, _impl->dst);
 
-    _impl->op->run(src, dst, {});
+    _impl->op->run(pack);
 }
 
 struct CLComplexPixelWiseMultiplication::Impl
@@ -208,9 +210,11 @@
 
 void CLComplexPixelWiseMultiplication::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC_0, _impl->src_0 }, { TensorType::ACL_SRC_1, _impl->src_1 } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
+    ITensorPack pack;
+    pack.add_tensor(TensorType::ACL_SRC_0, _impl->src_0);
+    pack.add_tensor(TensorType::ACL_SRC_1, _impl->src_1);
+    pack.add_tensor(TensorType::ACL_DST, _impl->dst);
 
-    _impl->op->run(src, dst, {});
+    _impl->op->run(pack);
 }
 } // namespace arm_compute
diff --git a/src/runtime/CL/functions/CLReshapeLayer.cpp b/src/runtime/CL/functions/CLReshapeLayer.cpp
index ac8b176..273a761 100644
--- a/src/runtime/CL/functions/CLReshapeLayer.cpp
+++ b/src/runtime/CL/functions/CLReshapeLayer.cpp
@@ -84,10 +84,10 @@
 
 void CLReshapeLayer::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC, _impl->src } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
-
-    _impl->op->run(src, dst, {});
+    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
 /** [CLReshapeLayer snippet] **/
diff --git a/src/runtime/CL/functions/CLSlice.cpp b/src/runtime/CL/functions/CLSlice.cpp
index 3689707..f36550b 100644
--- a/src/runtime/CL/functions/CLSlice.cpp
+++ b/src/runtime/CL/functions/CLSlice.cpp
@@ -97,8 +97,9 @@
 
 void CLSlice::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC, _impl->src } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
-    _impl->op->run(src, dst, {});
+    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/CL/functions/CLStridedSlice.cpp b/src/runtime/CL/functions/CLStridedSlice.cpp
index bdef078..b78073d 100644
--- a/src/runtime/CL/functions/CLStridedSlice.cpp
+++ b/src/runtime/CL/functions/CLStridedSlice.cpp
@@ -96,9 +96,9 @@
 
 void CLStridedSlice::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC, _impl->src } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
-
-    _impl->op->run(src, dst, {});
+    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/CL/tuners/BifrostTuner.cpp b/src/runtime/CL/tuners/BifrostTuner.cpp
index 1797c2c..52644bf 100644
--- a/src/runtime/CL/tuners/BifrostTuner.cpp
+++ b/src/runtime/CL/tuners/BifrostTuner.cpp
@@ -316,9 +316,9 @@
     ARM_COMPUTE_UNUSED(kernel);
 }
 
-void BifrostTuner::tune_kernel_dynamic(ICLKernel &kernel, const InputTensorMap &inputs, const OutputTensorMap &outputs)
+void BifrostTuner::tune_kernel_dynamic(ICLKernel &kernel, ITensorPack &tensors)
 {
-    ARM_COMPUTE_UNUSED(kernel, inputs, outputs);
+    ARM_COMPUTE_UNUSED(kernel, tensors);
 }
 } // namespace tuners
 } // namespace arm_compute
\ No newline at end of file
diff --git a/src/runtime/CL/tuners/MidgardTuner.cpp b/src/runtime/CL/tuners/MidgardTuner.cpp
index 68c98ce..e49e155 100644
--- a/src/runtime/CL/tuners/MidgardTuner.cpp
+++ b/src/runtime/CL/tuners/MidgardTuner.cpp
@@ -74,9 +74,9 @@
     ARM_COMPUTE_UNUSED(kernel);
 }
 
-void MidgardTuner::tune_kernel_dynamic(ICLKernel &kernel, const InputTensorMap &inputs, const OutputTensorMap &outputs)
+void MidgardTuner::tune_kernel_dynamic(ICLKernel &kernel, ITensorPack &tensors)
 {
-    ARM_COMPUTE_UNUSED(kernel, inputs, outputs);
+    ARM_COMPUTE_UNUSED(kernel, tensors);
 }
 } // namespace tuners
 } // namespace arm_compute
diff --git a/src/runtime/CPP/CPPScheduler.cpp b/src/runtime/CPP/CPPScheduler.cpp
index b07aa8c..55f62c1 100644
--- a/src/runtime/CPP/CPPScheduler.cpp
+++ b/src/runtime/CPP/CPPScheduler.cpp
@@ -405,7 +405,7 @@
 }
 #endif /* DOXYGEN_SKIP_THIS */
 
-void CPPScheduler::schedule_common(ICPPKernel *kernel, const Hints &hints, const InputTensorMap &inputs, const OutputTensorMap &outputs)
+void CPPScheduler::schedule_common(ICPPKernel *kernel, const Hints &hints, ITensorPack &tensors)
 {
     ARM_COMPUTE_ERROR_ON_MSG(!kernel, "The child class didn't set the kernel");
 
@@ -464,13 +464,13 @@
         {
             ThreadInfo info;
             info.cpu_info = &_cpu_info;
-            if(inputs.empty())
+            if(tensors.empty())
             {
                 kernel->run(max_window, info);
             }
             else
             {
-                kernel->run_op(inputs, outputs, max_window, info);
+                kernel->run_op(tensors, max_window, info);
             }
         }
         else
@@ -495,18 +495,18 @@
             for(unsigned int t = 0; t < num_windows; t++)
             {
                 //Capture 't' by copy, all the other variables by reference:
-                workloads[t] = [t, &hints, &max_window, &num_windows, &kernel, &inputs, &outputs](const ThreadInfo & info)
+                workloads[t] = [t, &hints, &max_window, &num_windows, &kernel, &tensors](const ThreadInfo & info)
                 {
                     Window win = max_window.split_window(hints.split_dimension(), t, num_windows);
                     win.validate();
 
-                    if(inputs.empty())
+                    if(tensors.empty())
                     {
                         kernel->run(win, info);
                     }
                     else
                     {
-                        kernel->run_op(inputs, outputs, win, info);
+                        kernel->run_op(tensors, win, info);
                     }
                 };
             }
@@ -515,15 +515,14 @@
     }
 }
 
-void CPPScheduler::schedule_op(ICPPKernel *kernel, const Hints &hints, const InputTensorMap &inputs, const OutputTensorMap &outputs)
+void CPPScheduler::schedule_op(ICPPKernel *kernel, const Hints &hints, ITensorPack &tensors)
 {
-    schedule_common(kernel, hints, inputs, outputs);
+    schedule_common(kernel, hints, tensors);
 }
 
 void CPPScheduler::schedule(ICPPKernel *kernel, const Hints &hints)
 {
-    const InputTensorMap inputs;
-    OutputTensorMap      outputs;
-    schedule_common(kernel, hints, inputs, outputs);
+    ITensorPack tensors;
+    schedule_common(kernel, hints, tensors);
 }
 } // namespace arm_compute
diff --git a/src/runtime/CPP/SingleThreadScheduler.cpp b/src/runtime/CPP/SingleThreadScheduler.cpp
index a672656..6f67bc0 100644
--- a/src/runtime/CPP/SingleThreadScheduler.cpp
+++ b/src/runtime/CPP/SingleThreadScheduler.cpp
@@ -49,12 +49,12 @@
     kernel->run(kernel->window(), info);
 }
 
-void SingleThreadScheduler::schedule_op(ICPPKernel *kernel, const Hints &hints, const InputTensorMap &inputs, const OutputTensorMap &outputs)
+void SingleThreadScheduler::schedule_op(ICPPKernel *kernel, const Hints &hints, ITensorPack &tensors)
 {
     ARM_COMPUTE_UNUSED(hints);
     ThreadInfo info;
     info.cpu_info = &_cpu_info;
-    kernel->run_op(inputs, outputs, kernel->window(), info);
+    kernel->run_op(tensors, kernel->window(), info);
 }
 
 void SingleThreadScheduler::run_workloads(std::vector<Workload> &workloads)
diff --git a/src/runtime/NEON/INEOperator.cpp b/src/runtime/NEON/INEOperator.cpp
index 1d81997..75068b1 100644
--- a/src/runtime/NEON/INEOperator.cpp
+++ b/src/runtime/NEON/INEOperator.cpp
@@ -33,19 +33,17 @@
 {
 }
 
-void INEOperator::run(InputTensorMap inputs, OutputTensorMap outputs, OperatorTensorMap workspace)
+void INEOperator::run(ITensorPack &tensors)
 {
-    ARM_COMPUTE_UNUSED(workspace);
-
-    if(inputs.empty() || outputs.empty())
+    if(tensors.empty())
     {
         ARM_COMPUTE_ERROR("No inputs provided");
     }
 
-    NEScheduler::get().schedule_op(_kernel.get(), Window::DimY, inputs, outputs);
+    NEScheduler::get().schedule_op(_kernel.get(), Window::DimY, tensors);
 }
 
-void INEOperator::prepare(OperatorTensorMap constants)
+void INEOperator::prepare(ITensorPack &constants)
 {
     ARM_COMPUTE_UNUSED(constants);
 }
diff --git a/src/runtime/NEON/functions/NEActivationLayer.cpp b/src/runtime/NEON/functions/NEActivationLayer.cpp
index 0e75e58..7f55edb 100644
--- a/src/runtime/NEON/functions/NEActivationLayer.cpp
+++ b/src/runtime/NEON/functions/NEActivationLayer.cpp
@@ -85,9 +85,9 @@
 
 void NEActivationLayer::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC, _impl->src } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
-
-    _impl->op->run(src, dst, {});
+    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/NEON/functions/NEArithmeticAddition.cpp b/src/runtime/NEON/functions/NEArithmeticAddition.cpp
index b18309e..4453a01 100644
--- a/src/runtime/NEON/functions/NEArithmeticAddition.cpp
+++ b/src/runtime/NEON/functions/NEArithmeticAddition.cpp
@@ -79,8 +79,10 @@
 
 void NEArithmeticAddition::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC_0, _impl->src_0 }, { TensorType::ACL_SRC_1, _impl->src_1 } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
-    _impl->op->run(src, dst, {});
+    ITensorPack pack;
+    pack.add_tensor(TensorType::ACL_SRC_0, _impl->src_0);
+    pack.add_tensor(TensorType::ACL_SRC_1, _impl->src_1);
+    pack.add_tensor(TensorType::ACL_DST, _impl->dst);
+    _impl->op->run(pack);
 }
 } // namespace arm_compute
diff --git a/src/runtime/NEON/functions/NEArithmeticSubtraction.cpp b/src/runtime/NEON/functions/NEArithmeticSubtraction.cpp
index c7f492b..1c95bbf 100644
--- a/src/runtime/NEON/functions/NEArithmeticSubtraction.cpp
+++ b/src/runtime/NEON/functions/NEArithmeticSubtraction.cpp
@@ -80,8 +80,10 @@
 
 void NEArithmeticSubtraction::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC_0, _impl->src_0 }, { TensorType::ACL_SRC_1, _impl->src_1 } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
-    _impl->op->run(src, dst, {});
+    ITensorPack pack;
+    pack.add_tensor(TensorType::ACL_SRC_0, _impl->src_0);
+    pack.add_tensor(TensorType::ACL_SRC_1, _impl->src_1);
+    pack.add_tensor(TensorType::ACL_DST, _impl->dst);
+    _impl->op->run(pack);
 }
 } // namespace arm_compute
diff --git a/src/runtime/NEON/functions/NEConcatenateLayer.cpp b/src/runtime/NEON/functions/NEConcatenateLayer.cpp
index 9f8a2a1..8df4f4c 100644
--- a/src/runtime/NEON/functions/NEConcatenateLayer.cpp
+++ b/src/runtime/NEON/functions/NEConcatenateLayer.cpp
@@ -146,16 +146,14 @@
     return Status{};
 }
 
-void NEConcatenation::run(InputTensorMap inputs, OutputTensorMap outputs, OperatorTensorMap workspace)
+void NEConcatenation::run(ITensorPack &tensors)
 {
-    ARM_COMPUTE_UNUSED(workspace);
-
-    if(inputs.empty() || outputs.empty())
+    if(tensors.empty())
     {
         ARM_COMPUTE_ERROR("No inputs provided");
     }
 
-    if(inputs.size() != _num_inputs)
+    if(static_cast<int>(tensors.size() - 1) != static_cast<int>(_num_inputs))
     {
         ARM_COMPUTE_ERROR("Configured with different number of inputs");
     }
@@ -163,8 +161,10 @@
     int i = 0;
     for(auto &k : _concat_kernels)
     {
-        const InputTensorMap input = { { TensorType::ACL_SRC, inputs.at(ACL_SRC_VEC + i) } };
-        NEScheduler::get().schedule_op(k.get(), Window::DimY, input, outputs);
+        ITensorPack pack;
+        pack.add_tensor(TensorType::ACL_SRC, tensors.get_const_tensor(ACL_SRC_VEC + i));
+        pack.add_tensor(TensorType::ACL_DST, tensors.get_tensor(ACL_DST));
+        NEScheduler::get().schedule_op(k.get(), Window::DimY, pack);
         ++i;
     }
 }
@@ -216,13 +216,13 @@
 
 void NEConcatenateLayer::run()
 {
-    InputTensorMap srcs;
+    ITensorPack pack;
     for(unsigned i = 0; i < _impl->num_inputs; ++i)
     {
-        srcs.insert(std::make_pair(TensorType::ACL_SRC_VEC + i, _impl->srcs.at(i)));
+        pack.add_tensor(TensorType::ACL_SRC_VEC + i, _impl->srcs.at(i));
     }
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
+    pack.add_tensor(TensorType::ACL_DST, _impl->dst);
 
-    _impl->op->run(srcs, dst, {});
+    _impl->op->run(pack);
 }
 } // namespace arm_compute
diff --git a/src/runtime/NEON/functions/NEElementwiseOperators.cpp b/src/runtime/NEON/functions/NEElementwiseOperators.cpp
index 9340cc0..d1f60c7 100644
--- a/src/runtime/NEON/functions/NEElementwiseOperators.cpp
+++ b/src/runtime/NEON/functions/NEElementwiseOperators.cpp
@@ -163,9 +163,11 @@
 
 void NEElementwiseMax::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC_0, _impl->src_0 }, { TensorType::ACL_SRC_1, _impl->src_1 } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
-    _impl->op->run(src, dst, {});
+    ITensorPack pack;
+    pack.add_tensor(TensorType::ACL_SRC_0, _impl->src_0);
+    pack.add_tensor(TensorType::ACL_SRC_1, _impl->src_1);
+    pack.add_tensor(TensorType::ACL_DST, _impl->dst);
+    _impl->op->run(pack);
 }
 
 struct NEElementwiseMin::Impl
@@ -202,9 +204,11 @@
 
 void NEElementwiseMin::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC_0, _impl->src_0 }, { TensorType::ACL_SRC_1, _impl->src_1 } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
-    _impl->op->run(src, dst, {});
+    ITensorPack pack;
+    pack.add_tensor(TensorType::ACL_SRC_0, _impl->src_0);
+    pack.add_tensor(TensorType::ACL_SRC_1, _impl->src_1);
+    pack.add_tensor(TensorType::ACL_DST, _impl->dst);
+    _impl->op->run(pack);
 }
 
 struct NEElementwiseSquaredDiff::Impl
@@ -241,9 +245,11 @@
 
 void NEElementwiseSquaredDiff::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC_0, _impl->src_0 }, { TensorType::ACL_SRC_1, _impl->src_1 } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
-    _impl->op->run(src, dst, {});
+    ITensorPack pack;
+    pack.add_tensor(TensorType::ACL_SRC_0, _impl->src_0);
+    pack.add_tensor(TensorType::ACL_SRC_1, _impl->src_1);
+    pack.add_tensor(TensorType::ACL_DST, _impl->dst);
+    _impl->op->run(pack);
 }
 
 struct NEElementwiseDivision::Impl
@@ -280,9 +286,11 @@
 
 void NEElementwiseDivision::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC_0, _impl->src_0 }, { TensorType::ACL_SRC_1, _impl->src_1 } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
-    _impl->op->run(src, dst, {});
+    ITensorPack pack;
+    pack.add_tensor(TensorType::ACL_SRC_0, _impl->src_0);
+    pack.add_tensor(TensorType::ACL_SRC_1, _impl->src_1);
+    pack.add_tensor(TensorType::ACL_DST, _impl->dst);
+    _impl->op->run(pack);
 }
 
 struct NEElementwisePower::Impl
@@ -319,9 +327,11 @@
 
 void NEElementwisePower::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC_0, _impl->src_0 }, { TensorType::ACL_SRC_1, _impl->src_1 } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
-    _impl->op->run(src, dst, {});
+    ITensorPack pack;
+    pack.add_tensor(TensorType::ACL_SRC_0, _impl->src_0);
+    pack.add_tensor(TensorType::ACL_SRC_1, _impl->src_1);
+    pack.add_tensor(TensorType::ACL_DST, _impl->dst);
+    _impl->op->run(pack);
 }
 
 template <ComparisonOperation COP>
@@ -364,9 +374,11 @@
 template <ComparisonOperation COP>
 void                          NEElementwiseComparisonStatic<COP>::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC_0, _impl->src_0 }, { TensorType::ACL_SRC_1, _impl->src_1 } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
-    _impl->op->run(src, dst, {});
+    ITensorPack pack;
+    pack.add_tensor(TensorType::ACL_SRC_0, _impl->src_0);
+    pack.add_tensor(TensorType::ACL_SRC_1, _impl->src_1);
+    pack.add_tensor(TensorType::ACL_DST, _impl->dst);
+    _impl->op->run(pack);
 }
 
 struct NEElementwiseComparison::Impl
@@ -401,9 +413,11 @@
 
 void NEElementwiseComparison::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC_0, _impl->src_0 }, { TensorType::ACL_SRC_1, _impl->src_1 } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
-    _impl->op->run(src, dst, {});
+    ITensorPack pack;
+    pack.add_tensor(TensorType::ACL_SRC_0, _impl->src_0);
+    pack.add_tensor(TensorType::ACL_SRC_1, _impl->src_1);
+    pack.add_tensor(TensorType::ACL_DST, _impl->dst);
+    _impl->op->run(pack);
 }
 
 // Supported Specializations
diff --git a/src/runtime/NEON/functions/NEPReluLayer.cpp b/src/runtime/NEON/functions/NEPReluLayer.cpp
index 15d9fd9..f9393a4 100644
--- a/src/runtime/NEON/functions/NEPReluLayer.cpp
+++ b/src/runtime/NEON/functions/NEPReluLayer.cpp
@@ -71,9 +71,11 @@
 
 void NEPReluLayer::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC_0, _impl->src_0 }, { TensorType::ACL_SRC_1, _impl->src_1 } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
-    _impl->op->run(src, dst, {});
+    ITensorPack pack;
+    pack.add_tensor(TensorType::ACL_SRC_0, _impl->src_0);
+    pack.add_tensor(TensorType::ACL_SRC_1, _impl->src_1);
+    pack.add_tensor(TensorType::ACL_DST, _impl->dst);
+    _impl->op->run(pack);
 }
 
 Status NEPReluLayer::validate(const ITensorInfo *input, const ITensorInfo *alpha, const ITensorInfo *output)
diff --git a/src/runtime/NEON/functions/NEPixelWiseMultiplication.cpp b/src/runtime/NEON/functions/NEPixelWiseMultiplication.cpp
index ba5dd7c..4208878 100644
--- a/src/runtime/NEON/functions/NEPixelWiseMultiplication.cpp
+++ b/src/runtime/NEON/functions/NEPixelWiseMultiplication.cpp
@@ -97,9 +97,11 @@
 
 void NEPixelWiseMultiplication::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC_0, _impl->src_0 }, { TensorType::ACL_SRC_1, _impl->src_1 } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
-    _impl->op->run(src, dst, {});
+    ITensorPack pack;
+    pack.add_tensor(TensorType::ACL_SRC_0, _impl->src_0);
+    pack.add_tensor(TensorType::ACL_SRC_1, _impl->src_1);
+    pack.add_tensor(TensorType::ACL_DST, _impl->dst);
+    _impl->op->run(pack);
 }
 
 struct NEComplexPixelWiseMultiplication::Impl
@@ -134,8 +136,10 @@
 
 void NEComplexPixelWiseMultiplication::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC_0, _impl->src_0 }, { TensorType::ACL_SRC_1, _impl->src_1 } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
-    _impl->op->run(src, dst, {});
+    ITensorPack pack;
+    pack.add_tensor(TensorType::ACL_SRC_0, _impl->src_0);
+    pack.add_tensor(TensorType::ACL_SRC_1, _impl->src_1);
+    pack.add_tensor(TensorType::ACL_DST, _impl->dst);
+    _impl->op->run(pack);
 }
 } // namespace arm_compute
diff --git a/src/runtime/NEON/functions/NEReshapeLayer.cpp b/src/runtime/NEON/functions/NEReshapeLayer.cpp
index 47d5519..c1c88c1 100644
--- a/src/runtime/NEON/functions/NEReshapeLayer.cpp
+++ b/src/runtime/NEON/functions/NEReshapeLayer.cpp
@@ -84,9 +84,9 @@
 
 void NEReshapeLayer::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC, _impl->src } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
-
-    _impl->op->run(src, dst, {});
+    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/NEON/functions/NESlice.cpp b/src/runtime/NEON/functions/NESlice.cpp
index 7c32521..2bacf2e 100644
--- a/src/runtime/NEON/functions/NESlice.cpp
+++ b/src/runtime/NEON/functions/NESlice.cpp
@@ -94,9 +94,10 @@
 
 void NESlice::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC, _impl->src } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
-    _impl->op->run(src, dst, {});
+    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/NEON/functions/NEStridedSlice.cpp b/src/runtime/NEON/functions/NEStridedSlice.cpp
index 37e3590..8bf81e8 100644
--- a/src/runtime/NEON/functions/NEStridedSlice.cpp
+++ b/src/runtime/NEON/functions/NEStridedSlice.cpp
@@ -76,9 +76,10 @@
 
 void NEStridedSlice::run()
 {
-    const InputTensorMap  src{ { TensorType::ACL_SRC, _impl->src } };
-    const OutputTensorMap dst{ { TensorType::ACL_DST, _impl->dst } };
-    _impl->op->run(src, dst, {});
+    ITensorPack pack;
+    pack.add_tensor(TensorType::ACL_SRC, _impl->src);
+    pack.add_tensor(TensorType::ACL_DST, _impl->dst);
+    _impl->op->run(pack);
 }
 
 Status NEStridedSlice::validate(const ITensorInfo *input, const ITensorInfo *output,
diff --git a/src/runtime/OMP/OMPScheduler.cpp b/src/runtime/OMP/OMPScheduler.cpp
index 1016018..11448e5 100644
--- a/src/runtime/OMP/OMPScheduler.cpp
+++ b/src/runtime/OMP/OMPScheduler.cpp
@@ -83,7 +83,7 @@
     }
 }
 
-void OMPScheduler::schedule_op(ICPPKernel *kernel, const Hints &hints, const InputTensorMap &inputs, const OutputTensorMap &outputs)
+void OMPScheduler::schedule_op(ICPPKernel *kernel, const Hints &hints, ITensorPack &tensors)
 {
     ARM_COMPUTE_ERROR_ON_MSG(!kernel, "The child class didn't set the kernel");
     ARM_COMPUTE_ERROR_ON_MSG(hints.strategy() == StrategyHint::DYNAMIC,
@@ -97,7 +97,7 @@
     {
         ThreadInfo info;
         info.cpu_info = &_cpu_info;
-        kernel->run_op(inputs, outputs, max_window, info);
+        kernel->run_op(tensors, max_window, info);
     }
     else
     {
@@ -106,11 +106,11 @@
         for(unsigned int t = 0; t < num_windows; t++)
         {
             //Capture 't' by copy, all the other variables by reference:
-            workloads[t] = [t, &hints, &max_window, &num_windows, &kernel, &inputs, &outputs](const ThreadInfo & info)
+            workloads[t] = [t, &hints, &max_window, &num_windows, &kernel, &tensors](const ThreadInfo & info)
             {
                 Window win = max_window.split_window(hints.split_dimension(), t, num_windows);
                 win.validate();
-                kernel->run_op(inputs, outputs, win, info);
+                kernel->run_op(tensors, win, info);
             };
         }
         run_workloads(workloads);