Remove dynamic fusion prototype with tests and examples

Public headers of the new experimental dynamic fusion can be found in arm_compute/dynamic_fusion/
New examples on how to use the interface can be found in tests/validation/dynamic_fusion/gpu/Integration.cpp

Resolves COMPMID-5683

Change-Id: I7ccb902a227fb487562df15fc3c30118d1d95bbd
Signed-off-by: SiCong Li <sicong.li@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/8671
Reviewed-by: Jakub Sujak <jakub.sujak@arm.com>
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>

6 files changed, 0 insertions, 1483 deletions

diff --git a/tests/validation/CL/UNIT/dynamic_fusion/ArbitraryElementwiseFusion.cpp b/tests/validation/CL/UNIT/dynamic_fusion/ArbitraryElementwiseFusion.cpp
deleted file mode 100644
index 1b1e8aa761..0000000000
--- a/tests/validation/CL/UNIT/dynamic_fusion/ArbitraryElementwiseFusion.cpp
+++ /dev/null
@@ -1,394 +0,0 @@
-/*
- * Copyright (c) 2022 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.
- */
-#ifdef ENABLE_EXPERIMENTAL_DYNAMIC_FUSION
-
-#include "src/core/experimental/dynamic_fusion/ClKernelBuildingAPI.h"
-#include "src/core/utils/helpers/float_ops.h"
-#include "tests/CL/CLAccessor.h"
-#include "tests/framework/Macros.h"
-#include "tests/validation/Validation.h"
-#include "tests/validation/reference/ConvolutionLayer.h"
-#include "tests/validation/reference/ElementwiseOperations.h"
-#include "tests/validation/reference/Permute.h"
-
-#include "arm_compute/runtime/experimental/ClCompositeOperator.h"
-#include "tests/validation/reference/Floor.h"
-
-#include "arm_compute/core/ITensor.h"
-#include "arm_compute/runtime/CL/CLTensor.h"
-#include "tests/validation/CL/UNIT/dynamic_fusion/Utils.h"
-
-using namespace arm_compute::experimental::dynamic_fusion;
-using namespace arm_compute::test::validation::utils;
-
-namespace arm_compute
-{
-namespace test
-{
-namespace validation
-{
-TEST_SUITE(CL)
-TEST_SUITE(UNIT)
-TEST_SUITE(DYNAMIC_FUSION)
-TEST_SUITE(ArbitraryFusion)
-
-TEST_CASE(ElementwiseBroadcasting, framework::DatasetMode::ALL)
-{
-    // Test elementwise broadcasting
-    const auto data_type   = DataType::F32;
-    const auto data_layout = DataLayout::NHWC;
-
-    const auto input_shape = TensorShape(7, 9, 5);
-    const auto rhs_shape   = TensorShape(7, 1, 1);
-    const auto dst_shape   = TensorShape(7, 9, 5);
-
-    // Tensor Info
-    auto input_info  = TensorInfo(input_shape, 1, data_type, data_layout);
-    auto addend_info = TensorInfo(rhs_shape, 1, data_type, data_layout);
-    auto dst_info    = TensorInfo();
-
-    ElementwiseDescriptor add_desc{ ArithmeticOperation::ADD };
-
-    CLScheduler::get().default_reinit();
-    const auto    cl_compile_ctx = CLKernelLibrary::get().get_compile_context();
-    OperatorGraph op_graph;
-
-    const auto op_input  = add_tensor(op_graph, input_info);
-    const auto op_addend = add_tensor(op_graph, addend_info);
-    const auto op_dst    = add_tensor(op_graph, dst_info);
-
-    add_op_elementwise_op(op_graph, add_desc, op_input, op_addend, op_dst);
-
-    const ClWorkloadContext workload_ctx{ GpuInfo{ CLScheduler::get().target() } };
-    ClWorkload              workload;
-    build(workload, op_graph, workload_ctx);
-
-    ClCompositeOperator op;
-    op.configure(cl_compile_ctx, workload);
-
-    // Construct tensors
-    CLTensor t_input{};
-    CLTensor t_addend{};
-    CLTensor t_dst{};
-
-    // Init tensors
-    t_input.allocator()->init(input_info);
-    t_addend.allocator()->init(addend_info);
-    t_dst.allocator()->init(dst_info);
-
-    // Allocate and fill tensors
-    t_input.allocator()->allocate();
-    t_addend.allocator()->allocate();
-    t_dst.allocator()->allocate();
-
-    // Fill
-    fill<float>(CLAccessor(t_input), 0, library.get());
-    fill<float>(CLAccessor(t_addend), 1, library.get());
-
-    // Pack tensors
-    OpTensorBinding bp_tensors({ { op_input, &t_input },
-        { op_addend, &t_addend },
-        { op_dst, &t_dst }
-    });
-
-    // Populate prepare and run pack-maps (including allocating aux tensors)
-    ClAuxTensorData aux_tensor_data{};
-    TensorPackMap   prepare_pack_map{};
-    TensorPackMap   run_pack_map{};
-    bind_tensors(aux_tensor_data, prepare_pack_map, run_pack_map, workload, bp_tensors);
-
-    op.prepare(prepare_pack_map);
-    op.run(run_pack_map);
-
-    // Create reference
-    SimpleTensor<float> ref_input{ input_shape, data_type, 1, QuantizationInfo(), DataLayout::NHWC };
-    SimpleTensor<float> ref_addend{ rhs_shape, data_type, 1, QuantizationInfo(), DataLayout::NHWC };
-
-    // Fill reference
-    fill<float>(ref_input, 0, library.get());
-    fill<float>(ref_addend, 1, library.get());
-
-    auto ref_input_nchw  = reference::permute(ref_input, PermutationVector(1U, 2U, 0U));
-    auto ref_addend_nchw = reference::permute(ref_addend, PermutationVector(1U, 2U, 0U));
-
-    auto dst_shape_nchw = dst_shape;
-    permute(dst_shape_nchw, PermutationVector(1U, 2U, 0U));
-
-    auto ref_t_dst_nchw = reference::arithmetic_operation(
-                              ArithmeticOperation::ADD,
-                              ref_input_nchw,
-                              ref_addend_nchw,
-                              data_type,
-                              ConvertPolicy{});
-
-    const auto ref_t_dst = reference::permute(ref_t_dst_nchw, PermutationVector(2U, 0U, 1U));
-
-    RelativeTolerance<float> tolerance_f32(0.001f);
-    validate(CLAccessor(t_dst), ref_t_dst_nchw, tolerance_f32);
-}
-TEST_CASE(DivFloor, framework::DatasetMode::ALL)
-{
-    // x = floor(div(input, input2))
-    const auto data_type    = DataType::F32;
-    const auto eltwise_info = ElementwiseDescriptor{ ArithmeticOperation::DIV };
-
-    // Tensor Values
-    const auto width  = 7U;
-    const auto height = 6U;
-
-    // Shapes
-    const auto input1_shape = TensorShape(width, height);
-    const auto input2_shape = TensorShape(width, height);
-    const auto dst_shape    = TensorShape(width, height);
-
-    // Create reference
-    SimpleTensor<float> ref_src_nhwc{ input1_shape, data_type, 1, QuantizationInfo(), DataLayout::NHWC };
-    SimpleTensor<float> ref_src2_nhwc{ input2_shape, data_type, 1, QuantizationInfo(), DataLayout::NHWC };
-
-    // Fill reference
-    fill<float>(ref_src_nhwc, 0, library.get());
-    fill<float>(ref_src2_nhwc, 1, library.get());
-
-    auto ref_src  = reference::permute(ref_src_nhwc, PermutationVector(1U, 2U, 0U));
-    auto ref_src2 = reference::permute(ref_src2_nhwc, PermutationVector(1U, 2U, 0U));
-
-    TensorShape dst_shape_nchw{ dst_shape };
-    permute(dst_shape_nchw, PermutationVector(1U, 2U, 0U));
-
-    const auto ref_dst_nchw = reference::floor_layer(reference::arithmetic_operation(
-                                                         ArithmeticOperation::DIV,
-                                                         ref_src,
-                                                         ref_src2,
-                                                         data_type,
-                                                         ConvertPolicy::SATURATE));
-
-    const auto ref_t_dst = reference::permute(ref_dst_nchw, PermutationVector(2U, 0U, 1U));
-
-    // Tensor Info
-    auto input1_info = TensorInfo(input1_shape, 1, data_type, DataLayout::NHWC);
-    auto input2_info = TensorInfo(input2_shape, 1, data_type, DataLayout::NHWC);
-    auto dst_info    = TensorInfo();
-    auto acc_info    = TensorInfo(); // Intermediate tensor for division
-
-    // Initialise Scheduler
-    CLScheduler::get().default_reinit();
-    const auto    cl_compile_ctx = CLKernelLibrary::get().get_compile_context();
-    OperatorGraph op_graph;
-
-    // add tensors
-    auto op_input1 = add_tensor(op_graph, input1_info);
-    auto op_input2 = add_tensor(op_graph, input2_info);
-    auto op_acc    = add_tensor(op_graph, acc_info);
-    auto op_dst    = add_tensor(op_graph, dst_info);
-
-    add_op_elementwise_op(op_graph, eltwise_info, op_input1, op_input2, op_acc);
-    add_op_floor(op_graph, FloorDescriptor(), op_acc, op_dst);
-
-    const ClWorkloadContext workload_ctx{ GpuInfo{ CLScheduler::get().target() } };
-    ClWorkload              workload;
-    build(workload, op_graph, workload_ctx);
-
-    ClCompositeOperator op;
-    op.configure(cl_compile_ctx, workload);
-
-    // Configure and add tensors.
-    CLTensor t_input1{};
-    CLTensor t_input2{};
-    CLTensor t_dst{};
-
-    // Init Tensors
-    t_input1.allocator()->init(input1_info);
-    t_input2.allocator()->init(input2_info);
-    t_dst.allocator()->init(dst_info);
-
-    // Allocate and fill tensors
-    t_input1.allocator()->allocate();
-    t_input2.allocator()->allocate();
-    t_dst.allocator()->allocate();
-
-    fill<float>(CLAccessor(t_input1), 0, library.get());
-    fill<float>(CLAccessor(t_input2), 1, library.get());
-
-    // "Pack" tensors
-    OpTensorBinding bp_tensors({ { op_input1, &t_input1 },
-        { op_input2, &t_input2 },
-        { op_dst, &t_dst }
-    });
-
-    // Populate prepare and run pack-maps (including allocating aux tensors)
-    ClAuxTensorData aux_tensor_data{};
-    TensorPackMap   prepare_pack_map{};
-    TensorPackMap   run_pack_map{};
-    bind_tensors(aux_tensor_data, prepare_pack_map, run_pack_map, workload, bp_tensors);
-
-    op.prepare(prepare_pack_map);
-    op.run(run_pack_map);
-
-    RelativeTolerance<float> tolerance_f32(0.001f); /**< Tolerance value for comparing reference's output against implementation's output for floating point data types */
-    validate(CLAccessor(t_dst), ref_dst_nchw, tolerance_f32);
-}
-TEST_CASE(Dconv2dAddDiv, framework::DatasetMode::ALL)
-{
-    // output = div(divend, add(addend, conv2d1x1(direct_conv)(input, weights, bias)))
-    const auto data_type   = DataType::F32;
-    const auto data_layout = DataLayout::NHWC;
-
-    const auto input_shape  = TensorShape(384, 12, 12);
-    const auto weight_shape = TensorShape(384, 1, 1, 16);
-    const auto dst_shape    = TensorShape(16, 12, 12);
-
-    // Tensor Info
-    auto input_info  = TensorInfo(input_shape, 1, data_type, data_layout);
-    auto weight_info = TensorInfo(weight_shape, 1, data_type, data_layout);
-    auto addend_info = TensorInfo(dst_shape, 1, data_type, data_layout);
-    auto divend_info = TensorInfo(dst_shape, 1, data_type, data_layout);
-    auto acc_info    = TensorInfo(); // Intermediate tensor for conv
-    auto acc_1_info  = TensorInfo();
-    auto dst_info    = TensorInfo();
-
-    Conv2dDescriptor      conv2d_desc{};
-    ElementwiseDescriptor add_desc{ ArithmeticOperation::ADD };
-    ElementwiseDescriptor div_desc{ ArithmeticOperation::DIV };
-
-    CLScheduler::get().default_reinit();
-    const auto    cl_compile_ctx = CLKernelLibrary::get().get_compile_context();
-    OperatorGraph op_graph;
-
-    const auto op_input  = add_tensor(op_graph, input_info);
-    const auto op_weight = add_tensor(op_graph, weight_info);
-    const auto op_addend = add_tensor(op_graph, addend_info);
-    const auto op_divend = add_tensor(op_graph, divend_info);
-    const auto op_acc    = add_tensor(op_graph, acc_info);   // temp accumulator; TensorInfo to be inferred
-    const auto op_acc_1  = add_tensor(op_graph, acc_1_info); // temp accumulator; TensorInfo to be inferred
-    const auto op_dst    = add_tensor(op_graph, dst_info);
-
-    auto conv2d = add_op_conv2d(op_graph, conv2d_desc, op_input, op_weight, op_acc);
-    force_conv2d_method(op_graph, conv2d, ConvolutionMethod::DIRECT);
-    add_op_elementwise_op(op_graph, add_desc, op_acc, op_addend, op_acc_1);
-    add_op_elementwise_op(op_graph, div_desc, op_acc_1, op_divend, op_dst);
-
-    const ClWorkloadContext workload_ctx{ GpuInfo{ CLScheduler::get().target() } };
-    ClWorkload              workload;
-    build(workload, op_graph, workload_ctx);
-
-    ClCompositeOperator op;
-    op.configure(cl_compile_ctx, workload);
-
-    // Construct tensors
-    CLTensor t_input{};
-    CLTensor t_weight{};
-    CLTensor t_addend{};
-    CLTensor t_divend{};
-    CLTensor t_dst{};
-
-    // Init tensors
-    t_input.allocator()->init(input_info);
-    t_weight.allocator()->init(weight_info);
-    t_divend.allocator()->init(divend_info);
-    t_addend.allocator()->init(addend_info);
-    t_dst.allocator()->init(dst_info);
-
-    // Allocate and fill tensors
-    t_input.allocator()->allocate();
-    t_weight.allocator()->allocate();
-    t_divend.allocator()->allocate();
-    t_addend.allocator()->allocate();
-    t_dst.allocator()->allocate();
-
-    // Fill
-    fill<float>(CLAccessor(t_input), 0, library.get());
-    fill<float>(CLAccessor(t_weight), 1, library.get());
-    fill<float>(CLAccessor(t_addend), 2, library.get());
-    fill<float>(CLAccessor(t_divend), 3, library.get());
-
-    // Pack tensors
-    OpTensorBinding bp_tensors({ { op_input, &t_input },
-        { op_weight, &t_weight },
-        { op_addend, &t_addend },
-        { op_divend, &t_divend },
-        { op_dst, &t_dst }
-    });
-
-    // Populate prepare and run pack-maps (including allocating aux tensors)
-    ClAuxTensorData aux_tensor_data{};
-    TensorPackMap   prepare_pack_map{};
-    TensorPackMap   run_pack_map{};
-    bind_tensors(aux_tensor_data, prepare_pack_map, run_pack_map, workload, bp_tensors);
-
-    op.prepare(prepare_pack_map);
-    op.run(run_pack_map);
-
-    // Create reference
-    SimpleTensor<float> ref_input{ input_shape, data_type, 1, QuantizationInfo(), DataLayout::NHWC };
-    SimpleTensor<float> ref_weight{ weight_shape, data_type, 1, QuantizationInfo(), DataLayout::NHWC };
-    SimpleTensor<float> ref_bias_placeholder{ dst_shape, data_type, 1, QuantizationInfo(), DataLayout::NHWC };
-    SimpleTensor<float> ref_addend{ dst_shape, data_type, 1, QuantizationInfo(), DataLayout::NHWC };
-    SimpleTensor<float> ref_divend{ dst_shape, data_type, 1, QuantizationInfo(), DataLayout::NHWC };
-
-    // Fill reference
-    fill<float>(ref_input, 0, library.get());
-    fill<float>(ref_weight, 1, library.get());
-    fill<float>(ref_addend, 2, library.get());
-    fill<float>(ref_divend, 3, library.get());
-
-    auto ref_input_nchw            = reference::permute(ref_input, PermutationVector(1U, 2U, 0U));
-    auto ref_weight_nchw           = reference::permute(ref_weight, PermutationVector(1U, 2U, 0U));
-    auto ref_bias_placeholder_nchw = reference::permute(ref_bias_placeholder, PermutationVector(1U, 2U, 0U));
-    auto ref_addend_nchw           = reference::permute(ref_addend, PermutationVector(1U, 2U, 0U));
-    auto ref_divend_nchw           = reference::permute(ref_divend, PermutationVector(1U, 2U, 0U));
-
-    auto dst_shape_nchw = dst_shape;
-    permute(dst_shape_nchw, PermutationVector(1U, 2U, 0U));
-
-    PadStrideInfo legacy_pad_stride(conv2d_desc.stride.x(), conv2d_desc.stride.y(), conv2d_desc.pad.left, conv2d_desc.pad.right, conv2d_desc.pad.top, conv2d_desc.pad.bottom, DimensionRoundingType{});
-    auto          ref_acc_nchw = reference::arithmetic_operation(
-                                     ArithmeticOperation::ADD,
-                                     ref_addend_nchw,
-                                     reference::convolution_layer(ref_input_nchw, ref_weight_nchw, ref_bias_placeholder_nchw, dst_shape_nchw, legacy_pad_stride, conv2d_desc.dilation),
-                                     data_type,
-                                     ConvertPolicy{});
-
-    auto ref_t_dst_nchw = reference::arithmetic_operation(
-                              ArithmeticOperation::DIV,
-                              ref_acc_nchw,
-                              ref_divend_nchw,
-                              data_type,
-                              ConvertPolicy{});
-
-    const auto ref_t_dst = reference::permute(ref_t_dst_nchw, PermutationVector(2U, 0U, 1U));
-
-    RelativeTolerance<float> tolerance_f32(0.001f);
-    validate(CLAccessor(t_dst), ref_t_dst_nchw, tolerance_f32);
-}
-
-TEST_SUITE_END() // ArbitraryFusion
-TEST_SUITE_END() // DYNAMIC_FUSION
-TEST_SUITE_END() // UNIT
-TEST_SUITE_END() // CL
-
-} // namespace validation
-} // namespace test
-} // namespace arm_compute
-
-#endif /* ENABLE_EXPERIMENTAL_DYNAMIC_FUSION */
diff --git a/tests/validation/CL/UNIT/dynamic_fusion/ClCompositeKernel.cpp b/tests/validation/CL/UNIT/dynamic_fusion/ClCompositeKernel.cpp
deleted file mode 100644
index dc98d72f4b..0000000000
--- a/tests/validation/CL/UNIT/dynamic_fusion/ClCompositeKernel.cpp
+++ /dev/null
@@ -1,215 +0,0 @@
-/*
- * Copyright (c) 2022 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.
- */
-#ifdef ENABLE_EXPERIMENTAL_DYNAMIC_FUSION
-
-#include "src/gpu/cl/kernels/experimental/dynamic_fusion/ClCompositeKernel.h"
-#include "src/core/experimental/dynamic_fusion/ClKernelBuildingAPI.h"
-
-#include "src/core/utils/helpers/float_ops.h"
-#include "src/gpu/cl/kernels/ClElementwiseKernel.h"
-#include "src/gpu/cl/kernels/ClGemmMatrixMultiplyNativeKernel.h"
-#include "tests/CL/CLAccessor.h"
-#include "tests/framework/Macros.h"
-#include "tests/framework/datasets/Datasets.h"
-#include "tests/validation/Validation.h"
-#include "tests/validation/reference/ConvolutionLayer.h"
-#include "tests/validation/reference/ElementwiseOperations.h"
-#include "tests/validation/reference/GEMM.h"
-#include "tests/validation/reference/Permute.h"
-
-#include "arm_compute/core/utils/misc/ShapeCalculator.h"
-#include "src/core/AccessWindowStatic.h"
-#include "src/core/helpers/AutoConfiguration.h"
-#include "src/core/helpers/WindowHelpers.h"
-
-#include "tests/validation/CL/UNIT/dynamic_fusion/Utils.h"
-
-#include <chrono>
-
-using namespace arm_compute::experimental::dynamic_fusion;
-using namespace arm_compute::test::validation::utils;
-
-namespace arm_compute
-{
-namespace test
-{
-namespace validation
-{
-TEST_SUITE(CL)
-TEST_SUITE(UNIT)
-TEST_SUITE(DYNAMIC_FUSION)
-TEST_SUITE(ClCompositeKernel)
-TEST_SUITE(Validate)
-
-TEST_CASE(MoveNet_SubGraph_1_DirectConv2d, framework::DatasetMode::ALL)
-{
-    /* Computation:
-     * out = add(addend, direct_conv2d(lhs, rhs, bias)) (non-broadcast)
-     */
-
-    ClCompositeKernel     kernel{};
-    ClKernelBlueprint     bp{};
-    ClKernelCode          cl_code{};
-    ClExecutionDescriptor exec_desc{};
-    Status                st{};
-
-    const auto data_type    = DataType::F32;
-    const auto conv_info    = Conv2dDescriptor{ Padding2D{ 1U, 1U, 1U, 1U }, { 1U, 1U } /* stride */ };
-    const auto eltwise_info = ElementwiseDescriptor{ ArithmeticOperation::ADD };
-
-    const auto width     = 7U;
-    const auto height    = 6U;
-    const auto IFM       = 5U;
-    const auto OFM       = 4U;
-    const auto kernel_sz = 3U;
-
-    const auto src_shape    = TensorShape(IFM, width, height);
-    const auto wei_shape    = TensorShape(IFM, kernel_sz, kernel_sz, OFM);
-    const auto bia_shape    = TensorShape(OFM);
-    const auto addend_shape = TensorShape(1, 1);
-    const auto dst_shape    = TensorShape(OFM, width, height);
-
-    auto src_info    = TensorInfo(src_shape, 1, data_type, DataLayout::NHWC);
-    auto wei_info    = TensorInfo(wei_shape, 1, data_type, DataLayout::NHWC);
-    auto bia_info    = TensorInfo(bia_shape, 1, data_type, DataLayout::NHWC);
-    auto addend_info = TensorInfo(addend_shape, 1, data_type, DataLayout::NHWC);
-    auto dst_info    = TensorInfo(dst_shape, 1, data_type, DataLayout::NHWC);
-
-    const auto n0 = std::min(OFM, 4u);
-    const auto m0 = (OFM > 16) ? ((data_type == DataType::F32) ? 2U : 4U) : 1U;
-
-    const ClDirectConv2dKernelDescriptor direct_conv2d_desc{ conv_info };
-    const ClElementwiseKernelDescriptor  eltwise_add_desc{ eltwise_info };
-    const TileDescriptor                 store_tile_info{ Size2D(n0, m0), Size2D(width, height), ClippingStrategy::TOP_LEFT };
-
-    ArgumentID src_id{ g_arg_placeholder };
-    ArgumentID wei_id{ g_arg_placeholder };
-    ArgumentID bia_id{ g_arg_placeholder };
-    ArgumentID acc_id{ g_arg_placeholder };
-    ArgumentID acc_1_id{ g_arg_placeholder };
-    ArgumentID addend_id{ g_arg_placeholder };
-    ArgumentID dst_id{ g_arg_placeholder };
-
-    st = add_tensor(bp, &src_info, src_id);
-    st = add_tensor(bp, &wei_info, wei_id);
-    st = add_tensor(bp, &bia_info, bia_id);
-    st = add_tensor(bp, &dst_info, acc_id);
-    st = add_tensor(bp, &dst_info, acc_1_id);
-    st = add_tensor(bp, &addend_info, addend_id);
-    st = add_tensor(bp, &dst_info, dst_id);
-
-    st = add_kcomp_direct_conv2d(bp, direct_conv2d_desc, src_id, wei_id, bia_id, acc_id);
-    st = add_kcomp_eltwise_op(bp, eltwise_add_desc, addend_id, acc_id, acc_1_id);
-    st = add_kcomp_store(bp, StoreType::TStoreIndirectWidthSelect, acc_1_id, dst_id);
-
-    exec_desc.skip_sliding_window = true;
-
-    st = set_tile_info(bp, store_tile_info);
-    st = build(cl_code, ClCodeBuilderContext{ GpuInfo{ GPUTarget::G71 } }, bp);
-    st = tune_static(exec_desc, cl_code);
-
-    CLScheduler::get().default_reinit();
-    kernel.configure(CLKernelLibrary::get().get_compile_context(), cl_code);
-
-    // Construct tensors
-    CLTensor src{};
-    CLTensor wei{};
-    CLTensor bia{};
-    CLTensor addend{};
-    CLTensor dst{};
-
-    // Init tensors
-    src.allocator()->init(src_info);
-    wei.allocator()->init(wei_info);
-    bia.allocator()->init(bia_info);
-    addend.allocator()->init(dst_info);
-    dst.allocator()->init(dst_info);
-
-    // "Pack" tensors
-    ITensorPack tensors{ { src_id, &src },
-        { wei_id, &wei },
-        { bia_id, &bia },
-        { addend_id, &addend },
-        { dst_id, &dst } };
-
-    // Allocate and fill tensors
-    src.allocator()->allocate();
-    wei.allocator()->allocate();
-    bia.allocator()->allocate();
-    addend.allocator()->allocate();
-    dst.allocator()->allocate();
-
-    fill<float>(CLAccessor(src), 0, library.get());
-    fill<float>(CLAccessor(wei), 1, library.get());
-    fill<float>(CLAccessor(bia), 2, library.get());
-    fill<float>(CLAccessor(addend), 3, library.get());
-
-    CLScheduler::get().enqueue_op(kernel, tensors, exec_desc, true);
-
-    // Create reference
-    SimpleTensor<float> ref_src_nhwc{ src_shape, data_type, 1, QuantizationInfo(), DataLayout::NHWC };
-    SimpleTensor<float> ref_wei_nhwc{ wei_shape, data_type, 1, QuantizationInfo(), DataLayout::NHWC };
-    SimpleTensor<float> ref_bia_nhwc{ bia_shape, data_type, 1, QuantizationInfo(), DataLayout::NHWC };
-    SimpleTensor<float> ref_addend_nhwc{ addend_shape, data_type, 1, QuantizationInfo(), DataLayout::NHWC };
-
-    // Fill reference
-    fill<float>(ref_src_nhwc, 0, library.get());
-    fill<float>(ref_wei_nhwc, 1, library.get());
-    fill<float>(ref_bia_nhwc, 2, library.get());
-    fill<float>(ref_addend_nhwc, 3, library.get());
-
-    auto ref_src    = reference::permute(ref_src_nhwc, PermutationVector(1U, 2U, 0U));
-    auto ref_wei    = reference::permute(ref_wei_nhwc, PermutationVector(1U, 2U, 0U));
-    auto ref_bia    = reference::permute(ref_bia_nhwc, PermutationVector(1U, 2U, 0U));
-    auto ref_addend = reference::permute(ref_addend_nhwc, PermutationVector(1U, 2U, 0U));
-
-    TensorShape dst_shape_nchw{ dst_shape };
-    permute(dst_shape_nchw, PermutationVector(1U, 2U, 0U));
-
-    const auto ref_dst = reference::arithmetic_operation(
-                             ArithmeticOperation::ADD,
-                             ref_addend,
-                             reference::convolution_layer<float>(ref_src, ref_wei, ref_bia, dst_shape_nchw,
-                                                                 PadStrideInfo
-    {
-        static_cast<unsigned int>(conv_info.stride.x()),
-        static_cast<unsigned int>(conv_info.stride.y()),
-        static_cast<unsigned int>(conv_info.pad.left),
-        static_cast<unsigned int>(conv_info.pad.top) }),
-    data_type,
-    ConvertPolicy::SATURATE);
-
-    RelativeTolerance<float> tolerance_f32(0.001f); /**< Tolerance value for comparing reference's output against implementation's output for floating point data types */
-    validate(CLAccessor(dst), ref_dst, tolerance_f32);
-}
-
-TEST_SUITE_END() // Validate
-TEST_SUITE_END() // ClCompositeKernel
-TEST_SUITE_END() // DYNAMIC_FUSION
-TEST_SUITE_END() // UNIT
-TEST_SUITE_END() // CL
-} // namespace validation
-} // namespace test
-} // namespace arm_compute
-#endif /* ENABLE_EXPERIMENTAL_DYNAMIC_FUSION */
\ No newline at end of file
diff --git a/tests/validation/CL/UNIT/dynamic_fusion/DependencyGraph.cpp b/tests/validation/CL/UNIT/dynamic_fusion/DependencyGraph.cpp
deleted file mode 100644
index 1824efff99..0000000000
--- a/tests/validation/CL/UNIT/dynamic_fusion/DependencyGraph.cpp
+++ /dev/null
@@ -1,266 +0,0 @@
-/*
- * Copyright (c) 2022 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.
- */
-#ifdef ENABLE_EXPERIMENTAL_DYNAMIC_FUSION
-#include "arm_compute/core/experimental/DependencyGraph.h"
-
-#include "tests/framework/Asserts.h"
-#include "tests/framework/Macros.h"
-
-using namespace arm_compute::experimental::dynamic_fusion;
-
-namespace arm_compute
-{
-namespace test
-{
-namespace validation
-{
-TEST_SUITE(CL)
-
-TEST_SUITE(UNIT)
-TEST_SUITE(DYNAMIC_FUSION)
-TEST_SUITE(DependencyGraph)
-
-TEST_CASE(Correct_Graph_Creation_Should_Pass, framework::DatasetMode::ALL)
-{
-    DependencyGraph graph{};
-    const auto      t0 = graph.add_tensor();
-    const auto      t1 = graph.add_tensor();
-    const auto      t2 = graph.add_tensor();
-    const auto      t3 = graph.add_tensor();
-    const auto      t4 = graph.add_tensor();
-
-    const auto o0 = graph.add_operator({ t0, t1 }, { t2 }).second;
-    const auto o1 = graph.add_operator({ t3, t2 }, { t4 }).second;
-
-    ARM_COMPUTE_EXPECT_EQUAL(graph.number_of_ops(), 2U, framework::LogLevel::ERRORS);
-    ARM_COMPUTE_EXPECT_EQUAL(graph.number_of_tensors(), 5U, framework::LogLevel::ERRORS);
-
-    const DependencyGraph ref_graph
-    {
-        {
-            // src_tensors
-            { o0, { t0, t1 } },
-            { o1, { t3, t2 } },
-        },
-        {
-            // dst_tensors
-            { o0, { t2 } },
-            { o1, { t4 } },
-        },
-        {
-            // src_ops
-            { t0, {} },
-            { t1, {} },
-            { t2, { o0 } },
-            { t3, {} },
-            { t4, { o1 } },
-        },
-        {
-            // dst_ops
-            { t0, { o0 } },
-            { t1, { o0 } },
-            { t2, { o1 } },
-            { t3, { o1 } },
-            { t4, {} },
-        }
-
-    };
-    ARM_COMPUTE_EXPECT(graph == ref_graph, framework::LogLevel::ERRORS);
-}
-
-TEST_CASE(Correct_Merge_Points_Should_Enable_Graph_Expansion, framework::DatasetMode::ALL)
-{
-    // Merge points are a simple way to collapse "graph of graphs" into a single graph
-    // Suppose we have a top-level graph g0
-    DependencyGraph g0{};
-    const auto      g0_t0 = g0.add_tensor();
-    const auto      g0_t1 = g0.add_tensor();
-    const auto      g0_t2 = g0.add_tensor();
-    const auto      g0_t3 = g0.add_tensor();
-    const auto      g0_t4 = g0.add_tensor();
-    g0.add_operator({ g0_t0, g0_t1 }, { g0_t2 }); // g0_o0
-    g0.add_operator({ g0_t3, g0_t2 }, { g0_t4 }); // g0_o1
-
-    // Then g0 expands into g1, with additional nodes added in-between "merge point tensors"
-    // Note that the expansion logic may be local to each operator node
-    DependencyGraph g1{};
-    // g0_o0 expands into g1_o0, g1_o1, g1_o2
-    const auto g1_t0 = g1.add_tensor(g0_t0);
-    const auto g1_t1 = g1.add_tensor(g0_t1);
-    const auto g1_t2 = g1.add_tensor();
-    const auto g1_t3 = g1.add_tensor();
-    const auto g1_t4 = g1.add_tensor(g0_t2);
-    const auto g1_o0 = g1.add_operator({ g1_t0 }, { g1_t2 }).second;
-    const auto g1_o1 = g1.add_operator({ g1_t1 }, { g1_t3 }).second;
-    const auto g1_o2 = g1.add_operator({ g1_t2, g1_t3 }, { g1_t4 }).second;
-
-    // g0_o1 expands into g1_o3
-    const auto g1_t5 = g1.add_tensor(g0_t3);
-    const auto g1_t6 = g1.add_tensor(g0_t2);
-    const auto g1_t7 = g1.add_tensor(g0_t4);
-    ARM_COMPUTE_EXPECT_EQUAL(g1_t4, g1_t6, framework::LogLevel::ERRORS); // both associate with the same merge point g0_t2, thus they should point to the same tensor in g1
-    const auto g1_o3 = g1.add_operator({ g1_t5, g1_t6 }, { g1_t7 }).second;
-
-    const DependencyGraph ref_graph
-    {
-        {
-            // src_tensors
-            { g1_o0, { g1_t0 } },
-            { g1_o1, { g1_t1 } },
-            { g1_o2, { g1_t2, g1_t3 } },
-            { g1_o3, { g1_t5, g1_t4 } },
-        },
-        {
-            // dst_tensors
-            { g1_o0, { g1_t2 } },
-            { g1_o1, { g1_t3 } },
-            { g1_o2, { g1_t4 } },
-            { g1_o3, { g1_t7 } },
-        },
-        {
-            // src_ops
-            { g1_t0, {} },
-            { g1_t1, {} },
-            { g1_t2, { g1_o0 } },
-            { g1_t3, { g1_o1 } },
-            { g1_t4, { g1_o2 } },
-            { g1_t5, {} },
-            { g1_t7, { g1_o3 } },
-        },
-        {
-            // dst_ops
-            { g1_t0, { g1_o0 } },
-            { g1_t1, { g1_o1 } },
-            { g1_t2, { g1_o2 } },
-            { g1_t3, { g1_o2 } },
-            { g1_t4, { g1_o3 } },
-            { g1_t5, { g1_o3 } },
-            { g1_t7, {} },
-        },
-        {
-            // merge points
-            { g0_t0, g1_t0 },
-            { g0_t1, g1_t1 },
-            { g0_t2, g1_t4 },
-            { g0_t3, g1_t5 },
-            { g0_t4, g1_t7 },
-        }
-    };
-    ARM_COMPUTE_EXPECT(g1 == ref_graph, framework::LogLevel::ERRORS);
-}
-
-TEST_CASE(Path_Existence_Check_0, framework::DatasetMode::ALL)
-{
-    DependencyGraph graph{};
-    const auto      t0 = graph.add_tensor();
-    const auto      t1 = graph.add_tensor();
-    const auto      t2 = graph.add_tensor();
-    const auto      t3 = graph.add_tensor();
-    const auto      t4 = graph.add_tensor();
-    const auto      t5 = graph.add_tensor();
-    const auto      t6 = graph.add_tensor();
-    const auto      t7 = graph.add_tensor();
-    const auto      o0 = graph.add_operator({ t1 }, { t3, t4 }).second;
-    const auto      o1 = graph.add_operator({ t3 }, { t5 }).second;
-    const auto      o2 = graph.add_operator({ t5, t6 }, { t7 }).second;
-    const auto      o3 = graph.add_operator({ t4 }, { t6 }).second;
-    const auto      o4 = graph.add_operator({ t0, t5 }, { t2 }).second;
-
-    ARM_COMPUTE_UNUSED(o1, o3);
-
-    ARM_COMPUTE_EXPECT((graph.path_exists_from_tensor_to_op(t3, o2)), framework::LogLevel::ERRORS);
-    ARM_COMPUTE_EXPECT((graph.path_exists_from_tensor_to_op(t1, o4)), framework::LogLevel::ERRORS);
-    ARM_COMPUTE_EXPECT(!(graph.path_exists_from_tensor_to_op(t2, o4)), framework::LogLevel::ERRORS);
-    ARM_COMPUTE_EXPECT(!(graph.path_exists_from_tensor_to_op(t0, o2)), framework::LogLevel::ERRORS);
-
-    ARM_COMPUTE_EXPECT((graph.path_exists_from_op_to_op(o0, o2)), framework::LogLevel::ERRORS);
-    ARM_COMPUTE_EXPECT(!(graph.path_exists_from_op_to_op(o2, o0)), framework::LogLevel::ERRORS);
-
-    ARM_COMPUTE_EXPECT(!(graph.path_exists_from_op_to_op(o2, o4)), framework::LogLevel::ERRORS);
-}
-
-TEST_CASE(Correct_Topological_Sort_Should_Pass, framework::DatasetMode::ALL)
-{
-    DependencyGraph graph{};
-    const auto      t0 = graph.add_tensor();
-    const auto      t1 = graph.add_tensor();
-    const auto      t2 = graph.add_tensor();
-    const auto      t3 = graph.add_tensor();
-    const auto      t4 = graph.add_tensor();
-    const auto      t5 = graph.add_tensor();
-    const auto      t6 = graph.add_tensor();
-    const auto      t7 = graph.add_tensor();
-    const auto      o0 = graph.add_operator({ t1 }, { t3, t4 }).second;
-    const auto      o1 = graph.add_operator({ t3 }, { t5 }).second;
-    const auto      o2 = graph.add_operator({ t5, t6 }, { t7 }).second;
-    const auto      o3 = graph.add_operator({ t4 }, { t6 }).second;
-    const auto      o4 = graph.add_operator({ t0, t5 }, { t2 }).second;
-
-    const auto res = graph.topological_sort();
-    ARM_COMPUTE_EXPECT(bool(res.first), framework::LogLevel::ERRORS);
-    std::vector<DependencyGraph::OpPack> ref_sorted_op_packs
-    {
-        { o0, { t1 }, { t3, t4 } },
-        { o1, { t3 }, { t5 } },
-        { o3, { t4 }, { t6 } },
-        { o4, { t0, t5 }, { t2 } },
-        { o2, { t5, t6 }, { t7 } },
-
-    };
-    ARM_COMPUTE_EXPECT((res.second == ref_sorted_op_packs), framework::LogLevel::ERRORS);
-}
-
-TEST_CASE(Cycles_Should_Fail, framework::DatasetMode::ALL)
-{
-    DependencyGraph graph{};
-    const auto      t0 = graph.add_tensor();
-    const auto      t1 = graph.add_tensor();
-    const auto      t2 = graph.add_tensor();
-    const auto      t3 = graph.add_tensor();
-
-    graph.add_operator({ t0, t1 }, { t2 });
-    graph.add_operator({ t2 }, { t1, t3 }); // Ideally error should occur here
-
-    const auto res = graph.topological_sort();
-    ARM_COMPUTE_EXPECT(!bool(res.first), framework::LogLevel::ERRORS);
-}
-TEST_CASE(Loops_Should_Fail, framework::DatasetMode::ALL)
-{
-    DependencyGraph graph{};
-    const auto      t0 = graph.add_tensor();
-    const auto      t1 = graph.add_tensor();
-    const auto      t2 = graph.add_tensor();
-
-    ARM_COMPUTE_EXPECT_THROW(graph.add_operator({ t0, t2 }, { t1, t2 }).first, framework::LogLevel::ERRORS);
-    ARM_COMPUTE_UNUSED(t0, t1, t2);
-}
-TEST_SUITE_END() // DependencyGraph
-TEST_SUITE_END() // DYNAMIC_FUSION
-TEST_SUITE_END() // UNIT
-
-TEST_SUITE_END() // CL
-} // namespace validation
-} // namespace test
-} // namespace arm_compute
-#endif /* ENABLE_EXPERIMENTAL_DYNAMIC_FUSION */
\ No newline at end of file
diff --git a/tests/validation/CL/UNIT/dynamic_fusion/Floor.cpp b/tests/validation/CL/UNIT/dynamic_fusion/Floor.cpp
deleted file mode 100644
index 2b8f69e5e7..0000000000
--- a/tests/validation/CL/UNIT/dynamic_fusion/Floor.cpp
+++ /dev/null
@@ -1,135 +0,0 @@
-/*
- * Copyright (c) 2022 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.
- */
-
-#ifdef ENABLE_EXPERIMENTAL_DYNAMIC_FUSION
-#include "arm_compute/core/TensorInfo.h"
-
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/experimental/ClWorkload.h"
-#include "arm_compute/runtime/CL/CLScheduler.h"
-#include "arm_compute/runtime/experimental/ClCompositeOperator.h"
-#include "src/core/experimental/dynamic_fusion/WorkloadImpl/ClKernelDescriptors.h"
-#include "tests/CL/CLAccessor.h"
-#include "tests/framework/Asserts.h"
-#include "tests/framework/Macros.h"
-#include "tests/validation/CL/UNIT/dynamic_fusion/Utils.h"
-#include "tests/validation/Validation.h"
-
-#include "tests/validation/reference/Floor.h"
-#include "tests/validation/reference/Permute.h"
-
-#ifdef ARM_COMPUTE_ASSERTS_ENABLED
-#include "tests/SimpleTensorPrinter.h"
-#endif /* ARM_COMPUTE_ASSERTS_ENABLED */
-
-using namespace arm_compute::experimental::dynamic_fusion;
-using namespace arm_compute::test::validation::utils;
-
-namespace arm_compute
-{
-namespace test
-{
-namespace validation
-{
-TEST_SUITE(CL)
-TEST_SUITE(UNIT)
-TEST_SUITE(DYNAMIC_FUSION)
-TEST_CASE(Operator_Floor_1_F32, framework::DatasetMode::ALL)
-{
-    /* Computation:
-     * out = floor(input)
-     */
-    const auto data_type    = DataType::F32;
-    const auto data_layout  = DataLayout::NHWC;
-    const auto t_shape      = TensorShape(32, 16);
-    auto       t_input_info = TensorInfo(t_shape, 1, data_type, data_layout);
-    auto       t_dst_info   = TensorInfo();
-
-    FloorDescriptor floor_desc{};
-
-    // Create reference
-    SimpleTensor<float> ref_t_input{ t_shape, data_type, 1, QuantizationInfo(), DataLayout::NHWC };
-
-    // Fill reference
-    fill<float>(ref_t_input, 0, library.get());
-
-    auto ref_t_input_nchw = reference::permute(ref_t_input, PermutationVector(1U, 2U, 0U));
-    auto t_dst_shape_nchw = t_shape;
-    permute(t_dst_shape_nchw, PermutationVector(1U, 2U, 0U));
-
-    auto       ref_t_dst_nchw = reference::floor_layer(ref_t_input_nchw);
-    const auto ref_t_dst      = reference::permute(ref_t_dst_nchw, PermutationVector(2U, 0U, 1U));
-
-    CLScheduler::get().default_reinit();
-    const auto    cl_compile_ctx = CLKernelLibrary::get().get_compile_context();
-    OperatorGraph op_graph;
-
-    const auto op_t_input = add_tensor(op_graph, t_input_info);
-    const auto op_t_dst   = add_tensor(op_graph, t_dst_info);
-
-    add_op_floor(op_graph, floor_desc, op_t_input, op_t_dst);
-
-    const ClWorkloadContext workload_ctx{ GpuInfo{ CLScheduler::get().target() } };
-    ClWorkload              workload;
-    build(workload, op_graph, workload_ctx);
-
-    ClCompositeOperator op;
-    op.configure(cl_compile_ctx, workload);
-
-    // Construct tensors
-    CLTensor t_input{};
-    CLTensor t_dst{};
-
-    // Init tensors
-    t_input.allocator()->init(t_input_info);
-    t_dst.allocator()->init(t_dst_info);
-
-    // Allocate and fill tensors
-    t_input.allocator()->allocate();
-    t_dst.allocator()->allocate();
-    fill<float>(CLAccessor(t_input), 0, library.get());
-    // "Pack" tensors
-    OpTensorBinding bp_tensors({ { op_t_input, &t_input },
-        { op_t_dst, &t_dst }
-    });
-
-    // Populate prepare and run pack-maps (including allocating aux tensors)
-    ClAuxTensorData aux_tensor_data{};
-    TensorPackMap   prepare_pack_map{};
-    TensorPackMap   run_pack_map{};
-    bind_tensors(aux_tensor_data, prepare_pack_map, run_pack_map, workload, bp_tensors);
-
-    op.prepare(prepare_pack_map);
-    op.run(run_pack_map);
-    RelativeTolerance<float> tolerance_f32(0.001f); /**< Tolerance value for comparing reference's output against implementation's output for floating point data types */
-    validate(CLAccessor(t_dst), ref_t_dst_nchw, tolerance_f32);
-}
-
-TEST_SUITE_END() // DYNAMIC_FUSION
-TEST_SUITE_END() // UNIT
-TEST_SUITE_END() // CL
-} // namespace validation
-} // namespace test
-} // namespace arm_compute
-#endif /* ENABLE_EXPERIMENTAL_DYNAMIC_FUSION */
\ No newline at end of file
diff --git a/tests/validation/CL/UNIT/dynamic_fusion/Integration_OperatorFuseMovenetSubGraph1.cpp b/tests/validation/CL/UNIT/dynamic_fusion/Integration_OperatorFuseMovenetSubGraph1.cpp
deleted file mode 100644
index 3a8b7c8ce8..0000000000
--- a/tests/validation/CL/UNIT/dynamic_fusion/Integration_OperatorFuseMovenetSubGraph1.cpp
+++ /dev/null
@@ -1,402 +0,0 @@
-/*
- * Copyright (c) 2022 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.
- */
-
-#ifdef ENABLE_EXPERIMENTAL_DYNAMIC_FUSION
-#include "arm_compute/core/TensorInfo.h"
-
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/experimental/ClWorkload.h"
-#include "arm_compute/runtime/CL/CLScheduler.h"
-#include "arm_compute/runtime/experimental/ClCompositeOperator.h"
-#include "src/core/experimental/dynamic_fusion/WorkloadImpl/ClKernelDescriptors.h"
-#include "src/gpu/cl/operators/ClAdd.h"
-#include "src/gpu/cl/operators/ClConv2d.h"
-#include "tests/CL/CLAccessor.h"
-#include "tests/framework/Asserts.h"
-#include "tests/framework/Macros.h"
-#include "tests/validation/CL/UNIT/dynamic_fusion/Utils.h"
-#include "tests/validation/Validation.h"
-
-#include "tests/validation/reference/ConvolutionLayer.h"
-#include "tests/validation/reference/ElementwiseOperations.h"
-#include "tests/validation/reference/Permute.h"
-
-#ifdef ARM_COMPUTE_ASSERTS_ENABLED
-#include "tests/SimpleTensorPrinter.h"
-#endif /* ARM_COMPUTE_ASSERTS_ENABLED */
-
-using namespace arm_compute::experimental::dynamic_fusion;
-using namespace arm_compute::test::validation::utils;
-
-namespace arm_compute
-{
-namespace test
-{
-namespace validation
-{
-TEST_SUITE(CL)
-TEST_SUITE(INTEGRATION)
-TEST_SUITE(DYNAMIC_FUSION)
-TEST_CASE(Operator_Fuse_Movenet_SubGraph_1_F32, framework::DatasetMode::ALL)
-{
-    // Please refer to: https://confluence.arm.com/pages/viewpage.action?pageId=886243697
-    /* Computation:
-     * out = add_desc(addend, conv2d1x1(direct_conv)(input, weights, bias))
-     */
-    const auto data_type     = DataType::F32;
-    const auto data_layout   = DataLayout::NHWC;
-    const auto t_input_shape = TensorShape(384, 12, 12);
-    // const auto t_weight_shape   = TensorShape(384, 1, 1, 64);
-    // const auto t_dst_shape      = TensorShape(64, 12, 12);
-    const auto t_weight_shape   = TensorShape(384, 1, 1, 16);
-    const auto t_dst_shape      = TensorShape(16, 12, 12);
-    auto       t_input_info     = TensorInfo(t_input_shape, 1, data_type, data_layout);
-    auto       t_weight_info    = TensorInfo(t_weight_shape, 1, data_type, data_layout);
-    auto       t_l1_addend_info = TensorInfo(t_dst_shape, 1, data_type, data_layout);
-    auto       t_acc_info       = TensorInfo(); // Intermediate tensor for cond3
-    auto       t_dst_info       = TensorInfo();
-
-    Conv2dDescriptor      conv2d_desc{};
-    ElementwiseDescriptor add_desc{ ArithmeticOperation::ADD };
-
-    // Create reference
-    SimpleTensor<float> ref_t_input{ t_input_shape, data_type, 1, QuantizationInfo(), DataLayout::NHWC };
-    SimpleTensor<float> ref_t_weight{ t_weight_shape, data_type, 1, QuantizationInfo(), DataLayout::NHWC };
-    SimpleTensor<float> ref_t_bias_placeholder{ t_dst_shape, data_type, 1, QuantizationInfo(), DataLayout::NHWC };
-    SimpleTensor<float> ref_t_l1_addend{ t_dst_shape, data_type, 1, QuantizationInfo(), DataLayout::NHWC };
-
-    // Fill reference
-    fill<float>(ref_t_input, 0, library.get());
-    fill<float>(ref_t_weight, 1, library.get());
-    fill<float>(ref_t_l1_addend, 2, library.get());
-
-    auto ref_t_input_nchw            = reference::permute(ref_t_input, PermutationVector(1U, 2U, 0U));
-    auto ref_t_weight_nchw           = reference::permute(ref_t_weight, PermutationVector(1U, 2U, 0U));
-    auto ref_t_bias_placeholder_nchw = reference::permute(ref_t_bias_placeholder, PermutationVector(1U, 2U, 0U));
-    auto ref_t_l1_addend_nchw        = reference::permute(ref_t_l1_addend, PermutationVector(1U, 2U, 0U));
-    auto t_dst_shape_nchw            = t_dst_shape;
-    permute(t_dst_shape_nchw, PermutationVector(1U, 2U, 0U));
-
-    PadStrideInfo legacy_pad_stride(conv2d_desc.stride.x(), conv2d_desc.stride.y(), conv2d_desc.pad.left, conv2d_desc.pad.right, conv2d_desc.pad.top, conv2d_desc.pad.bottom, DimensionRoundingType{});
-    auto          ref_t_dst_nchw = reference::arithmetic_operation(
-                                       ArithmeticOperation::ADD,
-                                       ref_t_l1_addend_nchw,
-                                       reference::convolution_layer(ref_t_input_nchw, ref_t_weight_nchw, ref_t_bias_placeholder_nchw, t_dst_shape_nchw, legacy_pad_stride, conv2d_desc.dilation),
-                                       data_type,
-                                       ConvertPolicy{});
-    const auto ref_t_dst = reference::permute(ref_t_dst_nchw, PermutationVector(2U, 0U, 1U));
-
-    CLScheduler::get().default_reinit();
-    const auto    cl_compile_ctx = CLKernelLibrary::get().get_compile_context();
-    OperatorGraph op_graph;
-
-    const auto op_t_input     = add_tensor(op_graph, t_input_info);
-    const auto op_t_weight    = add_tensor(op_graph, t_weight_info);
-    const auto op_t_l1_addend = add_tensor(op_graph, t_l1_addend_info);
-    const auto op_t_acc       = add_tensor(op_graph, t_acc_info); // temp accumulator; TensorInfo to be inferred
-    const auto op_t_dst       = add_tensor(op_graph, t_dst_info);
-
-    auto conv2d = add_op_conv2d(op_graph, conv2d_desc, op_t_input, op_t_weight, op_t_acc);
-    force_conv2d_method(op_graph, conv2d, ConvolutionMethod::DIRECT);
-    add_op_elementwise_op(op_graph, add_desc, op_t_acc, op_t_l1_addend, op_t_dst);
-
-    const ClWorkloadContext workload_ctx{ GpuInfo{ CLScheduler::get().target() } };
-    ClWorkload              workload;
-    build(workload, op_graph, workload_ctx);
-
-    ClCompositeOperator op;
-    op.configure(cl_compile_ctx, workload);
-
-    // Construct tensors
-    CLTensor t_input{};
-    CLTensor t_weight{};
-    CLTensor t_l1_addend{};
-    CLTensor t_dst{};
-
-    // Init tensors
-    t_input.allocator()->init(t_input_info);
-    t_weight.allocator()->init(t_weight_info);
-    t_l1_addend.allocator()->init(t_dst_info);
-    t_dst.allocator()->init(t_dst_info);
-
-    // Allocate and fill tensors
-    t_input.allocator()->allocate();
-    t_weight.allocator()->allocate();
-    t_l1_addend.allocator()->allocate();
-    t_dst.allocator()->allocate();
-    fill<float>(CLAccessor(t_input), 0, library.get());
-    fill<float>(CLAccessor(t_weight), 1, library.get());
-    fill<float>(CLAccessor(t_l1_addend), 2, library.get());
-    // "Pack" tensors
-    OpTensorBinding bp_tensors({ { op_t_input, &t_input },
-        { op_t_weight, &t_weight },
-        { op_t_l1_addend, &t_l1_addend },
-        { op_t_dst, &t_dst }
-    });
-
-    // Populate prepare and run pack-maps (including allocating aux tensors)
-    ClAuxTensorData aux_tensor_data{};
-    TensorPackMap   prepare_pack_map{};
-    TensorPackMap   run_pack_map{};
-    bind_tensors(aux_tensor_data, prepare_pack_map, run_pack_map, workload, bp_tensors);
-
-    op.prepare(prepare_pack_map);
-    op.run(run_pack_map);
-    RelativeTolerance<float> tolerance_f32(0.001f); /**< Tolerance value for comparing reference's output against implementation's output for floating point data types */
-    validate(CLAccessor(t_dst), ref_t_dst_nchw, tolerance_f32);
-}
-TEST_SUITE(Unsupported)
-TEST_CASE(DataType_QASYMM8, framework::DatasetMode::ALL)
-{
-    const auto data_type        = DataType::QASYMM8;
-    const auto data_layout      = DataLayout::NHWC;
-    const auto t_input_shape    = TensorShape(384, 12, 12);
-    const auto t_weight_shape   = TensorShape(384, 1, 1, 64);
-    const auto t_dst_shape      = TensorShape(64, 12, 12);
-    auto       t_input_info     = TensorInfo(t_input_shape, 1, data_type, data_layout);
-    auto       t_weight_info    = TensorInfo(t_weight_shape, 1, data_type, data_layout);
-    auto       t_l1_addend_info = TensorInfo(t_dst_shape, 1, data_type, data_layout);
-    auto       t_acc_info       = TensorInfo(t_dst_shape, 1, data_type, data_layout);
-    auto       t_dst_info       = TensorInfo(t_dst_shape, 1, data_type, data_layout);
-
-    Conv2dDescriptor      conv2d_desc{};
-    ElementwiseDescriptor add_desc{};
-
-    OperatorGraph op_graph;
-
-    const auto op_t_input     = add_tensor(op_graph, t_input_info);
-    const auto op_t_weight    = add_tensor(op_graph, t_weight_info);
-    const auto op_t_l1_addend = add_tensor(op_graph, t_l1_addend_info);
-    const auto op_t_acc       = add_tensor(op_graph, t_acc_info); // temp accumulator; TensorInfo to be inferred
-    const auto op_t_dst       = add_tensor(op_graph, t_dst_info);
-
-    auto conv2d = add_op_conv2d(op_graph, conv2d_desc, op_t_input, op_t_weight, op_t_acc);
-    add_op_elementwise_op(op_graph, add_desc, op_t_acc, op_t_l1_addend, op_t_dst);
-    force_conv2d_method(op_graph, conv2d, ConvolutionMethod::DIRECT);
-
-    const ClWorkloadContext workload_ctx{ GpuInfo{ CLScheduler::get().target() } };
-    ClWorkload              workload;
-    const auto              success = build(workload, op_graph, workload_ctx);
-
-    ARM_COMPUTE_EXPECT(!bool(success), framework::LogLevel::ERRORS);
-    ARM_COMPUTE_EXPECT(!bool(ClCompositeOperator::validate(workload)), framework::LogLevel::ERRORS);
-}
-TEST_CASE(DataLayout_NCHW, framework::DatasetMode::ALL)
-{
-    const auto data_type      = DataType::F32;
-    const auto data_layout    = DataLayout::NCHW;
-    const auto t_input_shape  = TensorShape(384, 12, 12);
-    const auto t_weight_shape = TensorShape(384, 1, 1, 64);
-    const auto t_dst_shape    = TensorShape(64, 12, 12);
-    auto       t_input_info   = TensorInfo(t_input_shape, 1, data_type, data_layout);
-    auto       t_weight_info  = TensorInfo(t_weight_shape, 1, data_type, data_layout);
-    auto       t_dst_info     = TensorInfo(t_dst_shape, 1, data_type, data_layout);
-
-    Conv2dDescriptor conv2d_desc{};
-
-    OperatorGraph op_graph;
-
-    const auto op_t_input  = add_tensor(op_graph, t_input_info);
-    const auto op_t_weight = add_tensor(op_graph, t_weight_info);
-    const auto op_t_dst    = add_tensor(op_graph, t_dst_info);
-
-    auto conv2d = add_op_conv2d(op_graph, conv2d_desc, op_t_input, op_t_weight, op_t_dst);
-    force_conv2d_method(op_graph, conv2d, ConvolutionMethod::DIRECT);
-    const ClWorkloadContext workload_ctx{ GpuInfo{ CLScheduler::get().target() } };
-    ClWorkload              workload;
-    const auto              success = build(workload, op_graph, workload_ctx);
-
-    ARM_COMPUTE_EXPECT(!bool(success), framework::LogLevel::ERRORS);
-    ARM_COMPUTE_EXPECT(!bool(ClCompositeOperator::validate(workload)), framework::LogLevel::ERRORS);
-}
-TEST_SUITE_END() // Unsupported
-
-TEST_SUITE(Invalid)
-TEST_CASE(Multiple_Complex_Ops_0, framework::DatasetMode::ALL)
-{
-    /* Computation:
-     * out = conv2d(conv2d(l0_input, l0_weight), l1_weight)
-     */
-    const auto data_type         = DataType::F32;
-    const auto data_layout       = DataLayout::NHWC;
-    const auto t_l0_input_shape  = TensorShape(1024, 56, 56);
-    const auto t_l0_weight_shape = TensorShape(512, 1024, 1, 1);
-    const auto t_l1_weight_shape = TensorShape(512, 256, 1, 1);
-
-    auto t_l0_input_info  = TensorInfo(t_l0_input_shape, 1, data_type, data_layout);
-    auto t_l0_weight_info = TensorInfo(t_l0_weight_shape, 1, data_type, data_layout);
-    auto t_l1_weight_info = TensorInfo(t_l1_weight_shape, 1, data_type, data_layout);
-    auto t_l0_dst_info    = TensorInfo();
-    auto t_dst_info       = TensorInfo();
-
-    OperatorGraph op_graph;
-    const auto    conv2d_desc = Conv2dDescriptor{};
-
-    const auto op_t_l0_input  = add_tensor(op_graph, t_l0_input_info);
-    const auto op_t_l0_weight = add_tensor(op_graph, t_l0_weight_info);
-    const auto op_t_l1_weight = add_tensor(op_graph, t_l1_weight_info);
-    const auto op_t_l0_dst    = add_tensor(op_graph, t_l0_dst_info); // temp accumulator; TensorInfo to be inferred
-    const auto op_t_dst       = add_tensor(op_graph, t_dst_info);
-
-    add_op_conv2d(op_graph, conv2d_desc, op_t_l0_input, op_t_l0_weight, op_t_l0_dst);
-    add_op_conv2d(op_graph, conv2d_desc, op_t_l0_dst, op_t_l1_weight, op_t_dst);
-
-    const ClWorkloadContext workload_ctx{ GpuInfo{ CLScheduler::get().target() } };
-    ClWorkload              workload;
-    const auto              success = build(workload, op_graph, workload_ctx);
-
-    ARM_COMPUTE_EXPECT(!bool(success), framework::LogLevel::ERRORS);
-    ARM_COMPUTE_EXPECT(!bool(ClCompositeOperator::validate(workload)), framework::LogLevel::ERRORS);
-}
-TEST_CASE(Enlarging_Execution_Space, framework::DatasetMode::ALL)
-{
-    /* Computation:
-     * out = add(l2_lhs, add(add(l0_lhs, l0_rhs), l1_rhs))
-     */
-    const auto data_type      = DataType::F32;
-    const auto data_layout    = DataLayout::NHWC;
-    const auto t_l0_lhs_shape = TensorShape(1, 256, 3);
-    const auto t_l0_rhs_shape = TensorShape(1, 256, 3);
-    const auto t_l1_rhs_shape = TensorShape(1, 1, 3);
-    const auto t_l2_lhs_shape = TensorShape(1024, 1, 3);
-
-    auto t_l0_lhs_info = TensorInfo(t_l0_lhs_shape, 1, data_type, data_layout);
-    auto t_l0_rhs_info = TensorInfo(t_l0_rhs_shape, 1, data_type, data_layout);
-    auto t_l1_rhs_info = TensorInfo(t_l1_rhs_shape, 1, data_type, data_layout);
-    auto t_l2_lhs_info = TensorInfo(t_l2_lhs_shape, 1, data_type, data_layout);
-    auto t_l0_dst_info = TensorInfo();
-    auto t_l1_dst_info = TensorInfo();
-    auto t_dst_info    = TensorInfo();
-
-    OperatorGraph op_graph;
-    const auto    add_desc = ElementwiseDescriptor{};
-
-    const auto op_t_l0_lhs = add_tensor(op_graph, t_l0_lhs_info);
-    const auto op_t_l0_rhs = add_tensor(op_graph, t_l0_rhs_info);
-    const auto op_t_l1_rhs = add_tensor(op_graph, t_l1_rhs_info);
-    const auto op_t_l2_lhs = add_tensor(op_graph, t_l2_lhs_info);
-    const auto op_t_l0_dst = add_tensor(op_graph, t_l0_dst_info); // temp accumulator; TensorInfo to be inferred
-    const auto op_t_l1_dst = add_tensor(op_graph, t_l1_dst_info); // temp accumulator; TensorInfo to be inferred
-    const auto op_t_dst    = add_tensor(op_graph, t_dst_info);
-
-    add_op_elementwise_op(op_graph, add_desc, op_t_l0_lhs, op_t_l0_rhs, op_t_l0_dst);
-    add_op_elementwise_op(op_graph, add_desc, op_t_l0_dst, op_t_l1_rhs, op_t_l1_dst);
-    add_op_elementwise_op(op_graph, add_desc, op_t_l1_dst, op_t_l2_lhs, op_t_dst);
-
-    const ClWorkloadContext workload_ctx{ GpuInfo{ CLScheduler::get().target() } };
-    ClWorkload              workload;
-    const auto              success = build(workload, op_graph, workload_ctx);
-
-    ARM_COMPUTE_EXPECT(!bool(success), framework::LogLevel::ERRORS);
-    ARM_COMPUTE_EXPECT(!bool(ClCompositeOperator::validate(workload)), framework::LogLevel::ERRORS);
-}
-TEST_CASE(Root_Simple_And_Complex, framework::DatasetMode::ALL)
-{
-    /* Computation:
-     * out = add(conv(l0_0_input, l0_0_weight), add(l0_1_lhs, l0_1_rhs))
-     */
-    const auto data_type   = DataType::F32;
-    const auto data_layout = DataLayout::NHWC;
-
-    const auto t_l0_0_input_shape  = TensorShape(128, 21, 21);
-    const auto t_l0_0_weight_shape = TensorShape(144, 128, 1, 1);
-    const auto t_l0_1_lhs_shape    = TensorShape(144, 21, 21);
-    const auto t_l0_1_rhs_shape    = TensorShape(1, 1, 21);
-
-    auto t_l0_0_input_info  = TensorInfo(t_l0_0_input_shape, 1, data_type, data_layout);
-    auto t_l0_0_weight_info = TensorInfo(t_l0_0_weight_shape, 1, data_type, data_layout);
-    auto t_l0_1_lhs_info    = TensorInfo(t_l0_1_lhs_shape, 1, data_type, data_layout);
-    auto t_l0_1_rhs_info    = TensorInfo(t_l0_1_rhs_shape, 1, data_type, data_layout);
-    auto t_l0_0_dst_info    = TensorInfo();
-    auto t_l0_1_dst_info    = TensorInfo();
-    auto t_dst_info         = TensorInfo();
-
-    OperatorGraph op_graph;
-    const auto    conv2d_desc = Conv2dDescriptor{};
-    const auto    add_desc    = ElementwiseDescriptor{};
-
-    const auto op_t_l0_0_input  = add_tensor(op_graph, t_l0_0_input_info);
-    const auto op_t_l0_0_weight = add_tensor(op_graph, t_l0_0_weight_info);
-    const auto op_t_l0_1_lhs    = add_tensor(op_graph, t_l0_1_lhs_info);
-    const auto op_t_l0_1_rhs    = add_tensor(op_graph, t_l0_1_rhs_info);
-    const auto op_t_l0_0_dst    = add_tensor(op_graph, t_l0_0_dst_info); // temp accumulator; TensorInfo to be inferred
-    const auto op_t_l0_1_dst    = add_tensor(op_graph, t_l0_1_dst_info); // temp accumulator; TensorInfo to be inferred
-    const auto op_t_dst         = add_tensor(op_graph, t_dst_info);
-
-    add_op_conv2d(op_graph, conv2d_desc, op_t_l0_0_input, op_t_l0_0_weight, op_t_l0_0_dst);
-    add_op_elementwise_op(op_graph, add_desc, op_t_l0_1_lhs, op_t_l0_1_rhs, op_t_l0_1_dst);
-    add_op_elementwise_op(op_graph, add_desc, op_t_l0_0_dst, op_t_l0_1_dst, op_t_dst);
-
-    const ClWorkloadContext workload_ctx{ GpuInfo{ CLScheduler::get().target() } };
-    ClWorkload              workload;
-    const auto              success = build(workload, op_graph, workload_ctx);
-
-    ARM_COMPUTE_EXPECT(!bool(success), framework::LogLevel::ERRORS);
-    ARM_COMPUTE_EXPECT(!bool(ClCompositeOperator::validate(workload)), framework::LogLevel::ERRORS);
-}
-TEST_CASE(Loop, framework::DatasetMode::ALL)
-{
-    /* Computation:
-     * tensor state0;
-     * state1 = add(l0_lhs, state0)
-     * state0 = add(l1_lhs, state1)
-     */
-    const auto data_type   = DataType::F32;
-    const auto data_layout = DataLayout::NHWC;
-
-    const auto t_shape = TensorShape(13, 21);
-
-    auto t_l0_lhs_info = TensorInfo(t_shape, 1, data_type, data_layout);
-    auto t_l1_lhs_info = TensorInfo(t_shape, 1, data_type, data_layout);
-    auto state0_info   = TensorInfo(t_shape, 1, data_type, data_layout);
-    auto state1_info   = TensorInfo();
-
-    OperatorGraph op_graph;
-    const auto    conv2d_desc = Conv2dDescriptor{};
-    const auto    add_desc    = ElementwiseDescriptor{};
-
-    const auto op_t_l0_lhs = add_tensor(op_graph, t_l0_lhs_info);
-    const auto op_t_l1_lhs = add_tensor(op_graph, t_l1_lhs_info);
-    const auto op_t_state0 = add_tensor(op_graph, state0_info);
-    const auto op_t_state1 = add_tensor(op_graph, state1_info);
-
-    add_op_conv2d(op_graph, conv2d_desc, op_t_l0_lhs, op_t_state0, op_t_state1);
-    add_op_elementwise_op(op_graph, add_desc, op_t_l1_lhs, op_t_state1, op_t_state0);
-
-    const ClWorkloadContext workload_ctx{ GpuInfo{ CLScheduler::get().target() } };
-    ClWorkload              workload;
-    const auto              success = build(workload, op_graph, workload_ctx);
-
-    ARM_COMPUTE_EXPECT(!bool(success), framework::LogLevel::ERRORS);
-    ARM_COMPUTE_EXPECT(!bool(ClCompositeOperator::validate(workload)), framework::LogLevel::ERRORS);
-}
-TEST_SUITE_END() // Invalid
-
-TEST_SUITE_END() // DYNAMIC_FUSION
-TEST_SUITE_END() // INTEGRATION
-TEST_SUITE_END() // CL
-} // namespace validation
-} // namespace test
-} // namespace arm_compute
-#endif /* ENABLE_EXPERIMENTAL_DYNAMIC_FUSION */
\ No newline at end of file
diff --git a/tests/validation/CL/UNIT/dynamic_fusion/Utils.h b/tests/validation/CL/UNIT/dynamic_fusion/Utils.h
deleted file mode 100644
index 4512305c1e..0000000000
--- a/tests/validation/CL/UNIT/dynamic_fusion/Utils.h
+++ /dev/null
@@ -1,71 +0,0 @@
-/*
- * Copyright (c) 2022 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 TESTS_VALIDATION_CL_DYNAMICFUSION_UTILS
-#define TESTS_VALIDATION_CL_DYNAMICFUSION_UTILS
-
-#include "tests/AssetsLibrary.h"
-#include "utils/Utils.h"
-
-#include <chrono>
-#include <limits>
-#include <type_traits>
-
-namespace arm_compute
-{
-namespace test
-{
-namespace validation
-{
-namespace utils
-{
-/** A pair of macros which measures the wall clock time, and records it into a map measurement_map with name clock_name
- *
- */
-#define TICK(clock_name) \
-    auto clock_name##_tick = std::chrono::high_resolution_clock::now();
-#define TOCK(clock_name, measurement_map)                                               \
-    auto clock_name##_tock                 = std::chrono::high_resolution_clock::now(); \
-    measurement_map["\"" #clock_name "\""] = duration_cast<microseconds>(clock_name##_tock - clock_name##_tick);
-#define TOCK_AVG(clock_name, measurement_map, num_iterations)                           \
-    auto clock_name##_tock                 = std::chrono::high_resolution_clock::now(); \
-    measurement_map["\"" #clock_name "\""] = duration_cast<microseconds>((clock_name##_tock - clock_name##_tick) / (num_iterations));
-
-template <typename T, typename U>
-void fill(U &&tensor, int seed, AssetsLibrary *library)
-{
-    static_assert(std::is_floating_point<T>::value || std::is_same<T, half>::value, "Only floating point data types supported.");
-    using DistributionType = typename std::conditional<std::is_same<T, half>::value, arm_compute::utils::uniform_real_distribution_16bit<T>, std::uniform_real_distribution<T>>::type;
-
-    DistributionType distribution{ T(-1.0f), T(1.0f) };
-    library->fill(tensor, distribution, seed);
-
-    // Fill border with infinity in order to check the presence of NaN values (i.e. inf * 0)
-    DistributionType distribution_inf{ T(std::numeric_limits<float>::infinity()), T(std::numeric_limits<float>::infinity()) };
-    library->fill_borders_with_garbage(tensor, distribution_inf, seed);
-}
-} // namespace utils
-} // namespace validation
-} // namespace test
-} // namespace arm_compute
-#endif //TESTS_VALIDATION_CL_DYNAMICFUSION_UTILS
\ No newline at end of file