Port CLGEMM to memory injecting interface

Moves the following kernels:
 - CLGEMMMatrixMultiplyKernel
 - CLGEMMMatrixMultiplyNativeKernel
 - CLGEMMMatrixMultipluReshapedKernel
 - CLGEMMMatrixMultiplyReshapedOnlyRHSKernel

 Moves the following functions
 - CLGEMM

Introduces facilities to easy handling of auxiliary temporary buffers
under then new run interface. Such are:
 - CLAuxTensorHandler: That allows wrapping of workspace buffers memory
 to CLBuffer objects
 - Ability to inject TensorInfo to allocator without transferring
 ownership. This reduce the copy overhead if needed.

Resolves: COMPMID-4188

Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com>
Change-Id: I7055435d831b05b749b26302082e4ac45f26dfb0
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5498
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Michalis Spyrou <michalis.spyrou@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>

16 files changed, 330 insertions, 183 deletions

diff --git a/tests/CL/Helper.h b/tests/CL/Helper.h
index 5153e98add..b99911e1e6 100644
--- a/tests/CL/Helper.h
+++ b/tests/CL/Helper.h
@@ -29,8 +29,11 @@
 #include "arm_compute/runtime/CL/functions/CLFill.h"
 #include "arm_compute/runtime/IFunction.h"
 #include "src/core/CL/kernels/CLFillBorderKernel.h"
+#include "src/runtime/gpu/cl/IClOperator.h"
+#include "src/runtime/gpu/cl/operators/ClFill.h"
 
 #include "src/core/CL/ICLKernel.h"
+#include "support/Cast.h"
 
 #include <memory>
 
@@ -38,6 +41,86 @@ namespace arm_compute
 {
 namespace test
 {
+/** This template synthetizes a simple IOperator which runs the given kernel K */
+template <typename K>
+class CLSynthetizeOperator : public opencl::IClOperator
+{
+public:
+    /** Configure the kernel.
+     *
+     * @param[in] args Configuration arguments.
+     */
+    template <typename... Args>
+    void configure(Args &&... args)
+    {
+        auto k = std::make_unique<K>();
+        k->configure(CLKernelLibrary::get().get_compile_context(), std::forward<Args>(args)...);
+        _kernel = std::move(k);
+    }
+    /** Configure the kernel setting the GPU target as well
+     *
+     * @param[in] gpu_target GPUTarget to set
+     * @param[in] args       Configuration arguments.
+     */
+    template <typename... Args>
+    void configure(GPUTarget gpu_target, Args &&... args)
+    {
+        auto k = std::make_unique<K>();
+        k->set_target(gpu_target);
+        k->configure(CLKernelLibrary::get().get_compile_context(), std::forward<Args>(args)...);
+        _kernel = std::move(k);
+    }
+    /** Validate input arguments
+     *
+     * @param[in] args Configuration arguments.
+     */
+    template <typename... Args>
+    static Status validate(Args &&... args)
+    {
+        return K::validate(std::forward<Args>(args)...);
+    }
+};
+
+/** As above but this also initializes to zero the input tensor */
+template <typename K, int bordersize>
+class CLSynthetizeOperatorInitOutputWithZeroAndWithZeroConstantBorder : public opencl::IClOperator
+{
+public:
+    /** Configure the kernel.
+     *
+     * @param[in] first  First input argument.
+     * @param[in] second Second input argument.
+     * @param[in] args   Rest of the configuration arguments.
+     */
+    template <typename T, typename... Args>
+    void configure(T first, T second, Args &&... args)
+    {
+        auto cctx = CLKernelLibrary::get().get_compile_context();
+        auto k    = std::make_unique<K>();
+        k->set_target(CLScheduler::get().target());
+        k->configure(cctx, first, second, std::forward<Args>(args)...);
+        _kernel = std::move(k);
+        _border_handler.configure(cctx, first, BorderSize(bordersize), BorderMode::CONSTANT, PixelValue());
+        _fill.configure(cctx, second, PixelValue());
+    }
+
+    // Inherited method overridden:
+    void run(ITensorPack &tensors) override final
+    {
+        ARM_COMPUTE_ERROR_ON_MSG(!_kernel, "The CL kernel or function isn't configured");
+
+        ITensorPack fill_pack = { { ACL_SRC, tensors.get_tensor(TensorType::ACL_DST) } };
+        _fill.run(fill_pack);
+        CLScheduler::get().enqueue_op(_border_handler, tensors);
+        CLScheduler::get().enqueue_op(*_kernel, tensors);
+    }
+
+private:
+    opencl::ClFill             _fill{};           /**< Kernel to initialize the tensor */
+    CLFillBorderKernel         _border_handler{}; /**< Kernel to handle  borders */
+    std::unique_ptr<ICLKernel> _kernel{};         /**< Kernel to run */
+};
+
 /** This template synthetizes an ICLSimpleFunction which runs the given kernel K */
 template <typename K>
 class CLSynthetizeFunction : public ICLSimpleFunction
diff --git a/tests/validation/CL/GEMMLowpMatrixMultiplyReshaped.cpp b/tests/validation/CL/GEMMLowpMatrixMultiplyReshaped.cpp
index 4873a291ab..68a7d055ad 100644
--- a/tests/validation/CL/GEMMLowpMatrixMultiplyReshaped.cpp
+++ b/tests/validation/CL/GEMMLowpMatrixMultiplyReshaped.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2019-2020 Arm Limited.
+ * Copyright (c) 2019-2021 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -24,8 +24,8 @@
 #include "arm_compute/core/Types.h"
 #include "arm_compute/core/utils/misc/ShapeCalculator.h"
 #include "src/core/CL/kernels/CLGEMMLowpMatrixMultiplyReshapedKernel.h"
-#include "src/core/CL/kernels/CLGEMMReshapeLHSMatrixKernel.h"
-#include "src/core/CL/kernels/CLGEMMReshapeRHSMatrixKernel.h"
+#include "src/core/gpu/cl/kernels/ClGemmReshapeLhsMatrixKernel.h"
+#include "src/core/gpu/cl/kernels/ClGemmReshapeRhsMatrixKernel.h"
 #include "tests/CL/CLAccessor.h"
 #include "tests/CL/Helper.h"
 #include "tests/framework/Asserts.h"
@@ -42,13 +42,13 @@ namespace validation
 {
 using namespace arm_compute::misc::shape_calculator;
 
-// Create function for CLGEMMReshapeLHSMatrixKernel
-using CLGEMMReshapeLHSMatrix = CLSynthetizeFunction<CLGEMMReshapeLHSMatrixKernel>;
+// Create function for ClGemmReshapeLhsMatrixKernel
+using CLGEMMReshapeLHSMatrix = CLSynthetizeOperator<opencl::kernels::ClGemmReshapeLhsMatrixKernel>;
 
-// Create function for CLGEMMReshapeRHSMatrixKernel
-using CLGEMMReshapeRHSMatrix = CLSynthetizeFunction<CLGEMMReshapeRHSMatrixKernel>;
+// Create function for ClGemmReshapeRhsMatrixKernel
+using CLGEMMReshapeRHSMatrix = CLSynthetizeOperator<opencl::kernels::ClGemmReshapeRhsMatrixKernel>;
 
-// Create function for CLGEMMMatrixMultiplyReshapedKernel
+// Create function for CLGEMMLowpMatrixMultiplyReshapedKernel
 using CLGEMMLowpMatrixMultiplyReshaped = CLSynthetizeFunction<CLGEMMLowpMatrixMultiplyReshapedKernel>;
 
 // Fixture for CLGEMMLowpMatrixMultiplyReshaped
diff --git a/tests/validation/CL/GEMMLowpMatrixMultiplyReshapedOnlyRHS.cpp b/tests/validation/CL/GEMMLowpMatrixMultiplyReshapedOnlyRHS.cpp
index fa256280ca..43b86b51e8 100644
--- a/tests/validation/CL/GEMMLowpMatrixMultiplyReshapedOnlyRHS.cpp
+++ b/tests/validation/CL/GEMMLowpMatrixMultiplyReshapedOnlyRHS.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2019-2020 Arm Limited.
+ * Copyright (c) 2019-2021 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -26,7 +26,7 @@
 #include "arm_compute/runtime/CL/CLTensor.h"
 #include "arm_compute/runtime/CL/CLTensorAllocator.h"
 #include "src/core/CL/kernels/CLGEMMLowpMatrixMultiplyReshapedOnlyRHSKernel.h"
-#include "src/core/CL/kernels/CLGEMMReshapeRHSMatrixKernel.h"
+#include "src/core/gpu/cl/kernels/ClGemmReshapeRhsMatrixKernel.h"
 #include "tests/CL/CLAccessor.h"
 #include "tests/CL/Helper.h"
 #include "tests/PaddingCalculator.h"
@@ -46,7 +46,7 @@ namespace validation
 using namespace arm_compute::misc::shape_calculator;
 
 // Create function for CLGEMMReshapeRHSMatrixKernel
-using CLGEMMReshapeRHSMatrix = CLSynthetizeFunction<CLGEMMReshapeRHSMatrixKernel>;
+using CLGEMMReshapeRHSMatrix = CLSynthetizeOperator<opencl::kernels::ClGemmReshapeRhsMatrixKernel>;
 
 // Create function for CLGEMMLowpMatrixMultiplyReshapedOnlyRHSKernel
 using CLGEMMLowpMatrixMultiplyReshapedOnlyRHS = CLSynthetizeFunction<CLGEMMLowpMatrixMultiplyReshapedOnlyRHSKernel>;
diff --git a/tests/validation/CL/GEMMMatrixMultiply.cpp b/tests/validation/CL/GEMMMatrixMultiply.cpp
index fdf7f503ec..21e085087d 100644
--- a/tests/validation/CL/GEMMMatrixMultiply.cpp
+++ b/tests/validation/CL/GEMMMatrixMultiply.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2019-2020 Arm Limited.
+ * Copyright (c) 2019-2021 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -26,7 +26,7 @@
 #include "arm_compute/core/utils/misc/ShapeCalculator.h"
 #include "arm_compute/runtime/CL/CLTensor.h"
 #include "arm_compute/runtime/CL/CLTensorAllocator.h"
-#include "src/core/CL/kernels/CLGEMMMatrixMultiplyKernel.h"
+#include "src/core/gpu/cl/kernels/ClGemmMatrixMultiplyKernel.h"
 #include "tests/CL/CLAccessor.h"
 #include "tests/CL/Helper.h"
 #include "tests/PaddingCalculator.h"
@@ -44,9 +44,10 @@ namespace test
 namespace validation
 {
 using namespace arm_compute::misc::shape_calculator;
+using namespace arm_compute::opencl::kernels;
 
 // Create function for CLGEMMMatrixMultiplyKernel
-using CLGEMMMatrixMultiplyNative = CLSynthetizeFunction<CLGEMMMatrixMultiplyKernel>;
+using CLGEMMMatrixMultiplyNative = CLSynthetizeOperator<ClGemmMatrixMultiplyKernel>;
 
 // Fixture for GEMMMatrixMultiplyValidationFixture
 template <typename T>
@@ -140,7 +141,7 @@ TEST_CASE(Negative, framework::DatasetMode::ALL)
         const bool is_interleaved_transposed = false;
         const GEMMReshapeInfo reshape_info = GEMMReshapeInfo(12, 14, 13, 1, 1, 0, false, false);
         const GPUTarget gpu_target           = GPUTarget::MIDGARD;
-        const auto status    = CLGEMMMatrixMultiplyKernel::validate(&lhs, &rhs, nullptr, &out, alpha, beta, is_interleaved_transposed, reshape_info, gpu_target);
+        const auto status    = ClGemmMatrixMultiplyKernel::validate(&lhs, &rhs, nullptr, &out, alpha, beta, is_interleaved_transposed, reshape_info, gpu_target);
         ARM_COMPUTE_EXPECT(bool(status) == false, framework::LogLevel::ERRORS);
     }
 
@@ -154,7 +155,7 @@ TEST_CASE(Negative, framework::DatasetMode::ALL)
         const bool is_interleaved_transposed = false;
         const GEMMReshapeInfo reshape_info = GEMMReshapeInfo(12, 14, 13, 1, 1, 0, false, false);
         const GPUTarget gpu_target           = GPUTarget::MIDGARD;
-        const auto status    = CLGEMMMatrixMultiplyKernel::validate(&lhs, &rhs, nullptr, &out, alpha, beta, is_interleaved_transposed, reshape_info, gpu_target);
+        const auto status    = ClGemmMatrixMultiplyKernel::validate(&lhs, &rhs, nullptr, &out, alpha, beta, is_interleaved_transposed, reshape_info, gpu_target);
         ARM_COMPUTE_EXPECT(bool(status) == false, framework::LogLevel::ERRORS);
     }
 
@@ -169,7 +170,7 @@ TEST_CASE(Negative, framework::DatasetMode::ALL)
         const GEMMReshapeInfo reshape_info  = GEMMReshapeInfo(12, 14, 13, 1, 1, 0, false, false);
         const GPUTarget gpu_target           = GPUTarget::MIDGARD;
         const bool fp_mixed_precision        = true;
-        const auto status    = CLGEMMMatrixMultiplyKernel::validate(&lhs, &rhs, nullptr, &out, alpha, beta, is_interleaved_transposed, reshape_info, gpu_target, fp_mixed_precision);
+        const auto status    = ClGemmMatrixMultiplyKernel::validate(&lhs, &rhs, nullptr, &out, alpha, beta, is_interleaved_transposed, reshape_info, gpu_target, fp_mixed_precision);
         ARM_COMPUTE_EXPECT(bool(status) == false, framework::LogLevel::ERRORS);
     }
 
@@ -183,7 +184,7 @@ TEST_CASE(Negative, framework::DatasetMode::ALL)
         const bool is_interleaved_transposed = false;
         const GEMMReshapeInfo reshape_info = GEMMReshapeInfo(12, 14, 13, 1, 1, 0, false, false);
         const GPUTarget gpu_target           = GPUTarget::MIDGARD;
-        const auto status    = CLGEMMMatrixMultiplyKernel::validate(&lhs, &rhs, nullptr, &out, alpha, beta, is_interleaved_transposed, reshape_info, gpu_target);
+        const auto status    = ClGemmMatrixMultiplyKernel::validate(&lhs, &rhs, nullptr, &out, alpha, beta, is_interleaved_transposed, reshape_info, gpu_target);
         ARM_COMPUTE_EXPECT(bool(status) == false, framework::LogLevel::ERRORS);
     }
 
@@ -197,7 +198,7 @@ TEST_CASE(Negative, framework::DatasetMode::ALL)
         const bool is_interleaved_transposed = false;
         const GEMMReshapeInfo reshape_info = GEMMReshapeInfo(12, 14, 13, 1, 1, 0, false, false);
         const GPUTarget gpu_target           = GPUTarget::MIDGARD;
-        const auto status    = CLGEMMMatrixMultiplyKernel::validate(&lhs, &rhs, nullptr, &out, alpha, beta, is_interleaved_transposed, reshape_info, gpu_target);
+        const auto status    = ClGemmMatrixMultiplyKernel::validate(&lhs, &rhs, nullptr, &out, alpha, beta, is_interleaved_transposed, reshape_info, gpu_target);
         ARM_COMPUTE_EXPECT(bool(status) == false, framework::LogLevel::ERRORS);
     }
 
@@ -214,7 +215,7 @@ TEST_CASE(Negative, framework::DatasetMode::ALL)
         const GEMMReshapeInfo reshape_info = GEMMReshapeInfo(12, 14, 13, 1, 1, 0, false, true);
         const GPUTarget gpu_target           = GPUTarget::MIDGARD;
         const bool fp_mixed_precision        = false;
-        const auto status    = CLGEMMMatrixMultiplyKernel::validate(&lhs, &rhs, &bias, &out, alpha, beta, is_interleaved_transposed, reshape_info, gpu_target, fp_mixed_precision);
+        const auto status    = ClGemmMatrixMultiplyKernel::validate(&lhs, &rhs, &bias, &out, alpha, beta, is_interleaved_transposed, reshape_info, gpu_target, fp_mixed_precision);
         ARM_COMPUTE_EXPECT(bool(status) == false, framework::LogLevel::ERRORS);
     }
 
@@ -231,7 +232,7 @@ TEST_CASE(Negative, framework::DatasetMode::ALL)
         const GEMMReshapeInfo reshape_info = GEMMReshapeInfo(12, 14, 13, 1, 1, 0, false, false);
         const GPUTarget gpu_target           = GPUTarget::MIDGARD;
         const bool fp_mixed_precision        = false;
-        const auto status    = CLGEMMMatrixMultiplyKernel::validate(&lhs, &rhs, &bias, &out, alpha, beta, is_interleaved_transposed, reshape_info, gpu_target, fp_mixed_precision);
+        const auto status    = ClGemmMatrixMultiplyKernel::validate(&lhs, &rhs, &bias, &out, alpha, beta, is_interleaved_transposed, reshape_info, gpu_target, fp_mixed_precision);
         ARM_COMPUTE_EXPECT(bool(status) == false, framework::LogLevel::ERRORS);
     }
 
@@ -246,7 +247,7 @@ TEST_CASE(Negative, framework::DatasetMode::ALL)
         const bool is_interleaved_transposed = false;
         const GEMMReshapeInfo reshape_info = GEMMReshapeInfo(12, 14, 13, 1, 1, 0, false, false);
         const GPUTarget gpu_target           = GPUTarget::MIDGARD;
-        const auto status    = CLGEMMMatrixMultiplyKernel::validate(&lhs, &rhs, nullptr, &out, alpha, beta, is_interleaved_transposed, reshape_info, gpu_target);
+        const auto status    = ClGemmMatrixMultiplyKernel::validate(&lhs, &rhs, nullptr, &out, alpha, beta, is_interleaved_transposed, reshape_info, gpu_target);
         ARM_COMPUTE_EXPECT(bool(status) == false, framework::LogLevel::ERRORS);
     }
 }
diff --git a/tests/validation/CL/GEMMMatrixMultiplyInterleavedTransposed.cpp b/tests/validation/CL/GEMMMatrixMultiplyInterleavedTransposed.cpp
index d6507a06c4..e47518ad7d 100644
--- a/tests/validation/CL/GEMMMatrixMultiplyInterleavedTransposed.cpp
+++ b/tests/validation/CL/GEMMMatrixMultiplyInterleavedTransposed.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2019-2020 Arm Limited.
+ * Copyright (c) 2019-2021 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -26,9 +26,9 @@
 #include "arm_compute/core/utils/misc/ShapeCalculator.h"
 #include "arm_compute/runtime/CL/CLTensor.h"
 #include "arm_compute/runtime/CL/CLTensorAllocator.h"
-#include "src/core/CL/kernels/CLGEMMMatrixMultiplyKernel.h"
-#include "src/core/CL/kernels/CLGEMMReshapeLHSMatrixKernel.h"
-#include "src/core/CL/kernels/CLGEMMReshapeRHSMatrixKernel.h"
+#include "src/core/gpu/cl/kernels/ClGemmMatrixMultiplyKernel.h"
+#include "src/core/gpu/cl/kernels/ClGemmReshapeLhsMatrixKernel.h"
+#include "src/core/gpu/cl/kernels/ClGemmReshapeRhsMatrixKernel.h"
 #include "tests/CL/CLAccessor.h"
 #include "tests/CL/Helper.h"
 #include "tests/PaddingCalculator.h"
@@ -46,15 +46,16 @@ namespace test
 namespace validation
 {
 using namespace arm_compute::misc::shape_calculator;
+using namespace arm_compute::opencl::kernels;
 
-// Create function for CLGEMMReshapeLHSMatrixKernel
-using CLGEMMReshapeLHSMatrix = CLSynthetizeFunction<CLGEMMReshapeLHSMatrixKernel>;
+// Create function for ClGemmReshapeLhsMatrixKernel
+using CLGEMMReshapeLHSMatrix = CLSynthetizeOperator<ClGemmReshapeLhsMatrixKernel>;
 
-// Create function for CLGEMMReshapeRHSMatrixKernel
-using CLGEMMReshapeRHSMatrix = CLSynthetizeFunction<CLGEMMReshapeRHSMatrixKernel>;
+// Create function for ClGemmReshapeRhsMatrixKernel
+using CLGEMMReshapeRHSMatrix = CLSynthetizeOperator<ClGemmReshapeRhsMatrixKernel>;
 
-// Create function for CLGEMMMatrixMultiplyKernel
-using CLGEMMMatrixMultiplyReshaped = CLSynthetizeFunction<CLGEMMMatrixMultiplyKernel>;
+// Create function for ClGemmMatrixMultiplyKernel
+using CLGEMMMatrixMultiplyReshaped = CLSynthetizeOperator<ClGemmMatrixMultiplyKernel>;
 
 // Fixture for GEMMMatrixMultiplyInterleavedTransposedValidationFixture
 template <typename T>
@@ -166,7 +167,7 @@ TEST_CASE(Negative, framework::DatasetMode::ALL)
         const GEMMReshapeInfo reshape_info = GEMMReshapeInfo(16, 24, 13, 2, 4, 0, false, false);
         const GPUTarget gpu_target           = GPUTarget::MIDGARD;
         const bool fp_mixed_precision        = false;
-        const auto status    = CLGEMMMatrixMultiplyKernel::validate(&lhs, &rhs, &bias, &out, alpha, beta, is_interleaved_transposed, reshape_info, gpu_target, fp_mixed_precision);
+        const auto status    = ClGemmMatrixMultiplyKernel::validate(&lhs, &rhs, &bias, &out, alpha, beta, is_interleaved_transposed, reshape_info, gpu_target, fp_mixed_precision);
         ARM_COMPUTE_EXPECT(bool(status) == false, framework::LogLevel::ERRORS);
     }
 
@@ -183,7 +184,7 @@ TEST_CASE(Negative, framework::DatasetMode::ALL)
         const GEMMReshapeInfo reshape_info = GEMMReshapeInfo(16, 24, 13, 2, 4, 0, false, false);
         const GPUTarget gpu_target           = GPUTarget::MIDGARD;
         const bool fp_mixed_precision        = false;
-        const auto status    = CLGEMMMatrixMultiplyKernel::validate(&lhs, &rhs, &bias, &out, alpha, beta, is_interleaved_transposed, reshape_info, gpu_target, fp_mixed_precision);
+        const auto status    = ClGemmMatrixMultiplyKernel::validate(&lhs, &rhs, &bias, &out, alpha, beta, is_interleaved_transposed, reshape_info, gpu_target, fp_mixed_precision);
         ARM_COMPUTE_EXPECT(bool(status) == false, framework::LogLevel::ERRORS);
     }
 
@@ -200,7 +201,7 @@ TEST_CASE(Negative, framework::DatasetMode::ALL)
         const GEMMReshapeInfo reshape_info = GEMMReshapeInfo(16, 24, 13, 2, 4, 0, false, true);
         const GPUTarget gpu_target           = GPUTarget::MIDGARD;
         const bool fp_mixed_precision        = false;
-        const auto status    = CLGEMMMatrixMultiplyKernel::validate(&lhs, &rhs, &bias, &out, alpha, beta, is_interleaved_transposed, reshape_info, gpu_target, fp_mixed_precision);
+        const auto status    = ClGemmMatrixMultiplyKernel::validate(&lhs, &rhs, &bias, &out, alpha, beta, is_interleaved_transposed, reshape_info, gpu_target, fp_mixed_precision);
         ARM_COMPUTE_EXPECT(bool(status) == false, framework::LogLevel::ERRORS);
     }
 
@@ -217,7 +218,7 @@ TEST_CASE(Negative, framework::DatasetMode::ALL)
         const GEMMReshapeInfo reshape_info = GEMMReshapeInfo(16, 24, 13, 2, 4, 0, false, false);
         const GPUTarget gpu_target           = GPUTarget::MIDGARD;
         const bool fp_mixed_precision        = false;
-        const auto status    = CLGEMMMatrixMultiplyKernel::validate(&lhs, &rhs, &bias, &out, alpha, beta, is_interleaved_transposed, reshape_info, gpu_target, fp_mixed_precision);
+        const auto status    = ClGemmMatrixMultiplyKernel::validate(&lhs, &rhs, &bias, &out, alpha, beta, is_interleaved_transposed, reshape_info, gpu_target, fp_mixed_precision);
         ARM_COMPUTE_EXPECT(bool(status) == false, framework::LogLevel::ERRORS);
     }
 
@@ -234,7 +235,7 @@ TEST_CASE(Negative, framework::DatasetMode::ALL)
         const GEMMReshapeInfo reshape_info = GEMMReshapeInfo(16, 24, 13, 2, 4, 0, false, false);
         const GPUTarget gpu_target           = GPUTarget::MIDGARD;
         const bool fp_mixed_precision        = false;
-        const auto status    = CLGEMMMatrixMultiplyKernel::validate(&lhs, &rhs, &bias, &out, alpha, beta, is_interleaved_transposed, reshape_info, gpu_target, fp_mixed_precision);
+        const auto status    = ClGemmMatrixMultiplyKernel::validate(&lhs, &rhs, &bias, &out, alpha, beta, is_interleaved_transposed, reshape_info, gpu_target, fp_mixed_precision);
         ARM_COMPUTE_EXPECT(bool(status) == false, framework::LogLevel::ERRORS);
     }
 }
diff --git a/tests/validation/CL/GEMMMatrixMultiplyNative.cpp b/tests/validation/CL/GEMMMatrixMultiplyNative.cpp
index ec6b87fbae..a737c687c4 100644
--- a/tests/validation/CL/GEMMMatrixMultiplyNative.cpp
+++ b/tests/validation/CL/GEMMMatrixMultiplyNative.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2019-2020 Arm Limited.
+ * Copyright (c) 2019-2021 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -26,7 +26,7 @@
 #include "arm_compute/core/utils/misc/ShapeCalculator.h"
 #include "arm_compute/runtime/CL/CLTensor.h"
 #include "arm_compute/runtime/CL/CLTensorAllocator.h"
-#include "src/core/CL/kernels/CLGEMMMatrixMultiplyNativeKernel.h"
+#include "src/core/gpu/cl/kernels/ClGemmMatrixMultiplyNativeKernel.h"
 #include "tests/CL/CLAccessor.h"
 #include "tests/CL/Helper.h"
 #include "tests/PaddingCalculator.h"
@@ -44,9 +44,10 @@ namespace test
 namespace validation
 {
 using namespace arm_compute::misc::shape_calculator;
+using namespace arm_compute::opencl::kernels;
 
-// Create function for CLGEMMMatrixMultiplyNativeKernel
-using CLGEMMMatrixMultiplyNative = CLSynthetizeFunction<CLGEMMMatrixMultiplyNativeKernel>;
+// Create function for ClGemmMatrixMultiplyNativeKernel
+using CLGEMMMatrixMultiplyNative = CLSynthetizeOperator<ClGemmMatrixMultiplyNativeKernel>;
 
 // Fixture for CLGEMMMatrixMultiplyNative
 template <typename T>
@@ -184,7 +185,7 @@ void validate_configuration(unsigned int m_value, unsigned int n_value, unsigned
 
     // Create and configure function
     CLGEMMMatrixMultiplyNative gemm;
-    gemm.configure(&lhs, &rhs, &bias, &dst, 1.0f, 1.0f, lhs_info, rhs_info, kernel_info);
+    gemm.configure(lhs.info(), rhs.info(), bias.info(), dst.info(), 1.0f, 1.0f, lhs_info, rhs_info, kernel_info);
 }
 } // namespace
 
diff --git a/tests/validation/CL/GEMMMatrixMultiplyReshaped.cpp b/tests/validation/CL/GEMMMatrixMultiplyReshaped.cpp
index 52afb716e4..6f368a9650 100644
--- a/tests/validation/CL/GEMMMatrixMultiplyReshaped.cpp
+++ b/tests/validation/CL/GEMMMatrixMultiplyReshaped.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2018-2020 Arm Limited.
+ * Copyright (c) 2018-2021 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -26,9 +26,9 @@
 #include "arm_compute/core/utils/misc/ShapeCalculator.h"
 #include "arm_compute/runtime/CL/CLTensor.h"
 #include "arm_compute/runtime/CL/CLTensorAllocator.h"
-#include "src/core/CL/kernels/CLGEMMMatrixMultiplyReshapedKernel.h"
-#include "src/core/CL/kernels/CLGEMMReshapeLHSMatrixKernel.h"
-#include "src/core/CL/kernels/CLGEMMReshapeRHSMatrixKernel.h"
+#include "src/core/gpu/cl/kernels/ClGemmMatrixMultiplyReshapedKernel.h"
+#include "src/core/gpu/cl/kernels/ClGemmReshapeLhsMatrixKernel.h"
+#include "src/core/gpu/cl/kernels/ClGemmReshapeRhsMatrixKernel.h"
 #include "tests/CL/CLAccessor.h"
 #include "tests/CL/Helper.h"
 #include "tests/PaddingCalculator.h"
@@ -46,15 +46,16 @@ namespace test
 namespace validation
 {
 using namespace arm_compute::misc::shape_calculator;
+using namespace arm_compute::opencl::kernels;
 
-// Create function for CLGEMMReshapeLHSMatrixKernel
-using CLGEMMReshapeLHSMatrix = CLSynthetizeFunction<CLGEMMReshapeLHSMatrixKernel>;
+// Create function for ClGemmReshapeLhsMatrixKernel
+using CLGEMMReshapeLHSMatrix = CLSynthetizeOperator<ClGemmReshapeLhsMatrixKernel>;
 
-// Create function for CLGEMMReshapeRHSMatrixKernel
-using CLGEMMReshapeRHSMatrix = CLSynthetizeFunction<CLGEMMReshapeRHSMatrixKernel>;
+// Create function for ClGemmReshapeRhsMatrixKernel
+using CLGEMMReshapeRHSMatrix = CLSynthetizeOperator<ClGemmReshapeRhsMatrixKernel>;
 
-// Create function for CLGEMMMatrixMultiplyReshapedKernel
-using CLGEMMMatrixMultiplyReshaped = CLSynthetizeFunction<CLGEMMMatrixMultiplyReshapedKernel>;
+// Create function for ClGemmMatrixMultiplyReshapedKernel
+using CLGEMMMatrixMultiplyReshaped = CLSynthetizeOperator<ClGemmMatrixMultiplyReshapedKernel>;
 
 // Fixture for CLGEMMMatrixMultiplyReshaped
 template <typename T>
@@ -327,7 +328,7 @@ DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip(zip(zip(zip(zip(zi
                framework::dataset::make("Expected", { true, true, false, false, false, true, true,true})),
                     input0_info ,input1_info, input2_info, output_info, lhs_info, rhs_info, gemm_info, expected)
 {
-   ARM_COMPUTE_EXPECT(bool(CLGEMMMatrixMultiplyReshapedKernel::validate(&input0_info.clone()->set_is_resizable(true),
+   ARM_COMPUTE_EXPECT(bool(ClGemmMatrixMultiplyReshapedKernel::validate(&input0_info.clone()->set_is_resizable(true),
                                                           &input1_info.clone()->set_is_resizable(true),
                                                           &input2_info.clone()->set_is_resizable(true),
                                                           &output_info.clone()->set_is_resizable(true),1.f,1.f,
@@ -562,7 +563,7 @@ DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip(zip(zip(zip(zip(zi
                                                       false})),
                     input0_info ,input1_info, input2_info, output_info, lhs_info, rhs_info, gemm_info, expected)
 {
-   ARM_COMPUTE_EXPECT(bool(CLGEMMMatrixMultiplyReshapedKernel::validate(&input0_info.clone()->set_is_resizable(true),
+   ARM_COMPUTE_EXPECT(bool(ClGemmMatrixMultiplyReshapedKernel::validate(&input0_info.clone()->set_is_resizable(true),
                                                           &input1_info.clone()->set_is_resizable(true),
                                                           &input2_info.clone()->set_is_resizable(true),
                                                           &output_info.clone()->set_is_resizable(true),1.f,1.f,
@@ -933,7 +934,7 @@ DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip(zip(zip(zip(zip(zi
                                                       false})),
                     input0_info ,input1_info, input2_info, output_info, lhs_info, rhs_info, gemm_info, expected)
 {
-   ARM_COMPUTE_EXPECT(bool(CLGEMMMatrixMultiplyReshapedKernel::validate(&input0_info.clone()->set_is_resizable(true),
+   ARM_COMPUTE_EXPECT(bool(ClGemmMatrixMultiplyReshapedKernel::validate(&input0_info.clone()->set_is_resizable(true),
                                                           &input1_info.clone()->set_is_resizable(true),
                                                           &input2_info.clone()->set_is_resizable(true),
                                                           &output_info.clone()->set_is_resizable(true),1.f,1.f,
diff --git a/tests/validation/CL/GEMMMatrixMultiplyReshapedOnlyRHS.cpp b/tests/validation/CL/GEMMMatrixMultiplyReshapedOnlyRHS.cpp
index ebcecb8b78..88e99bcfef 100644
--- a/tests/validation/CL/GEMMMatrixMultiplyReshapedOnlyRHS.cpp
+++ b/tests/validation/CL/GEMMMatrixMultiplyReshapedOnlyRHS.cpp
@@ -26,8 +26,8 @@
 #include "arm_compute/core/utils/misc/ShapeCalculator.h"
 #include "arm_compute/runtime/CL/CLTensor.h"
 #include "arm_compute/runtime/CL/CLTensorAllocator.h"
-#include "src/core/CL/kernels/CLGEMMMatrixMultiplyReshapedOnlyRHSKernel.h"
-#include "src/core/CL/kernels/CLGEMMReshapeRHSMatrixKernel.h"
+#include "src/core/gpu/cl/kernels/ClGemmMatrixMultiplyReshapedOnlyRhsKernel.h"
+#include "src/core/gpu/cl/kernels/ClGemmReshapeRhsMatrixKernel.h"
 #include "tests/CL/CLAccessor.h"
 #include "tests/CL/Helper.h"
 #include "tests/PaddingCalculator.h"
@@ -45,12 +45,13 @@ namespace test
 namespace validation
 {
 using namespace arm_compute::misc::shape_calculator;
+using namespace arm_compute::opencl::kernels;
 
-// Create function for CLGEMMReshapeRHSMatrixKernel
-using CLGEMMReshapeRHSMatrix = CLSynthetizeFunction<CLGEMMReshapeRHSMatrixKernel>;
+// Create function for ClGemmReshapeRhsMatrixKernel
+using CLGEMMReshapeRHSMatrix = CLSynthetizeOperator<ClGemmReshapeRhsMatrixKernel>;
 
-// Create function for CLGEMMMatrixMultiplyReshapedOnlyRHSKernel
-using CLGEMMMatrixMultiplyReshapedOnlyRHS = CLSynthetizeFunction<CLGEMMMatrixMultiplyReshapedOnlyRHSKernel>;
+// Create function for ClGemmMatrixMultiplyReshapedOnlyRhsKernel
+using CLGEMMMatrixMultiplyReshapedOnlyRHS = CLSynthetizeOperator<ClGemmMatrixMultiplyReshapedOnlyRhsKernel>;
 
 // Fixture for CLGEMMMatrixMultiplyReshapedOnlyRHS
 template <typename T>
diff --git a/tests/validation/CL/GEMMReshapeLHSMatrix.cpp b/tests/validation/CL/GEMMReshapeLHSMatrix.cpp
index 34c37dffde..f995608308 100644
--- a/tests/validation/CL/GEMMReshapeLHSMatrix.cpp
+++ b/tests/validation/CL/GEMMReshapeLHSMatrix.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2018-2020 Arm Limited.
+ * Copyright (c) 2018-2021 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -25,7 +25,7 @@
 #include "arm_compute/core/utils/misc/ShapeCalculator.h"
 #include "arm_compute/runtime/CL/CLTensor.h"
 #include "arm_compute/runtime/CL/CLTensorAllocator.h"
-#include "src/core/CL/kernels/CLGEMMReshapeLHSMatrixKernel.h"
+#include "src/core/gpu/cl/kernels/ClGemmReshapeLhsMatrixKernel.h"
 #include "tests/CL/CLAccessor.h"
 #include "tests/CL/Helper.h"
 #include "tests/PaddingCalculator.h"
@@ -43,9 +43,10 @@ namespace test
 namespace validation
 {
 using namespace arm_compute::misc::shape_calculator;
+using namespace arm_compute::opencl::kernels;
 
 // Initialize the output tensor with zero and fill the border with zero
-using CLGEMMReshapeLHSMatrix = CLSynthetizeFunctionInitOutputWithZeroAndWithZeroConstantBorder<CLGEMMReshapeLHSMatrixKernel, 16>;
+using CLGEMMReshapeLHSMatrix = CLSynthetizeOperatorInitOutputWithZeroAndWithZeroConstantBorder<ClGemmReshapeLhsMatrixKernel, 16>;
 
 template <typename T>
 using CLGEMMReshapeLHSMatrixFixture = GEMMReshapeLHSMatrixValidationFixture<CLTensor, CLAccessor, CLGEMMReshapeLHSMatrix, T, false>;
diff --git a/tests/validation/CL/GEMMReshapeRHSMatrix.cpp b/tests/validation/CL/GEMMReshapeRHSMatrix.cpp
index 14048e81ec..ff1240ea2e 100644
--- a/tests/validation/CL/GEMMReshapeRHSMatrix.cpp
+++ b/tests/validation/CL/GEMMReshapeRHSMatrix.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2018-2020 Arm Limited.
+ * Copyright (c) 2018-2021 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -25,7 +25,7 @@
 #include "arm_compute/core/utils/misc/ShapeCalculator.h"
 #include "arm_compute/runtime/CL/CLTensor.h"
 #include "arm_compute/runtime/CL/CLTensorAllocator.h"
-#include "src/core/CL/kernels/CLGEMMReshapeRHSMatrixKernel.h"
+#include "src/core/gpu/cl/kernels/ClGemmReshapeRhsMatrixKernel.h"
 #include "tests/CL/CLAccessor.h"
 #include "tests/CL/Helper.h"
 #include "tests/PaddingCalculator.h"
@@ -73,9 +73,10 @@ const auto i_values = framework::dataset::make("interleave", { true, false });
 } // namespace
 
 using namespace arm_compute::misc::shape_calculator;
+using namespace arm_compute::opencl::kernels;
 
 // Initialize the output tensor with zero and fill the border with zero
-using CLGEMMReshapeRHSMatrix = CLSynthetizeFunctionInitOutputWithZeroAndWithZeroConstantBorder<CLGEMMReshapeRHSMatrixKernel, 16>;
+using CLGEMMReshapeRHSMatrix = CLSynthetizeOperatorInitOutputWithZeroAndWithZeroConstantBorder<ClGemmReshapeRhsMatrixKernel, 16>;
 
 template <typename T>
 using CLGEMMReshapeRHSMatrixFixture = GEMMReshapeRHSMatrixValidationFixture<CLTensor, CLAccessor, CLGEMMReshapeRHSMatrix, T>;
@@ -117,7 +118,7 @@ DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip(zip(zip(zip(
     rhs_info.transpose = true;
     rhs_info.interleave = true;
 
-    bool has_error = bool(CLGEMMReshapeRHSMatrixKernel::validate(&input_info.clone()->set_is_resizable(false), (output_info.total_size() == 0) ? nullptr : &output_info.clone()->set_is_resizable(false), rhs_info));
+    bool has_error = bool(ClGemmReshapeRhsMatrixKernel::validate(&input_info.clone()->set_is_resizable(false), (output_info.total_size() == 0) ? nullptr : &output_info.clone()->set_is_resizable(false), rhs_info));
     ARM_COMPUTE_EXPECT(has_error == expected, framework::LogLevel::ERRORS);
 }
 
@@ -158,9 +159,9 @@ DATA_TEST_CASE(ValidatePadding, framework::DatasetMode::ALL, combine(combine(com
         padding = round_up_width - output_shape[0];
     }
 
-    CLGEMMReshapeRHSMatrixKernel kernel;
+    ClGemmReshapeRhsMatrixKernel kernel;
 
-    kernel.configure(&input, &output, rhs_info);
+    kernel.configure(CLKernelLibrary::get().get_compile_context(), input.info(), output.info(), rhs_info);
 
     ARM_COMPUTE_EXPECT((output.info()->padding().right == padding), framework::LogLevel::ERRORS);
 }
diff --git a/tests/validation/CL/UNIT/DynamicTensor.cpp b/tests/validation/CL/UNIT/DynamicTensor.cpp
index 833256039e..ad2d4892ba 100644
--- a/tests/validation/CL/UNIT/DynamicTensor.cpp
+++ b/tests/validation/CL/UNIT/DynamicTensor.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2019-2020 Arm Limited.
+ * Copyright (c) 2019-2021 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -29,7 +29,6 @@
 #include "arm_compute/runtime/MemoryManagerOnDemand.h"
 #include "arm_compute/runtime/PoolManager.h"
 #include "src/core/CL/kernels/CLFillBorderKernel.h"
-#include "src/core/CL/kernels/CLGEMMReshapeRHSMatrixKernel.h"
 #include "src/core/CL/kernels/CLIm2ColKernel.h"
 #include "src/core/CL/kernels/CLL2NormalizeLayerKernel.h"
 #include "src/core/CL/kernels/CLReductionOperationKernel.h"
diff --git a/tests/validation/CL/UNIT/WeightsRetention.cpp b/tests/validation/CL/UNIT/WeightsRetention.cpp
index acf795e48b..1965f0f7a5 100644
--- a/tests/validation/CL/UNIT/WeightsRetention.cpp
+++ b/tests/validation/CL/UNIT/WeightsRetention.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2018-2020 Arm Limited.
+ * Copyright (c) 2018-2021 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -24,16 +24,6 @@
 #include "arm_compute/runtime/CL/functions/CLFullyConnectedLayer.h"
 #include "src/core/CL/kernels/CLDepthConvertLayerKernel.h"
 #include "src/core/CL/kernels/CLFillBorderKernel.h"
-#include "src/core/CL/kernels/CLGEMMLowpMatrixMultiplyNativeKernel.h"
-#include "src/core/CL/kernels/CLGEMMLowpMatrixMultiplyReshapedOnlyRHSKernel.h"
-#include "src/core/CL/kernels/CLGEMMLowpOffsetContributionKernel.h"
-#include "src/core/CL/kernels/CLGEMMLowpOffsetContributionOutputStageKernel.h"
-#include "src/core/CL/kernels/CLGEMMLowpReductionKernel.h"
-#include "src/core/CL/kernels/CLGEMMMatrixMultiplyKernel.h"
-#include "src/core/CL/kernels/CLGEMMMatrixMultiplyReshapedKernel.h"
-#include "src/core/CL/kernels/CLGEMMMatrixMultiplyReshapedOnlyRHSKernel.h"
-#include "src/core/CL/kernels/CLGEMMReshapeLHSMatrixKernel.h"
-#include "src/core/CL/kernels/CLGEMMReshapeRHSMatrixKernel.h"
 #include "tests/AssetsLibrary.h"
 #include "tests/CL/CLAccessor.h"
 #include "tests/Globals.h"
diff --git a/tests/validation/fixtures/GEMMFixture.h b/tests/validation/fixtures/GEMMFixture.h
index 9ad27c782e..c118da66ae 100644
--- a/tests/validation/fixtures/GEMMFixture.h
+++ b/tests/validation/fixtures/GEMMFixture.h
@@ -175,7 +175,7 @@ protected:
     SimpleTensor<T> _reference{};
 };
 
-template <typename TensorType, typename AccessorType, typename T, typename GEMMFunctionType>
+template <typename TensorType, typename AccessorType, typename T, typename GEMMOperatorType>
 class GEMMMatrixMultiplyValidationFixture : public framework::Fixture
 {
 public:
@@ -226,8 +226,8 @@ protected:
         // The output tensor will be auto-initialized within the function
 
         // Create and configure function
-        GEMMFunctionType gemm;
-        gemm.configure(gpu_arch, &lhs, &rhs, &bias, &dst, alpha, beta, false, reshape_info, fp16_mixed_precision, act_info);
+        GEMMOperatorType gemm;
+        gemm.configure(gpu_arch, lhs.info(), rhs.info(), bias.info(), dst.info(), alpha, beta, false, reshape_info, fp16_mixed_precision, act_info);
 
         ARM_COMPUTE_ASSERT(lhs.info()->is_resizable());
         ARM_COMPUTE_ASSERT(rhs.info()->is_resizable());
@@ -252,7 +252,12 @@ protected:
         fill(AccessorType(bias), 2);
 
         // Compute GEMM
-        gemm.run();
+        ITensorPack gemm_pack({ { ACL_SRC_0, &lhs },
+            { ACL_SRC_1, &rhs },
+            { ACL_SRC_2, &bias },
+            { ACL_DST, &dst }
+        });
+        gemm.run(gemm_pack);
 
         return dst;
     }
@@ -294,7 +299,7 @@ protected:
     SimpleTensor<T> _reference{};
 };
 
-template <typename TensorType, typename AccessorType, typename T, typename GEMMFunctionType>
+template <typename TensorType, typename AccessorType, typename T, typename GEMMOperatorType>
 class GEMMMatrixMultiply3DValidationFixture : public framework::Fixture
 {
 public:
@@ -344,8 +349,8 @@ protected:
         // The output tensor will be auto-initialized within the function
 
         // Create and configure function
-        GEMMFunctionType gemm;
-        gemm.configure(gpu_arch, &lhs, &rhs, &bias, &dst, alpha, beta, false, reshape_info, fp16_mixed_precision, act_info);
+        GEMMOperatorType gemm;
+        gemm.configure(gpu_arch, lhs.info(), rhs.info(), bias.info(), dst.info(), alpha, beta, false, reshape_info, fp16_mixed_precision, act_info);
 
         ARM_COMPUTE_ASSERT(lhs.info()->is_resizable());
         ARM_COMPUTE_ASSERT(rhs.info()->is_resizable());
@@ -370,7 +375,12 @@ protected:
         fill(AccessorType(bias), 2);
 
         // Compute GEMM
-        gemm.run();
+        ITensorPack gemm_pack({ { ACL_SRC_0, &lhs },
+            { ACL_SRC_1, &rhs },
+            { ACL_SRC_2, &bias },
+            { ACL_DST, &dst }
+        });
+        gemm.run(gemm_pack);
 
         return dst;
     }
@@ -411,7 +421,7 @@ protected:
     SimpleTensor<T> _reference{};
 };
 
-template <typename TensorType, typename AccessorType, typename T, typename ReshapeLHSFunctionType, typename ReshapeRHSFunctionType, typename GEMMFunctionType>
+template <typename TensorType, typename AccessorType, typename T, typename ReshapeLHSOperatorType, typename ReshapeRHSOperatorType, typename GEMMOperatorType>
 class GEMMMatrixMultiplyInterleavedTransposedValidationFixture : public framework::Fixture
 {
 public:
@@ -478,12 +488,12 @@ protected:
         // The output tensor will be auto-initialized within the function
 
         // Create and configure function
-        ReshapeLHSFunctionType reshape_lhs;
-        ReshapeRHSFunctionType reshape_rhs;
-        GEMMFunctionType       gemm;
-        reshape_lhs.configure(&lhs, &lhs_reshaped, lhs_info);
-        reshape_rhs.configure(&rhs, &rhs_reshaped, rhs_info);
-        gemm.configure(gpu_arch, &lhs_reshaped, &rhs_reshaped, &bias, &dst, alpha, beta, true, reshape_info, fp16_mixed_precision, act_info);
+        ReshapeLHSOperatorType reshape_lhs;
+        ReshapeRHSOperatorType reshape_rhs;
+        GEMMOperatorType       gemm;
+        reshape_lhs.configure(lhs.info(), lhs_reshaped.info(), lhs_info);
+        reshape_rhs.configure(rhs.info(), rhs_reshaped.info(), rhs_info);
+        gemm.configure(gpu_arch, lhs_reshaped.info(), rhs_reshaped.info(), bias.info(), dst.info(), alpha, beta, true, reshape_info, fp16_mixed_precision, act_info);
 
         ARM_COMPUTE_ASSERT(lhs.info()->is_resizable());
         ARM_COMPUTE_ASSERT(rhs.info()->is_resizable());
@@ -516,9 +526,16 @@ protected:
         fill(AccessorType(bias), 2);
 
         // Compute GEMM
-        reshape_lhs.run();
-        reshape_rhs.run();
-        gemm.run();
+        ITensorPack reshape_lhs_pack = { { ACL_SRC, &lhs }, { ACL_DST, &lhs_reshaped } };
+        reshape_lhs.run(reshape_lhs_pack);
+        ITensorPack reshape_rhs_pack = { { ACL_SRC, &rhs }, { ACL_DST, &rhs_reshaped } };
+        reshape_rhs.run(reshape_rhs_pack);
+        ITensorPack gemm_pack({ { ACL_SRC_0, &lhs_reshaped },
+            { ACL_SRC_1, &rhs_reshaped },
+            { ACL_SRC_2, &bias },
+            { ACL_DST, &dst }
+        });
+        gemm.run(gemm_pack);
 
         return dst;
     }
@@ -560,7 +577,7 @@ protected:
     SimpleTensor<T> _reference{};
 };
 
-template <typename TensorType, typename AccessorType, typename T, typename ReshapeLHSFunctionType, typename ReshapeRHSFunctionType, typename GEMMFunctionType>
+template <typename TensorType, typename AccessorType, typename T, typename ReshapeLHSOperatorType, typename ReshapeRHSOperatorType, typename GEMMOperatorType>
 class GEMMMatrixMultiplyInterleavedTransposed3DValidationFixture : public framework::Fixture
 {
 public:
@@ -626,12 +643,12 @@ protected:
         // The output tensor will be auto-initialized within the function
 
         // Create and configure function
-        ReshapeLHSFunctionType reshape_lhs;
-        ReshapeRHSFunctionType reshape_rhs;
-        GEMMFunctionType       gemm;
-        reshape_lhs.configure(&lhs, &lhs_reshaped, lhs_info);
-        reshape_rhs.configure(&rhs, &rhs_reshaped, rhs_info);
-        gemm.configure(gpu_arch, &lhs_reshaped, &rhs_reshaped, &bias, &dst, alpha, beta, true, reshape_info, fp16_mixed_precision, act_info);
+        ReshapeLHSOperatorType reshape_lhs;
+        ReshapeRHSOperatorType reshape_rhs;
+        GEMMOperatorType       gemm;
+        reshape_lhs.configure(lhs.info(), lhs_reshaped.info(), lhs_info);
+        reshape_rhs.configure(rhs.info(), rhs_reshaped.info(), rhs_info);
+        gemm.configure(gpu_arch, lhs_reshaped.info(), rhs_reshaped.info(), bias.info(), dst.info(), alpha, beta, true, reshape_info, fp16_mixed_precision, act_info);
 
         ARM_COMPUTE_ASSERT(lhs.info()->is_resizable());
         ARM_COMPUTE_ASSERT(rhs.info()->is_resizable());
@@ -664,9 +681,16 @@ protected:
         fill(AccessorType(bias), 2);
 
         // Compute GEMM
-        reshape_lhs.run();
-        reshape_rhs.run();
-        gemm.run();
+        ITensorPack reshape_lhs_pack = { { ACL_SRC, &lhs }, { ACL_DST, &lhs_reshaped } };
+        reshape_lhs.run(reshape_lhs_pack);
+        ITensorPack reshape_rhs_pack = { { ACL_SRC, &rhs }, { ACL_DST, &rhs_reshaped } };
+        reshape_rhs.run(reshape_rhs_pack);
+        ITensorPack gemm_pack({ { ACL_SRC_0, &lhs_reshaped },
+            { ACL_SRC_1, &rhs_reshaped },
+            { ACL_SRC_2, &bias },
+            { ACL_DST, &dst }
+        });
+        gemm.run(gemm_pack);
 
         return dst;
     }
@@ -707,7 +731,7 @@ protected:
     SimpleTensor<T> _reference{};
 };
 
-template <typename TensorType, typename AccessorType, typename T, typename ReshapeLHSFunctionType, typename ReshapeRHSFunctionType, typename GEMMFunctionType, bool fp_mixed_precision = false>
+template <typename TensorType, typename AccessorType, typename T, typename ReshapeLHSOperatorType, typename ReshapeRHSOperatorType, typename GEMMOperatorType, bool fp_mixed_precision = false>
 class GEMMMatrixMultiplyReshapedValidationFixture : public framework::Fixture
 {
 public:
@@ -786,9 +810,9 @@ protected:
         // The output tensor will be auto-initialized within the function
 
         // Create and configure function
-        ReshapeLHSFunctionType reshape_lhs;
-        ReshapeRHSFunctionType reshape_rhs;
-        GEMMFunctionType       gemm;
+        ReshapeLHSOperatorType reshape_lhs;
+        ReshapeRHSOperatorType reshape_rhs;
+        GEMMOperatorType       gemm;
 
         validate_result = bool(reshape_rhs.validate(rhs.info(), rhs_reshaped.info(), rhs_info));
         validate_result = validate_result || !rhs_info.export_to_cl_image;
@@ -797,9 +821,9 @@ protected:
             return nullptr;
         }
 
-        reshape_lhs.configure(&lhs, &lhs_reshaped, lhs_info);
-        reshape_rhs.configure(&rhs, &rhs_reshaped, rhs_info);
-        gemm.configure(&lhs_reshaped, &rhs_reshaped, &bias, &dst, alpha, beta, lhs_info, rhs_info, kernel_info);
+        reshape_lhs.configure(lhs.info(), lhs_reshaped.info(), lhs_info);
+        reshape_rhs.configure(rhs.info(), rhs_reshaped.info(), rhs_info);
+        gemm.configure(lhs_reshaped.info(), rhs_reshaped.info(), bias.info(), dst.info(), alpha, beta, lhs_info, rhs_info, kernel_info);
 
         ARM_COMPUTE_ASSERT(lhs.info()->is_resizable());
         ARM_COMPUTE_ASSERT(rhs.info()->is_resizable());
@@ -832,9 +856,16 @@ protected:
         fill(AccessorType(bias), 2);
 
         // Compute GEMM
-        reshape_lhs.run();
-        reshape_rhs.run();
-        gemm.run();
+        ITensorPack reshape_lhs_pack = { { ACL_SRC, &lhs }, { ACL_DST, &lhs_reshaped } };
+        reshape_lhs.run(reshape_lhs_pack);
+        ITensorPack reshape_rhs_pack = { { ACL_SRC, &rhs }, { ACL_DST, &rhs_reshaped } };
+        reshape_rhs.run(reshape_rhs_pack);
+        ITensorPack gemm_pack({ { ACL_SRC_0, &lhs_reshaped },
+            { ACL_SRC_1, &rhs_reshaped },
+            { ACL_SRC_2, &bias },
+            { ACL_DST, &dst }
+        });
+        gemm.run(gemm_pack);
 
         return dst;
     }
@@ -884,7 +915,7 @@ protected:
     SimpleTensor<T> _reference{};
 };
 
-template <typename TensorType, typename AccessorType, typename T, typename ReshapeLHSFunctionType, typename ReshapeRHSFunctionType, typename GEMMFunctionType, bool fp_mixed_precision = false>
+template <typename TensorType, typename AccessorType, typename T, typename ReshapeLHSOperatorType, typename ReshapeRHSOperatorType, typename GEMMOperatorType, bool fp_mixed_precision = false>
 class GEMMMatrixMultiplyReshaped3DValidationFixture : public framework::Fixture
 {
 public:
@@ -960,9 +991,9 @@ protected:
         // The output tensor will be auto-initialized within the function
 
         // Create and configure function
-        ReshapeLHSFunctionType reshape_lhs;
-        ReshapeRHSFunctionType reshape_rhs;
-        GEMMFunctionType       gemm;
+        ReshapeLHSOperatorType reshape_lhs;
+        ReshapeRHSOperatorType reshape_rhs;
+        GEMMOperatorType       gemm;
 
         validate_result = bool(reshape_rhs.validate(rhs.info(), rhs_reshaped.info(), rhs_info));
         validate_result = validate_result || !rhs_info.export_to_cl_image;
@@ -971,9 +1002,9 @@ protected:
             return nullptr;
         }
 
-        reshape_lhs.configure(&lhs, &lhs_reshaped, lhs_info);
-        reshape_rhs.configure(&rhs, &rhs_reshaped, rhs_info);
-        gemm.configure(&lhs_reshaped, &rhs_reshaped, &bias, &dst, alpha, beta, lhs_info, rhs_info, kernel_info);
+        reshape_lhs.configure(lhs.info(), lhs_reshaped.info(), lhs_info);
+        reshape_rhs.configure(rhs.info(), rhs_reshaped.info(), rhs_info);
+        gemm.configure(lhs_reshaped.info(), rhs_reshaped.info(), bias.info(), dst.info(), alpha, beta, lhs_info, rhs_info, kernel_info);
 
         ARM_COMPUTE_ASSERT(lhs.info()->is_resizable());
         ARM_COMPUTE_ASSERT(rhs.info()->is_resizable());
@@ -1006,9 +1037,16 @@ protected:
         fill(AccessorType(bias), 2);
 
         // Compute GEMM
-        reshape_lhs.run();
-        reshape_rhs.run();
-        gemm.run();
+        ITensorPack reshape_lhs_pack = { { ACL_SRC, &lhs }, { ACL_DST, &lhs_reshaped } };
+        reshape_lhs.run(reshape_lhs_pack);
+        ITensorPack reshape_rhs_pack = { { ACL_SRC, &rhs }, { ACL_DST, &rhs_reshaped } };
+        reshape_rhs.run(reshape_rhs_pack);
+        ITensorPack gemm_pack({ { ACL_SRC_0, &lhs_reshaped },
+            { ACL_SRC_1, &rhs_reshaped },
+            { ACL_SRC_2, &bias },
+            { ACL_DST, &dst }
+        });
+        gemm.run(gemm_pack);
 
         return dst;
     }
@@ -1057,7 +1095,7 @@ protected:
     SimpleTensor<T> _reference{};
 };
 
-template <typename TensorType, typename AccessorType, typename T, typename ReshapeRHSFunctionType, typename GEMMFunctionType>
+template <typename TensorType, typename AccessorType, typename T, typename ReshapeRHSOperatorType, typename GEMMOperatorType>
 class GEMMMatrixMultiplyReshapedOnlyRHSValidationFixture : public framework::Fixture
 {
 public:
@@ -1131,8 +1169,8 @@ protected:
         // The output tensor will be auto-initialized within the function
 
         // Create and configure function
-        ReshapeRHSFunctionType reshape_rhs;
-        GEMMFunctionType       gemm;
+        ReshapeRHSOperatorType reshape_rhs;
+        GEMMOperatorType       gemm;
 
         validate_result = bool(reshape_rhs.validate(rhs.info(), rhs_reshaped.info(), rhs_info));
         validate_result = validate_result || !rhs_info.export_to_cl_image;
@@ -1141,8 +1179,8 @@ protected:
             return nullptr;
         }
 
-        reshape_rhs.configure(&rhs, &rhs_reshaped, rhs_info);
-        gemm.configure(&lhs, &rhs_reshaped, &bias, &dst, alpha, beta, lhs_info, rhs_info, kernel_info);
+        reshape_rhs.configure(rhs.info(), rhs_reshaped.info(), rhs_info);
+        gemm.configure(lhs.info(), rhs_reshaped.info(), bias.info(), dst.info(), alpha, beta, lhs_info, rhs_info, kernel_info);
 
         ARM_COMPUTE_ASSERT(lhs.info()->is_resizable());
         ARM_COMPUTE_ASSERT(rhs.info()->is_resizable());
@@ -1173,8 +1211,14 @@ protected:
         fill(AccessorType(bias), 2);
 
         // Compute GEMM
-        reshape_rhs.run();
-        gemm.run();
+        ITensorPack reshape_rhs_pack = { { ACL_SRC, &rhs }, { ACL_DST, &rhs_reshaped } };
+        reshape_rhs.run(reshape_rhs_pack);
+        ITensorPack gemm_pack({ { ACL_SRC_0, &lhs },
+            { ACL_SRC_1, &rhs_reshaped },
+            { ACL_SRC_2, &bias },
+            { ACL_DST, &dst }
+        });
+        gemm.run(gemm_pack);
 
         return dst;
     }
@@ -1217,7 +1261,7 @@ protected:
     SimpleTensor<T> _reference{};
 };
 
-template <typename TensorType, typename AccessorType, typename T, typename ReshapeRHSFunctionType, typename GEMMFunctionType>
+template <typename TensorType, typename AccessorType, typename T, typename ReshapeRHSOperatorType, typename GEMMOperatorType>
 class GEMMMatrixMultiplyReshapedOnlyRHS3DValidationFixture : public framework::Fixture
 {
 public:
@@ -1289,8 +1333,8 @@ protected:
 
         // The output tensor will be auto-initialized within the function
         // Create and configure function
-        ReshapeRHSFunctionType reshape_rhs;
-        GEMMFunctionType       gemm;
+        ReshapeRHSOperatorType reshape_rhs;
+        GEMMOperatorType       gemm;
 
         validate_result = bool(reshape_rhs.validate(rhs.info(), rhs_reshaped.info(), rhs_info));
         validate_result = validate_result || !rhs_info.export_to_cl_image;
@@ -1299,8 +1343,8 @@ protected:
             return nullptr;
         }
 
-        reshape_rhs.configure(&rhs, &rhs_reshaped, rhs_info);
-        gemm.configure(&lhs, &rhs_reshaped, &bias, &dst, alpha, beta, lhs_info, rhs_info, kernel_info);
+        reshape_rhs.configure(rhs.info(), rhs_reshaped.info(), rhs_info);
+        gemm.configure(lhs.info(), rhs_reshaped.info(), bias.info(), dst.info(), alpha, beta, lhs_info, rhs_info, kernel_info);
 
         if(has_pad_y)
         {
@@ -1338,8 +1382,14 @@ protected:
         fill(AccessorType(bias), 2);
 
         // Compute GEMM
-        reshape_rhs.run();
-        gemm.run();
+        ITensorPack reshape_rhs_pack = { { ACL_SRC, &rhs }, { ACL_DST, &rhs_reshaped } };
+        reshape_rhs.run(reshape_rhs_pack);
+        ITensorPack gemm_pack({ { ACL_SRC_0, &lhs },
+            { ACL_SRC_1, &rhs_reshaped },
+            { ACL_SRC_2, &bias },
+            { ACL_DST, &dst }
+        });
+        gemm.run(gemm_pack);
 
         return dst;
     }
@@ -1381,7 +1431,7 @@ protected:
     SimpleTensor<T> _reference{};
 };
 
-template <typename TensorType, typename AccessorType, typename T, typename GEMMFunctionType>
+template <typename TensorType, typename AccessorType, typename T, typename GEMMOperatorType>
 class GEMMMatrixMultiplyNativeValidationFixture : public framework::Fixture
 {
 public:
@@ -1445,8 +1495,8 @@ protected:
         kernel_info.activation_info         = act_info;
 
         // Create and configure function
-        GEMMFunctionType gemm;
-        gemm.configure(&lhs, &rhs, &bias, &dst, alpha, beta, lhs_info, rhs_info, kernel_info);
+        GEMMOperatorType gemm;
+        gemm.configure(lhs.info(), rhs.info(), bias.info(), dst.info(), alpha, beta, lhs_info, rhs_info, kernel_info);
 
         ARM_COMPUTE_ASSERT(lhs.info()->is_resizable());
         ARM_COMPUTE_ASSERT(rhs.info()->is_resizable());
@@ -1471,7 +1521,12 @@ protected:
         fill(AccessorType(bias), 2);
 
         // Compute GEMM
-        gemm.run();
+        ITensorPack gemm_pack({ { ACL_SRC_0, &lhs },
+            { ACL_SRC_1, &rhs },
+            { ACL_SRC_2, &bias },
+            { ACL_DST, &dst }
+        });
+        gemm.run(gemm_pack);
 
         return dst;
     }
@@ -1513,7 +1568,7 @@ protected:
     SimpleTensor<T> _reference{};
 };
 
-template <typename TensorType, typename AccessorType, typename T, typename GEMMFunctionType>
+template <typename TensorType, typename AccessorType, typename T, typename GEMMOperatorType>
 class GEMMMatrixMultiplyNative3DValidationFixture : public framework::Fixture
 {
 public:
@@ -1576,8 +1631,8 @@ protected:
         // The output tensor will be auto-initialized within the function
 
         // Create and configure function
-        GEMMFunctionType gemm;
-        gemm.configure(&lhs, &rhs, &bias, &dst, alpha, beta, lhs_info, rhs_info, kernel_info);
+        GEMMOperatorType gemm;
+        gemm.configure(lhs.info(), rhs.info(), bias.info(), dst.info(), alpha, beta, lhs_info, rhs_info, kernel_info);
 
         ARM_COMPUTE_ASSERT(lhs.info()->is_resizable());
         ARM_COMPUTE_ASSERT(rhs.info()->is_resizable());
@@ -1602,7 +1657,12 @@ protected:
         fill(AccessorType(bias), 2);
 
         // Compute GEMM
-        gemm.run();
+        ITensorPack gemm_pack({ { ACL_SRC_0, &lhs },
+            { ACL_SRC_1, &rhs },
+            { ACL_SRC_2, &bias },
+            { ACL_DST, &dst }
+        });
+        gemm.run(gemm_pack);
 
         return dst;
     }
diff --git a/tests/validation/fixtures/GEMMLowpFixture.h b/tests/validation/fixtures/GEMMLowpFixture.h
index d7fe96cd3d..5cf210bab4 100644
--- a/tests/validation/fixtures/GEMMLowpFixture.h
+++ b/tests/validation/fixtures/GEMMLowpFixture.h
@@ -869,7 +869,7 @@ protected:
     SimpleTensor<int16_t> _reference{};
 };
 
-template <typename TensorType, typename AccessorType, typename ReshapeLHSFunctionType, typename ReshapeRHSFunctionType, typename GEMMFunctionType>
+template <typename TensorType, typename AccessorType, typename ReshapeLHSOperatorType, typename ReshapeRHSOperatorType, typename GEMMFunctionType>
 class GEMMLowpMatrixMultiplyReshapedValidationFixture : public framework::Fixture
 {
 public:
@@ -939,11 +939,11 @@ protected:
         // The output tensor will be auto-initialized within the function
 
         // Create and configure function
-        ReshapeLHSFunctionType reshape_lhs;
-        ReshapeRHSFunctionType reshape_rhs;
+        ReshapeLHSOperatorType reshape_lhs;
+        ReshapeRHSOperatorType reshape_rhs;
         GEMMFunctionType       gemm;
-        reshape_lhs.configure(&lhs, &lhs_reshaped, lhs_info);
-        reshape_rhs.configure(&rhs, &rhs_reshaped, rhs_info);
+        reshape_lhs.configure(lhs.info(), lhs_reshaped.info(), lhs_info);
+        reshape_rhs.configure(rhs.info(), rhs_reshaped.info(), rhs_info);
         gemm.configure(&lhs_reshaped, &rhs_reshaped, &dst, lhs_info, rhs_info, GEMMReshapeInfo(M, N, K));
 
         ARM_COMPUTE_ASSERT(lhs.info()->is_resizable());
@@ -969,8 +969,10 @@ protected:
         fill(AccessorType(rhs), 1);
 
         // Compute GEMM
-        reshape_lhs.run();
-        reshape_rhs.run();
+        ITensorPack reshape_lhs_pack = { { ACL_SRC, &lhs }, { ACL_DST, &lhs_reshaped } };
+        reshape_lhs.run(reshape_lhs_pack);
+        ITensorPack reshape_rhs_pack = { { ACL_SRC, &rhs }, { ACL_DST, &rhs_reshaped } };
+        reshape_rhs.run(reshape_rhs_pack);
         gemm.run();
 
         return dst;
@@ -1017,7 +1019,7 @@ protected:
     SimpleTensor<int32_t> _reference{};
 };
 
-template <typename TensorType, typename AccessorType, typename ReshapeLHSFunctionType, typename ReshapeRHSFunctionType, typename GEMMFunctionType>
+template <typename TensorType, typename AccessorType, typename ReshapeLHSOperatorType, typename ReshapeRHSOperatorType, typename GEMMFunctionType>
 class GEMMLowpMatrixMultiplyReshaped3DValidationFixture : public framework::Fixture
 {
 public:
@@ -1091,11 +1093,11 @@ protected:
         // The output tensor will be auto-initialized within the function
 
         // Create and configure function
-        ReshapeLHSFunctionType reshape_lhs;
-        ReshapeRHSFunctionType reshape_rhs;
+        ReshapeLHSOperatorType reshape_lhs;
+        ReshapeRHSOperatorType reshape_rhs;
         GEMMFunctionType       gemm;
-        reshape_lhs.configure(&lhs, &lhs_reshaped, lhs_info);
-        reshape_rhs.configure(&rhs, &rhs_reshaped, rhs_info);
+        reshape_lhs.configure(lhs.info(), lhs_reshaped.info(), lhs_info);
+        reshape_rhs.configure(rhs.info(), rhs_reshaped.info(), rhs_info);
         gemm.configure(&lhs_reshaped, &rhs_reshaped, &dst, lhs_info, rhs_info, GEMMReshapeInfo(M, N, K, 1, 1, m_h));
 
         ARM_COMPUTE_ASSERT(lhs.info()->is_resizable());
@@ -1121,8 +1123,10 @@ protected:
         fill(AccessorType(rhs), 1);
 
         // Compute GEMM
-        reshape_lhs.run();
-        reshape_rhs.run();
+        ITensorPack reshape_lhs_pack = { { ACL_SRC, &lhs }, { ACL_DST, &lhs_reshaped } };
+        reshape_lhs.run(reshape_lhs_pack);
+        ITensorPack reshape_rhs_pack = { { ACL_SRC, &rhs }, { ACL_DST, &rhs_reshaped } };
+        reshape_rhs.run(reshape_rhs_pack);
         gemm.run();
 
         return dst;
@@ -1171,7 +1175,7 @@ protected:
     SimpleTensor<int32_t> _reference{};
 };
 
-template <typename TensorType, typename AccessorType, typename ReshapeRHSFunctionType, typename GEMMFunctionType>
+template <typename TensorType, typename AccessorType, typename ReshapeRHSOperatorType, typename GEMMFunctionType>
 class GEMMLowpMatrixMultiplyReshapedOnlyRHSValidationFixture : public framework::Fixture
 {
 public:
@@ -1244,9 +1248,9 @@ protected:
         // The output tensor will be auto-initialized within the function
 
         // Create and configure function
-        ReshapeRHSFunctionType reshape_rhs;
+        ReshapeRHSOperatorType reshape_rhs;
         GEMMFunctionType       gemm;
-        reshape_rhs.configure(&rhs, &rhs_reshaped, rhs_info);
+        reshape_rhs.configure(rhs.info(), rhs_reshaped.info(), rhs_info);
         gemm.configure(&lhs, &rhs_reshaped, &dst, gemm_info);
 
         ARM_COMPUTE_ASSERT(lhs.info()->is_resizable());
@@ -1270,7 +1274,8 @@ protected:
         fill(AccessorType(rhs), 1);
 
         // Compute GEMM
-        reshape_rhs.run();
+        ITensorPack reshape_rhs_pack = { { ACL_SRC, &rhs }, { ACL_DST, &rhs_reshaped } };
+        reshape_rhs.run(reshape_rhs_pack);
         gemm.run();
 
         return dst;
@@ -1312,7 +1317,7 @@ protected:
     SimpleTensor<int32_t> _reference{};
 };
 
-template <typename TensorType, typename AccessorType, typename ReshapeRHSFunctionType, typename GEMMFunctionType>
+template <typename TensorType, typename AccessorType, typename ReshapeRHSOperatorType, typename GEMMFunctionType>
 class GEMMLowpMatrixMultiplyReshapedOnlyRHS3DValidationFixture : public framework::Fixture
 {
 public:
@@ -1389,9 +1394,9 @@ protected:
         // The output tensor will be auto-initialized within the function
 
         // Create and configure function
-        ReshapeRHSFunctionType reshape_rhs;
+        ReshapeRHSOperatorType reshape_rhs;
         GEMMFunctionType       gemm;
-        reshape_rhs.configure(&rhs, &rhs_reshaped, rhs_info);
+        reshape_rhs.configure(rhs.info(), rhs_reshaped.info(), rhs_info);
         gemm.configure(&lhs, &rhs_reshaped, &dst, gemm_info);
 
         ARM_COMPUTE_ASSERT(lhs.info()->is_resizable());
@@ -1415,7 +1420,8 @@ protected:
         fill(AccessorType(rhs), 1);
 
         // Compute GEMM
-        reshape_rhs.run();
+        ITensorPack reshape_rhs_pack = { { ACL_SRC, &rhs }, { ACL_DST, &rhs_reshaped } };
+        reshape_rhs.run(reshape_rhs_pack);
         gemm.run();
 
         return dst;
diff --git a/tests/validation/fixtures/GEMMReshapeLHSMatrixFixture.h b/tests/validation/fixtures/GEMMReshapeLHSMatrixFixture.h
index 70bafcc143..a9d6c9b6aa 100644
--- a/tests/validation/fixtures/GEMMReshapeLHSMatrixFixture.h
+++ b/tests/validation/fixtures/GEMMReshapeLHSMatrixFixture.h
@@ -46,7 +46,7 @@ namespace validation
 {
 using namespace arm_compute::misc::shape_calculator;
 
-template <typename TensorType, typename AccessorType, typename FunctionType, typename T, bool reinterpret_input_as_3d = false>
+template <typename TensorType, typename AccessorType, typename OperatorType, typename T, bool reinterpret_input_as_3d = false>
 class GEMMReshapeLHSMatrixValidationFixture : public framework::Fixture
 {
 public:
@@ -86,8 +86,8 @@ protected:
         // The output tensor will be auto-initialized within the function
 
         // Create and configure function
-        FunctionType gemm_lhs_reshape;
-        gemm_lhs_reshape.configure(&src, &dst, lhs_info, reinterpret_input_as_3d);
+        OperatorType gemm_lhs_reshape;
+        gemm_lhs_reshape.configure(src.info(), dst.info(), lhs_info, reinterpret_input_as_3d);
 
         ARM_COMPUTE_ASSERT(src.info()->is_resizable());
 
@@ -104,7 +104,8 @@ protected:
         fill(AccessorType(src));
 
         // Compute GEMM LHS matrix reshape function
-        gemm_lhs_reshape.run();
+        ITensorPack tensors = { { ACL_SRC, &src }, { ACL_DST, &dst } };
+        gemm_lhs_reshape.run(tensors);
 
         return dst;
     }
diff --git a/tests/validation/fixtures/GEMMReshapeRHSMatrixFixture.h b/tests/validation/fixtures/GEMMReshapeRHSMatrixFixture.h
index 1428adb3a7..cdb3ec3944 100644
--- a/tests/validation/fixtures/GEMMReshapeRHSMatrixFixture.h
+++ b/tests/validation/fixtures/GEMMReshapeRHSMatrixFixture.h
@@ -46,7 +46,7 @@ namespace validation
 {
 using namespace arm_compute::misc::shape_calculator;
 
-template <typename TensorType, typename AccessorType, typename FunctionType, typename T>
+template <typename TensorType, typename AccessorType, typename OperatorType, typename T>
 class GEMMReshapeRHSMatrixValidationFixture : public framework::Fixture
 {
 public:
@@ -85,8 +85,8 @@ protected:
         // The output tensor will be auto-initialized within the function
 
         // Create and configure function
-        FunctionType gemm_rhs_reshape;
-        gemm_rhs_reshape.configure(&src, &dst, rhs_info);
+        OperatorType gemm_rhs_reshape;
+        gemm_rhs_reshape.configure(src.info(), dst.info(), rhs_info);
 
         ARM_COMPUTE_ASSERT(src.info()->is_resizable());
 
@@ -103,7 +103,8 @@ protected:
         fill(AccessorType(src));
 
         // Compute GEMM RHS matrix reshape function
-        gemm_rhs_reshape.run();
+        ITensorPack tensors = { { ACL_SRC, &src }, { ACL_DST, &dst } };
+        gemm_rhs_reshape.run(tensors);
 
         return dst;
     }