diff --git a/arm_compute/runtime/CL/CLFunctions.h b/arm_compute/runtime/CL/CLFunctions.h
index a01e4a7..fe90b09 100644
--- a/arm_compute/runtime/CL/CLFunctions.h
+++ b/arm_compute/runtime/CL/CLFunctions.h
@@ -79,6 +79,7 @@
 #include "arm_compute/runtime/CL/functions/CLHistogram.h"
 #include "arm_compute/runtime/CL/functions/CLIntegralImage.h"
 #include "arm_compute/runtime/CL/functions/CLL2NormalizeLayer.h"
+#include "arm_compute/runtime/CL/functions/CLLSTMLayer.h"
 #include "arm_compute/runtime/CL/functions/CLLaplacianPyramid.h"
 #include "arm_compute/runtime/CL/functions/CLLaplacianReconstruct.h"
 #include "arm_compute/runtime/CL/functions/CLLocallyConnectedLayer.h"
diff --git a/arm_compute/runtime/CL/functions/CLLSTMLayer.h b/arm_compute/runtime/CL/functions/CLLSTMLayer.h
new file mode 100644
index 0000000..cf47f34
--- /dev/null
+++ b/arm_compute/runtime/CL/functions/CLLSTMLayer.h
@@ -0,0 +1,346 @@
+/*
+ * Copyright (c) 2018 ARM Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#ifndef __ARM_COMPUTE_CLLSTMLAYER_H__
+#define __ARM_COMPUTE_CLLSTMLAYER_H__
+
+#include "arm_compute/runtime/IFunction.h"
+
+#include "arm_compute/core/CL/kernels/CLActivationLayerKernel.h"
+#include "arm_compute/core/CL/kernels/CLArithmeticAdditionKernel.h"
+#include "arm_compute/core/CL/kernels/CLArithmeticSubtractionKernel.h"
+#include "arm_compute/core/CL/kernels/CLCopyKernel.h"
+#include "arm_compute/core/CL/kernels/CLPixelWiseMultiplicationKernel.h"
+#include "arm_compute/core/Types.h"
+#include "arm_compute/runtime/CL/CLMemoryGroup.h"
+#include "arm_compute/runtime/CL/CLTensor.h"
+#include "arm_compute/runtime/CL/functions/CLArithmeticAddition.h"
+#include "arm_compute/runtime/CL/functions/CLFullyConnectedLayer.h"
+#include "arm_compute/runtime/CL/functions/CLGEMM.h"
+#include "arm_compute/runtime/CL/functions/CLWidthConcatenateLayer.h"
+#include "arm_compute/runtime/IMemoryManager.h"
+
+#include <memory>
+
+namespace arm_compute
+{
+class ICLTensor;
+
+template <typename T>
+class LSTMParams
+{
+public:
+    /** Constructor */
+    LSTMParams()
+        : _input_to_input_weights(nullptr), _recurrent_to_input_weights(nullptr), _cell_to_input_weights(nullptr), _input_gate_bias(nullptr), _cell_to_forget_weights(nullptr),
+          _cell_to_output_weights(nullptr), _projection_weights(nullptr), _projection_bias(nullptr), _has_peephole_opt(false), _has_projection(false), _has_cifg_opt(true)
+    {
+    }
+    /** Prevent instances of this class from being copied (As this class contains pointers) */
+    LSTMParams(const LSTMParams &) = delete;
+    /** Prevent instances of this class from being copied (As this class contains pointers) */
+    LSTMParams &operator=(const LSTMParams &) = delete;
+    /** Default destructor */
+    ~LSTMParams() = default;
+    /** Set CIFG tensor parameters.
+     *
+     * @param[in] input_to_input_weights     2D weights tensor with dimensions [input_size, num_units]. Data types supported: F16/F32.
+     * @param[in] recurrent_to_input_weights 2D weights tensor with dimensions [output_size, num_units]. Data type supported: Same as @p input_to_input_weights.
+     * @param[in] cell_to_input_weights      1D weights tensor with dimensions [num_units]. Can be nullptr. Data type supported: Same as @p input_to_input_weights.
+     * @param[in] input_gate_bias            1D weights tensor with dimensions [num_units]. Data type supported: Same as @p input_to_input_weights
+     *
+     * @return Reference to this LSTMParams object
+     */
+    LSTMParams &set_cifg_params(const T *input_to_input_weights, const T *recurrent_to_input_weights, const T *cell_to_input_weights, const T *input_gate_bias)
+    {
+        _input_to_input_weights     = input_to_input_weights;
+        _recurrent_to_input_weights = recurrent_to_input_weights;
+        _cell_to_input_weights      = cell_to_input_weights;
+        _input_gate_bias            = input_gate_bias;
+        _has_cifg_opt               = false;
+        return *this;
+    }
+    /** Set projection tensor parameters.
+     *
+     * @param[in] projection_weights 2D weights tensor with dimensions [output_size, num_units]. Data type supported: Data types supported: F16/F32.
+     * @param[in] projection_bias    1D weights tensor with dimensions [output_size]. Data type supported: Same as @p projection_weights.
+     *
+     * @return Reference to this LSTMParams object
+     */
+    LSTMParams &set_projection_params(const T *projection_weights, const T *projection_bias)
+    {
+        _projection_weights = projection_weights;
+        _projection_bias    = projection_bias;
+        _has_projection     = true;
+        return *this;
+    }
+    /** Set peephole tensor parameters.
+     *
+     * @param[in] cell_to_input_weights  1D weights tensor with dimensions [num_units]. Data type supported: Data types supported: F16/F32.
+     * @param[in] cell_to_forget_weights 1D weights tensor with dimensions [num_units]. Data type supported: Same as @p cell_to_input_weights.
+     * @param[in] cell_to_output_weights 1D weights tensor with dimensions [num_units]. Data type supported: Same as @p cell_to_input_weights.
+     *
+     * @return Reference to this LSTMParams object
+     */
+    LSTMParams &set_peephole_params(const T *cell_to_input_weights, const T *cell_to_forget_weights, const T *cell_to_output_weights)
+    {
+        _cell_to_input_weights  = cell_to_input_weights;
+        _cell_to_forget_weights = cell_to_forget_weights;
+        _cell_to_output_weights = cell_to_output_weights;
+        _has_peephole_opt       = true;
+        return *this;
+    }
+
+    const T *input_to_input_weights() const
+    {
+        return _input_to_input_weights;
+    }
+
+    const T *recurrent_to_input_weights() const
+    {
+        return _recurrent_to_input_weights;
+    }
+
+    const T *cell_to_input_weights() const
+    {
+        return _cell_to_input_weights;
+    }
+
+    const T *input_gate_bias() const
+    {
+        return _input_gate_bias;
+    }
+
+    const T *cell_to_forget_weights() const
+    {
+        return _cell_to_forget_weights;
+    }
+
+    const T *cell_to_output_weights() const
+    {
+        return _cell_to_output_weights;
+    }
+
+    const T *projection_weights() const
+    {
+        return _projection_weights;
+    }
+
+    const T *projection_bias() const
+    {
+        return _projection_bias;
+    }
+
+    bool has_peephole_opt() const
+    {
+        return _has_peephole_opt;
+    }
+
+    bool has_projection() const
+    {
+        return _has_projection;
+    }
+
+    bool has_cifg_opt() const
+    {
+        return _has_cifg_opt;
+    }
+
+private:
+    const T *_input_to_input_weights;
+    const T *_recurrent_to_input_weights;
+    const T *_cell_to_input_weights;
+    const T *_input_gate_bias;
+    const T *_cell_to_forget_weights;
+    const T *_cell_to_output_weights;
+    const T *_projection_weights;
+    const T *_projection_bias;
+    bool     _has_peephole_opt;
+    bool     _has_projection;
+    bool     _has_cifg_opt;
+};
+
+/** This function performs a single time step in a Long Short-Term Memory (LSTM) layer.
+ *
+ */
+class CLLSTMLayer : public IFunction
+{
+public:
+    /** Default constructor */
+    CLLSTMLayer(std::shared_ptr<IMemoryManager> memory_manager = nullptr);
+    /** Initialize function's tensors.
+     *
+     * @param[in]      input                       Source tensor. Input is a 2D tensor with dimensions [input_size, batch_size]. Data types supported: F16/F32.
+     * @param[in]      input_to_forget_weights     2D weights tensor with dimensions [input_size, num_units]. Data type supported: Same as @p input.
+     * @param[in]      input_to_cell_weights       2D weights tensor with dimensions [input_size, num_units]. Data type supported: Same as @p input.
+     * @param[in]      input_to_output_weights     2D weights tensor with dimensions [input_size, num_units]. Data type supported: Same as @p input.
+     * @param[in]      recurrent_to_forget_weights 2D weights tensor with dimensions [output_size, num_units]. Data type supported: Same as @p input.
+     * @param[in]      recurrent_to_cell_weights   2D weights tensor with dimensions [output_size, num_units]. Data type supported: Same as @p input.
+     * @param[in]      recurrent_to_output_weights 2D weights tensor with dimensions [output_size, num_units]. Data type supported: Same as @p input.
+     * @param[in]      forget_gate_bias            1D weights tensor with dimensions [num_units]. Data type supported: Same as @p input.
+     * @param[in]      cell_bias                   1D weights tensor with dimensions [num_units]. Data type supported: Same as @p input.
+     * @param[in]      output_gate_bias            1D weights tensor with dimensions [num_units]. Data type supported: Same as @p input.
+     * @param[in, out] output_state                2D weights tensor with dimensions [output_size, batch_size]. Data type supported: Same as @p input.
+     * @param[in, out] cell_state                  2D tensor with dimensions [num_units, batch_size]. Data type supported: Same as @p input.
+     * @param[out]     scratch_buffer              2D tensor with dimensions [num_units * 4, batch_size] with CIFG or [num_units * 3, batch_size] without CIGF. Data type supported: Same as @p input.
+     * @param[out]     output                      Destination tensor. Output is a 2D tensor with dimensions [output_size, batch_size].
+     *                                             Data types supported: Same as @p input.
+     * @param[in]      lstm_params                 (Optional) Weights tensors used in peephole optimization:
+     *                                             input_to_input_weights       2D weights tensor with dimensions [input_size, num_units]. Data type supported: Same as @p input.
+     *                                             recurrent_to_input_weights   2D weights tensor with dimensions [output_size, num_units]. Data type supported: Same as @p input.
+     *                                             cell_to_input_weights        1D weights tensor with dimensions [num_units]. Can be nullptr. Data type supported: Same as @p input.
+     *                                             cell_to_forget_weights       1D weights tensor with dimensions [num_units]. Data type supported: Same as @p input.
+     *                                             cell_to_output_weights       1D weights tensor with dimensions [num_units]. Data type supported: Same as @p input.
+     *                                             input_gate_bias              1D weights tensor with dimensions [num_units]. Data type supported: Same as @p input
+     *                                             projection_weights           2D weights tensor with dimensions [output_size, num_units]. Data type supported: Same as @p input.
+     *                                             projection_bias              1D weights tensor with dimensions [output_size]. Data type supported: Same as @p input.
+     * @param[in]      activation_info             Contains activation information described in @ref ActivationLayerInfo.
+     * @param[in]      cell_threshold              The clipping threshold for the cell state, such that values are bound within [-cell_clip, cell_clip]. If set to 0.0 then clipping is disabled.
+     * @param[in]      projection_threshold        The clipping threshold for the output from the projection layer, such that values are bound within [-proj_clip, proj_clip]. If set to 0.0 then clipping is disabled.
+     */
+    void configure(const ICLTensor *input, const ICLTensor *input_to_forget_weights, const ICLTensor *input_to_cell_weights, const ICLTensor *input_to_output_weights,
+                   const ICLTensor *recurrent_to_forget_weights, const ICLTensor *recurrent_to_cell_weights, const ICLTensor *recurrent_to_output_weights,
+                   const ICLTensor *forget_gate_bias, const ICLTensor *cell_bias, const ICLTensor *output_gate_bias, ICLTensor *output_state, ICLTensor *cell_state, ICLTensor *scratch_buffer, ICLTensor *output,
+                   const LSTMParams<ICLTensor> &lstm_params, const ActivationLayerInfo &activation_info, float cell_threshold = 0.f, float projection_threshold = 0.f);
+
+    /** Static function to check if given info will lead to a valid configuration of @ref CLLSTMLayer
+     *
+     * @param[in] input                       Source tensor info. Input is a 2D tensor info with dimensions [input_size, batch_size]. Data types supported: F16/F32.
+     * @param[in] input_to_forget_weights     2D weights tensor info with dimensions [input_size, num_units]. Data type supported: Same as @p input.
+     * @param[in] input_to_cell_weights       2D weights tensor info with dimensions [input_size, num_units]. Data type supported: Same as @p input.
+     * @param[in] input_to_output_weights     2D weights tensor info with dimensions [input_size, num_units]. Data type supported: Same as @p input.
+     * @param[in] recurrent_to_forget_weights 2D weights tensor info with dimensions [output_size, num_units]. Data type supported: Same as @p input.
+     * @param[in] recurrent_to_cell_weights   2D weights tensor info with dimensions [output_size, num_units]. Data type supported: Same as @p input.
+     * @param[in] recurrent_to_output_weights 2D weights tensor info with dimensions [output_size, num_units]. Data type supported: Same as @p input.
+     * @param[in] forget_gate_bias            1D weights tensor info with dimensions [num_units]. Data type supported: Same as @p input.
+     * @param[in] cell_bias                   1D weights tensor info with dimensions [num_units]. Data type supported: Same as @p input.
+     * @param[in] output_gate_bias            1D weights tensor info with dimensions [num_units]. Data type supported: Same as @p input.
+     * @param[in] output_state                2D weights tensor info with dimensions [output_size, batch_size]. Data type supported: Same as @p input.
+     * @param[in] cell_state                  2D tensor info with dimensions [num_units, batch_size]. Data type supported: Same as @p input.
+     * @param[in] scratch_buffer              2D tensor info with dimensions [num_units * 4, batch_size] with CIFG or [num_units * 3, batch_size] without CIGF. Data type supported: Same as @p input.
+     * @param[in] output                      Destination tensor info. Output is a 2D tensor with dimensions [output_size, batch_size].
+     *                                        Data types supported: Same as @p input.
+     * @param[in] lstm_params                 (Optional) Weights tensors used in peephole optimization:
+     *                                        input_to_input_weights       2D weights tensor with dimensions [input_size, num_units]. Data type supported: Same as @p input.
+     *                                        recurrent_to_input_weights   2D weights tensor with dimensions [output_size, num_units]. Data type supported: Same as @p input.
+     *                                        cell_to_input_weights        1D weights tensor with dimensions [num_units]. Can be nullptr. Data type supported: Same as @p input.
+     *                                        cell_to_forget_weights       1D weights tensor with dimensions [num_units]. Data type supported: Same as @p input.
+     *                                        cell_to_output_weights       1D weights tensor with dimensions [num_units]. Data type supported: Same as @p input.
+     *                                        input_gate_bias              1D weights tensor with dimensions [num_units]. Data type supported: Same as @p input
+     *                                        projection_weights           2D weights tensor with dimensions [output_size, num_units]. Data type supported: Same as @p input.
+     *                                        projection_bias              1D weights tensor with dimensions [output_size]. Data type supported: Same as @p input.
+     * @param[in] activation_info             Contains activation information described in @ref ActivationLayerInfo.
+     * @param[in] cell_threshold              The clipping threshold for the cell state, such that values are bound within [-cell_clip, cell_clip]. If set to 0.0 then clipping is disabled.
+     * @param[in] projection_threshold        The clipping threshold for the output from the projection layer, such that values are bound within [-proj_clip, proj_clip]. If set to 0.0 then clipping is disabled.
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *input, const ITensorInfo *input_to_forget_weights, const ITensorInfo *input_to_cell_weights, const ITensorInfo *input_to_output_weights,
+                           const ITensorInfo *recurrent_to_forget_weights, const ITensorInfo *recurrent_to_cell_weights, const ITensorInfo *recurrent_to_output_weights,
+                           const ITensorInfo *forget_gate_bias, const ITensorInfo *cell_bias, const ITensorInfo *output_gate_bias,
+                           const ITensorInfo *output_state, const ITensorInfo *cell_state, const ITensorInfo *scratch_buffer, const ITensorInfo *output,
+                           const LSTMParams<ITensorInfo> &lstm_params, const ActivationLayerInfo &activation_info, float cell_threshold = 0.f, float projection_threshold = 0.f);
+
+    // Inherited methods overridden:
+    void run() override;
+
+private:
+    CLMemoryGroup                   _memory_group;
+    CLFullyConnectedLayer           _fully_connected_input_gate;
+    CLGEMM                          _gemm_input_gate1;
+    CLGEMM                          _gemm_input_gate2;
+    CLTransposeKernel               _transpose_input_gate1;
+    CLTransposeKernel               _transpose_input_gate2;
+    CLArithmeticAdditionKernel      _accum_input_gate1;
+    CLArithmeticAddition            _accum_input_gate2;
+    CLArithmeticSubtractionKernel   _subtract_input_gate;
+    CLActivationLayerKernel         _activation_input_gate;
+    CLFullyConnectedLayer           _fully_connected_forget_gate;
+    CLGEMM                          _gemm_forget_gate1;
+    CLGEMM                          _gemm_forget_gate2;
+    CLTransposeKernel               _transpose_forget_gate1;
+    CLTransposeKernel               _transpose_forget_gate2;
+    CLArithmeticAdditionKernel      _accum_forget_gate1;
+    CLArithmeticAddition            _accum_forget_gate2;
+    CLActivationLayerKernel         _activation_forget_gate;
+    CLFullyConnectedLayer           _fully_connected_cell_state;
+    CLGEMM                          _gemm_cell_state1;
+    CLGEMM                          _gemm_cell_state2;
+    CLTransposeKernel               _transpose_cell_state1;
+    CLArithmeticAdditionKernel      _accum_cell_state1;
+    CLArithmeticAdditionKernel      _accum_cell_state2;
+    CLPixelWiseMultiplicationKernel _pixelwise_mul_cell_state1;
+    CLActivationLayerKernel         _activation_cell_state;
+    CLActivationLayerKernel         _cell_clip;
+    CLPixelWiseMultiplicationKernel _pixelwise_mul_cell_state2;
+    CLFullyConnectedLayer           _fully_connected_output;
+    CLGEMM                          _gemm_output1;
+    CLGEMM                          _gemm_output2;
+    CLTransposeKernel               _transpose_output1;
+    CLTransposeKernel               _transpose_output2;
+    CLArithmeticAdditionKernel      _accum_output1;
+    CLArithmeticAddition            _accum_output2;
+    CLActivationLayerKernel         _activation_output;
+    CLActivationLayerKernel         _activation_output_state;
+    CLPixelWiseMultiplicationKernel _pixelwise_mul_output_state;
+    CLFullyConnectedLayer           _fully_connected_output_state;
+    CLGEMM                          _gemm_output_state;
+    CLArithmeticAdditionKernel      _accum_output_state;
+    CLActivationLayerKernel         _projection_clip;
+    CLCopyKernel                    _copy_cell_state;
+    CLCopyKernel                    _copy_output;
+    CLWidthConcatenateLayer         _concat_scratch_buffer;
+    CLTensor                        _input_gate_out1;
+    CLTensor                        _input_gate_out2;
+    CLTensor                        _input_gate_out3;
+    CLTensor                        _input_gate_out4;
+    CLTensor                        _input_gate_out5;
+    CLTensor                        _input_gate_out6;
+    CLTensor                        _forget_gate_out1;
+    CLTensor                        _forget_gate_out2;
+    CLTensor                        _forget_gate_out3;
+    CLTensor                        _forget_gate_out4;
+    CLTensor                        _forget_gate_out5;
+    CLTensor                        _forget_gate_out6;
+    CLTensor                        _cell_state_out1;
+    CLTensor                        _cell_state_out2;
+    CLTensor                        _cell_state_out3;
+    CLTensor                        _cell_state_out4;
+    CLTensor                        _cell_state_out5;
+    CLTensor                        _output1;
+    CLTensor                        _output2;
+    CLTensor                        _output3;
+    CLTensor                        _output4;
+    CLTensor                        _output5;
+    CLTensor                        _output6;
+    CLTensor                        _cell_state_activation;
+    CLTensor                        _output_projection1;
+    CLTensor                        _ones;
+    bool                            _run_peephole_opt;
+    bool                            _run_cifg_opt;
+    bool                            _perform_cell_clipping;
+    bool                            _has_projection_weights;
+    bool                            _perform_projection_clipping;
+};
+}
+#endif /* __ARM_COMPUTE_CLLSTMLAYER_H__ */
diff --git a/src/runtime/CL/functions/CLLSTMLayer.cpp b/src/runtime/CL/functions/CLLSTMLayer.cpp
new file mode 100644
index 0000000..930d311
--- /dev/null
+++ b/src/runtime/CL/functions/CLLSTMLayer.cpp
@@ -0,0 +1,508 @@
+/*
+ * Copyright (c) 2018 ARM Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#include "arm_compute/runtime/CL/functions/CLLSTMLayer.h"
+
+#include "arm_compute/core/PixelValue.h"
+#include "arm_compute/core/Utils.h"
+#include "arm_compute/core/Validate.h"
+#include "arm_compute/core/utils/misc/ShapeCalculator.h"
+#include "arm_compute/core/utils/quantization/AsymmHelpers.h"
+#include "arm_compute/runtime/CL/CLScheduler.h"
+
+#include <cmath>
+#include <memory>
+#include <tuple>
+
+using namespace arm_compute;
+using namespace arm_compute::misc::shape_calculator;
+
+CLLSTMLayer::CLLSTMLayer(std::shared_ptr<IMemoryManager> memory_manager)
+    : _memory_group(std::move(memory_manager)), _fully_connected_input_gate(), _gemm_input_gate1(), _gemm_input_gate2(), _transpose_input_gate1(), _transpose_input_gate2(), _accum_input_gate1(),
+      _accum_input_gate2(), _subtract_input_gate(), _activation_input_gate(), _fully_connected_forget_gate(), _gemm_forget_gate1(), _gemm_forget_gate2(), _transpose_forget_gate1(),
+      _transpose_forget_gate2(), _accum_forget_gate1(), _accum_forget_gate2(), _activation_forget_gate(), _fully_connected_cell_state(), _gemm_cell_state1(), _gemm_cell_state2(), _transpose_cell_state1(),
+      _accum_cell_state1(), _accum_cell_state2(), _pixelwise_mul_cell_state1(), _activation_cell_state(), _cell_clip(), _pixelwise_mul_cell_state2(), _fully_connected_output(), _gemm_output1(),
+      _gemm_output2(), _transpose_output1(), _transpose_output2(), _accum_output1(), _accum_output2(), _activation_output(), _activation_output_state(), _pixelwise_mul_output_state(),
+      _fully_connected_output_state(), _gemm_output_state(), _accum_output_state(), _projection_clip(), _copy_cell_state(), _copy_output(), _concat_scratch_buffer(), _input_gate_out1(), _input_gate_out2(),
+      _input_gate_out3(), _input_gate_out4(), _input_gate_out5(), _input_gate_out6(), _forget_gate_out1(), _forget_gate_out2(), _forget_gate_out3(), _forget_gate_out4(), _forget_gate_out5(),
+      _forget_gate_out6(), _cell_state_out1(), _cell_state_out2(), _cell_state_out3(), _cell_state_out4(), _cell_state_out5(), _output1(), _output2(), _output3(), _output4(), _output5(), _output6(),
+      _cell_state_activation(), _output_projection1(), _ones(), _run_peephole_opt(false), _run_cifg_opt(false), _perform_cell_clipping(false), _has_projection_weights(false),
+      _perform_projection_clipping(false)
+{
+}
+
+void CLLSTMLayer::configure(const ICLTensor *input, const ICLTensor *input_to_forget_weights, const ICLTensor *input_to_cell_weights, const ICLTensor *input_to_output_weights,
+                            const ICLTensor *recurrent_to_forget_weights, const ICLTensor *recurrent_to_cell_weights, const ICLTensor *recurrent_to_output_weights,
+                            const ICLTensor *forget_gate_bias, const ICLTensor *cell_bias, const ICLTensor *output_gate_bias,
+                            ICLTensor *output_state, ICLTensor *cell_state, ICLTensor *scratch_buffer, ICLTensor *output, const LSTMParams<ICLTensor> &lstm_params, const ActivationLayerInfo &activation_info,
+                            float cell_threshold, float projection_threshold)
+{
+    ARM_COMPUTE_ERROR_ON_NULLPTR(input, input_to_forget_weights, input_to_cell_weights, input_to_output_weights, recurrent_to_forget_weights, recurrent_to_cell_weights, recurrent_to_output_weights,
+                                 forget_gate_bias, cell_bias, output_gate_bias, output_state, cell_state);
+    LSTMParams<ITensorInfo> lstm_params_info;
+    if(lstm_params.has_peephole_opt())
+    {
+        lstm_params_info.set_peephole_params(lstm_params.cell_to_input_weights()->info(), lstm_params.cell_to_forget_weights()->info(), lstm_params.cell_to_output_weights()->info());
+    }
+    if(lstm_params.has_projection())
+    {
+        lstm_params_info.set_projection_params(lstm_params.projection_weights()->info(), lstm_params.projection_bias()->info());
+    }
+    if(!lstm_params.has_cifg_opt())
+    {
+        lstm_params_info.set_cifg_params(lstm_params.input_to_input_weights()->info(), lstm_params.recurrent_to_input_weights()->info(),
+                                         lstm_params.cell_to_input_weights()->info(), lstm_params.input_gate_bias()->info());
+    }
+    ARM_COMPUTE_ERROR_THROW_ON(CLLSTMLayer::validate(input->info(), input_to_forget_weights->info(),
+                                                     input_to_cell_weights->info(), input_to_output_weights->info(),
+                                                     recurrent_to_forget_weights->info(), recurrent_to_cell_weights->info(), recurrent_to_output_weights->info(),
+                                                     forget_gate_bias->info(), cell_bias->info(), output_gate_bias->info(),
+                                                     output_state->info(), cell_state->info(), scratch_buffer->info(), output->info(), lstm_params_info,
+                                                     activation_info, cell_threshold, projection_threshold));
+
+    const TensorShape cell_state_shape = cell_state->info()->tensor_shape();
+
+    TensorShape forget_gate1_shape = compute_transposed_shape(*recurrent_to_output_weights->info());
+    TensorShape forget_gate2_shape = compute_transposed_shape(*forget_gate_bias->info());
+    TensorShape forget_gate3_shape{ 1, output_state->info()->dimension(1) };
+    _forget_gate_out1.allocator()->init(TensorInfo(cell_state_shape, 1, input->info()->data_type()));
+    _forget_gate_out2.allocator()->init(TensorInfo(forget_gate1_shape, 1, input->info()->data_type()));
+    _forget_gate_out3.allocator()->init(TensorInfo(cell_state_shape, 1, input->info()->data_type()));
+    _forget_gate_out6.allocator()->init(TensorInfo(cell_state_shape, 1, input->info()->data_type()));
+
+    // Configure block that calculates the forget gate
+    // forget_gate = Activation(input * input_to_forget_weights + output_state * recurrent_to_forget_weights + cell_state * cell_to_forget_weights + forget_gate_bias)
+    _memory_group.manage(&_forget_gate_out1);
+    _fully_connected_forget_gate.configure(input, input_to_forget_weights, forget_gate_bias, &_forget_gate_out1, true, false);
+    _memory_group.manage(&_forget_gate_out2);
+    _transpose_forget_gate1.configure(recurrent_to_forget_weights, &_forget_gate_out2);
+    _memory_group.manage(&_forget_gate_out3);
+    _gemm_forget_gate1.configure(output_state, &_forget_gate_out2, nullptr, &_forget_gate_out3, 1.f, 0.f);
+    _forget_gate_out2.allocator()->allocate();
+    _memory_group.manage(&_forget_gate_out6);
+    _accum_forget_gate1.configure(&_forget_gate_out1, &_forget_gate_out3, &_forget_gate_out6, ConvertPolicy::SATURATE);
+    CLTensor *forget_gate_out = &_forget_gate_out6;
+
+    if(lstm_params.has_peephole_opt())
+    {
+        _forget_gate_out4.allocator()->init(TensorInfo(forget_gate2_shape, 1, input->info()->data_type()));
+        _forget_gate_out5.allocator()->init(TensorInfo(forget_gate3_shape, 1, input->info()->data_type()));
+
+        _run_peephole_opt = true;
+        _memory_group.manage(&_forget_gate_out4);
+        _transpose_forget_gate2.configure(lstm_params.cell_to_forget_weights(), &_forget_gate_out4);
+        _memory_group.manage(&_forget_gate_out5);
+        _gemm_forget_gate2.configure(cell_state, &_forget_gate_out4, nullptr, &_forget_gate_out5, 1.f, 0.f);
+        _forget_gate_out4.allocator()->allocate();
+        _accum_forget_gate2.configure(&_forget_gate_out6, &_forget_gate_out5, &_forget_gate_out3, ConvertPolicy::SATURATE);
+        _forget_gate_out5.allocator()->allocate();
+        _forget_gate_out6.allocator()->allocate();
+        forget_gate_out = &_forget_gate_out3;
+    }
+    else
+    {
+        _forget_gate_out3.allocator()->allocate();
+    }
+    _activation_forget_gate.configure(forget_gate_out, &_forget_gate_out1, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LOGISTIC));
+    forget_gate_out->allocator()->allocate();
+
+    TensorShape input_gate3_shape{ 1, output_state->info()->dimension(1) };
+    _input_gate_out1.allocator()->init(TensorInfo(cell_state_shape, 1, input->info()->data_type()));
+    _input_gate_out5.allocator()->init(TensorInfo(input_gate3_shape, 1, input->info()->data_type()));
+
+    // Configure block that calculates the input gate
+    // input_gate = Activation(input * input_to_input_weights + output_state * recurrent_to_input_weights + cell_state * cell_to_input_weights + input_gate_bias), without CIFG
+    // input_gate = 1 - forget_gate, with CIFG
+    if(lstm_params.has_cifg_opt())
+    {
+        _memory_group.manage(&_input_gate_out1);
+        _ones.allocator()->init(TensorInfo(cell_state_shape, 1, input->info()->data_type()));
+        _subtract_input_gate.configure(&_ones, &_forget_gate_out1, &_input_gate_out1, ConvertPolicy::SATURATE);
+        _ones.allocator()->allocate();
+        _run_cifg_opt = true;
+    }
+    else
+    {
+        TensorShape input_gate1_shape = compute_transposed_shape(*recurrent_to_output_weights->info());
+        TensorShape input_gate2_shape = compute_transposed_shape(*lstm_params.cell_to_input_weights()->info());
+
+        _input_gate_out2.allocator()->init(TensorInfo(input_gate1_shape, 1, input->info()->data_type()));
+        _input_gate_out3.allocator()->init(TensorInfo(cell_state_shape, 1, input->info()->data_type()));
+        _input_gate_out4.allocator()->init(TensorInfo(input_gate2_shape, 1, input->info()->data_type()));
+        _input_gate_out6.allocator()->init(TensorInfo(cell_state_shape, 1, input->info()->data_type()));
+
+        _memory_group.manage(&_input_gate_out1);
+        _fully_connected_input_gate.configure(input, lstm_params.input_to_input_weights(), lstm_params.input_gate_bias(), &_input_gate_out1, true, false);
+        _memory_group.manage(&_input_gate_out2);
+        _transpose_input_gate1.configure(lstm_params.recurrent_to_input_weights(), &_input_gate_out2);
+        _memory_group.manage(&_input_gate_out3);
+        _gemm_input_gate1.configure(output_state, &_input_gate_out2, nullptr, &_input_gate_out3, 1.f, 0.f);
+        _input_gate_out2.allocator()->allocate();
+        _memory_group.manage(&_input_gate_out4);
+        _transpose_input_gate2.configure(lstm_params.cell_to_input_weights(), &_input_gate_out4);
+        _memory_group.manage(&_input_gate_out5);
+        _gemm_input_gate2.configure(cell_state, &_input_gate_out4, nullptr, &_input_gate_out5, 1.f, 0.f);
+        _input_gate_out4.allocator()->allocate();
+        _memory_group.manage(&_input_gate_out6);
+        _accum_input_gate1.configure(&_input_gate_out1, &_input_gate_out3, &_input_gate_out6, ConvertPolicy::SATURATE);
+        _input_gate_out3.allocator()->allocate();
+        _accum_input_gate2.configure(&_input_gate_out6, &_input_gate_out5, &_input_gate_out1, ConvertPolicy::SATURATE);
+        _input_gate_out5.allocator()->allocate();
+        _input_gate_out6.allocator()->allocate();
+        _activation_input_gate.configure(&_input_gate_out1, nullptr, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LOGISTIC));
+    }
+
+    TensorShape cell_state1_shape = compute_transposed_shape(*recurrent_to_output_weights->info());
+    _cell_state_out1.allocator()->init(TensorInfo(cell_state_shape, 1, input->info()->data_type()));
+    _cell_state_out2.allocator()->init(TensorInfo(cell_state1_shape, 1, input->info()->data_type()));
+    _cell_state_out3.allocator()->init(TensorInfo(cell_state_shape, 1, input->info()->data_type()));
+    _cell_state_out4.allocator()->init(TensorInfo(cell_state_shape, 1, input->info()->data_type()));
+    _cell_state_out5.allocator()->init(TensorInfo(cell_state_shape, 1, input->info()->data_type()));
+
+    // Configure block that calculates the cell state
+    // cell_state = Clip((RixelwiseMul(input_gate, Activation(input * input_to_cell_weights + output_state * recurrent_to_cell_weights + cell_bias)) + PixelwiseMul(forget_gate, cell_state)), cell_threshold)
+    _memory_group.manage(&_cell_state_out1);
+    _fully_connected_cell_state.configure(input, input_to_cell_weights, cell_bias, &_cell_state_out1, true, false);
+    _memory_group.manage(&_cell_state_out2);
+    _transpose_cell_state1.configure(recurrent_to_cell_weights, &_cell_state_out2);
+    _memory_group.manage(&_cell_state_out3);
+    _gemm_cell_state1.configure(output_state, &_cell_state_out2, nullptr, &_cell_state_out3, 1.f, 0.f);
+    _cell_state_out2.allocator()->allocate();
+    _memory_group.manage(&_cell_state_out4);
+    _accum_cell_state1.configure(&_cell_state_out1, &_cell_state_out3, &_cell_state_out4, ConvertPolicy::SATURATE);
+    _activation_cell_state.configure(&_cell_state_out4, nullptr, activation_info);
+    _memory_group.manage(&_cell_state_out5);
+    _pixelwise_mul_cell_state1.configure(&_cell_state_out4, &_input_gate_out1, &_cell_state_out5, 1, ConvertPolicy::SATURATE, RoundingPolicy::TO_NEAREST_EVEN);
+    _input_gate_out1.allocator()->allocate();
+    _cell_state_out4.allocator()->allocate();
+    _pixelwise_mul_cell_state2.configure(&_forget_gate_out1, cell_state, &_cell_state_out3, 1, ConvertPolicy::SATURATE, RoundingPolicy::TO_NEAREST_EVEN);
+    _forget_gate_out1.allocator()->allocate();
+    _accum_cell_state2.configure(&_cell_state_out5, &_cell_state_out3, &_cell_state_out1, ConvertPolicy::SATURATE);
+    _cell_state_out3.allocator()->allocate();
+    _cell_state_out5.allocator()->allocate();
+
+    // Perform clipping
+    if(cell_threshold != 0.f)
+    {
+        _perform_cell_clipping = true;
+        _cell_clip.configure(&_cell_state_out1, nullptr, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LU_BOUNDED_RELU, -cell_threshold, cell_threshold));
+    }
+
+    TensorShape output1_shape = compute_transposed_shape(*recurrent_to_output_weights->info());
+    TensorShape output2_shape = compute_transposed_shape(*cell_bias->info());
+    TensorShape output3_shape{ 1, output_state->info()->dimension(1) };
+    _output1.allocator()->init(TensorInfo(cell_state_shape, 1, input->info()->data_type()));
+    _output2.allocator()->init(TensorInfo(output1_shape, 1, input->info()->data_type()));
+    _output3.allocator()->init(TensorInfo(cell_state_shape, 1, input->info()->data_type()));
+    _output6.allocator()->init(TensorInfo(cell_state_shape, 1, input->info()->data_type()));
+
+    // Configure block that calculates the output
+    // output_gate = Activation(input * input_to_output_weights + output_state * recurrent_to_output_weights + cell_state * cell_to_output_weights + output_gate_bias)
+    _memory_group.manage(&_output1);
+    _fully_connected_output.configure(input, input_to_output_weights, output_gate_bias, &_output1, true, false);
+    _memory_group.manage(&_output2);
+    _transpose_output1.configure(recurrent_to_output_weights, &_output2);
+    _memory_group.manage(&_output3);
+    _gemm_output1.configure(output_state, &_output2, nullptr, &_output3, 1.f, 0.f);
+    _output2.allocator()->allocate();
+    _memory_group.manage(&_output6);
+    _accum_output1.configure(&_output1, &_output3, &_output6, ConvertPolicy::SATURATE);
+    _output3.allocator()->allocate();
+    CLTensor *output_gate_out = &_output6;
+    if(lstm_params.has_peephole_opt())
+    {
+        _output4.allocator()->init(TensorInfo(output2_shape, 1, input->info()->data_type()));
+        _output5.allocator()->init(TensorInfo(output3_shape, 1, input->info()->data_type()));
+
+        _memory_group.manage(&_output4);
+        _transpose_output2.configure(lstm_params.cell_to_output_weights(), &_output4);
+        _memory_group.manage(&_output5);
+        _gemm_output2.configure(&_cell_state_out1, &_output4, nullptr, &_output5, 1.f, 0.f);
+        _accum_output2.configure(&_output6, &_output5, &_output1, ConvertPolicy::SATURATE);
+        _output6.allocator()->allocate();
+        output_gate_out = &_output1;
+
+        // Allocate intermediate buffers
+        _output4.allocator()->allocate();
+        _output5.allocator()->allocate();
+    }
+    else
+    {
+        _output1.allocator()->allocate();
+    }
+    _activation_output.configure(output_gate_out, output, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LOGISTIC));
+    output_gate_out->allocator()->allocate();
+
+    _cell_state_activation.allocator()->init(TensorInfo(cell_state_shape, 1, input->info()->data_type()));
+
+    // Configure block that calculates the output state
+    /** lstm_res = PixelwiseMul(output, Activation(cell_state))
+     *
+     *                      -- Clip(lstm_res * projection_weights + projection_bias, projection_threshold) , if there is a projection
+     *                     /
+     *  output_state =  --
+     *                     \
+     *                      -- lstm_res , otherwise
+     */
+    _memory_group.manage(&_cell_state_activation);
+    _activation_output_state.configure(&_cell_state_out1, &_cell_state_activation, activation_info);
+    _pixelwise_mul_output_state.configure(&_cell_state_activation, output, output_state, 1, ConvertPolicy::SATURATE, RoundingPolicy::TO_NEAREST_EVEN);
+    _cell_state_activation.allocator()->allocate();
+
+    if(lstm_params.has_projection())
+    {
+        _has_projection_weights = true;
+        _output_projection1.allocator()->init(TensorInfo(cell_state_shape, 1, input->info()->data_type()));
+        _memory_group.manage(&_output_projection1);
+        _fully_connected_output_state.configure(output_state, lstm_params.projection_weights(), lstm_params.projection_bias(), &_output_projection1, true, false);
+        // Perform clipping
+        if(projection_threshold != 0.f)
+        {
+            _perform_projection_clipping = true;
+            _projection_clip.configure(&_output_projection1, output_state, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LU_BOUNDED_RELU, -projection_threshold, projection_threshold));
+        }
+
+        // Allocate intermediate buffer
+        _output_projection1.allocator()->allocate();
+    }
+
+    // Copy cell state and output
+    _copy_cell_state.configure(&_cell_state_out1, cell_state);
+    _cell_state_out1.allocator()->allocate();
+    _copy_output.configure(output_state, output);
+
+    // Vector for holding the tensors to store in scratch buffer
+    std::vector<ICLTensor *> scratch_inputs;
+    if(lstm_params.has_cifg_opt())
+    {
+        scratch_inputs.emplace_back(&_input_gate_out1);
+    }
+    scratch_inputs.emplace_back(&_cell_state_out1);
+    scratch_inputs.emplace_back(forget_gate_out);
+    scratch_inputs.emplace_back(output_gate_out);
+    _concat_scratch_buffer.configure(scratch_inputs, scratch_buffer);
+}
+
+Status CLLSTMLayer::validate(const ITensorInfo *input, const ITensorInfo *input_to_forget_weights, const ITensorInfo *input_to_cell_weights, const ITensorInfo *input_to_output_weights,
+                             const ITensorInfo *recurrent_to_forget_weights, const ITensorInfo *recurrent_to_cell_weights, const ITensorInfo *recurrent_to_output_weights,
+                             const ITensorInfo *forget_gate_bias, const ITensorInfo *cell_bias, const ITensorInfo *output_gate_bias,
+                             const ITensorInfo *output_state, const ITensorInfo *cell_state, const ITensorInfo *scratch_buffer, const ITensorInfo *output,
+                             const LSTMParams<ITensorInfo> &lstm_params, const ActivationLayerInfo &activation_info, float cell_threshold, float projection_threshold)
+{
+    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, input_to_forget_weights, input_to_cell_weights, input_to_output_weights, recurrent_to_forget_weights, recurrent_to_cell_weights, recurrent_to_output_weights,
+                                        forget_gate_bias, cell_bias, output_gate_bias, output_state, cell_state);
+    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::F16, DataType::F32);
+    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, input_to_forget_weights, input_to_cell_weights, input_to_output_weights, recurrent_to_forget_weights, recurrent_to_cell_weights,
+                                                       recurrent_to_output_weights, forget_gate_bias, cell_bias, output_gate_bias, output_state, cell_state);
+    ARM_COMPUTE_RETURN_ERROR_ON(input->num_dimensions() != 2);
+    ARM_COMPUTE_RETURN_ERROR_ON(input_to_forget_weights->num_dimensions() != 2);
+    ARM_COMPUTE_RETURN_ERROR_ON(input_to_cell_weights->num_dimensions() != 2);
+    ARM_COMPUTE_RETURN_ERROR_ON(input_to_output_weights->num_dimensions() != 2);
+    ARM_COMPUTE_RETURN_ERROR_ON(recurrent_to_forget_weights->num_dimensions() != 2);
+    ARM_COMPUTE_RETURN_ERROR_ON(recurrent_to_cell_weights->num_dimensions() != 2);
+    ARM_COMPUTE_RETURN_ERROR_ON(recurrent_to_output_weights->num_dimensions() != 2);
+    ARM_COMPUTE_RETURN_ERROR_ON(forget_gate_bias->num_dimensions() != 1);
+    ARM_COMPUTE_RETURN_ERROR_ON(cell_bias->num_dimensions() != 1);
+    ARM_COMPUTE_RETURN_ERROR_ON(output_gate_bias->num_dimensions() != 1);
+    ARM_COMPUTE_RETURN_ERROR_ON(output_state->num_dimensions() != 2);
+    ARM_COMPUTE_RETURN_ERROR_ON(cell_state->num_dimensions() != 2);
+    ARM_COMPUTE_RETURN_ERROR_ON(scratch_buffer->num_dimensions() != 2);
+    ARM_COMPUTE_RETURN_ERROR_ON(output->num_dimensions() != 2);
+    ARM_COMPUTE_RETURN_ERROR_ON(cell_bias->dimension(0) * 4 != scratch_buffer->dimension(0) && cell_bias->dimension(0) * 3 != scratch_buffer->dimension(0));
+
+    if(lstm_params.has_peephole_opt())
+    {
+        ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(lstm_params.cell_to_input_weights(), lstm_params.cell_to_output_weights(), lstm_params.cell_to_forget_weights());
+        ARM_COMPUTE_RETURN_ERROR_ON(lstm_params.cell_to_input_weights()->num_dimensions() != 1);
+        ARM_COMPUTE_RETURN_ERROR_ON(lstm_params.cell_to_forget_weights()->num_dimensions() != 1);
+        ARM_COMPUTE_RETURN_ERROR_ON(lstm_params.cell_to_output_weights()->num_dimensions() != 1);
+    }
+
+    TensorShape      units_out_transposed_shape = compute_transposed_shape(*recurrent_to_output_weights);
+    TensorShape      gemmv_shape{ 1, output_state->dimension(1) };
+    TensorShape      num_units_transposed_shape = compute_transposed_shape(*forget_gate_bias);
+    const TensorInfo units_out_transposed_info  = TensorInfo(units_out_transposed_shape, 1, input->data_type());
+    const TensorInfo gemmv_shape_info           = TensorInfo(gemmv_shape, 1, input->data_type());
+    const TensorInfo num_units_transposed_info  = TensorInfo(num_units_transposed_shape, 1, input->data_type());
+
+    // Validate forget gate
+    ARM_COMPUTE_RETURN_ON_ERROR(CLFullyConnectedLayer::validate(input, input_to_forget_weights, forget_gate_bias, cell_state, true, false));
+    ARM_COMPUTE_RETURN_ON_ERROR(CLGEMM::validate(output_state, &units_out_transposed_info, nullptr, cell_state, 1.f, 0.f, GEMMInfo()));
+    ARM_COMPUTE_RETURN_ON_ERROR(CLArithmeticAdditionKernel::validate(cell_state, cell_state, cell_state, ConvertPolicy::SATURATE));
+    if(lstm_params.has_peephole_opt())
+    {
+        ARM_COMPUTE_RETURN_ON_ERROR(CLGEMM::validate(cell_state, &num_units_transposed_info, nullptr, &gemmv_shape_info, 1.f, 0.f, GEMMInfo()));
+        ARM_COMPUTE_RETURN_ON_ERROR(CLArithmeticAddition::validate(cell_state, &gemmv_shape_info, cell_state, ConvertPolicy::SATURATE));
+    }
+    ARM_COMPUTE_RETURN_ON_ERROR(CLActivationLayerKernel::validate(cell_state, cell_state, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LOGISTIC)));
+
+    // Validate input gate
+    if(!lstm_params.has_cifg_opt())
+    {
+        ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(lstm_params.input_to_input_weights(), lstm_params.recurrent_to_input_weights(), lstm_params.cell_to_input_weights(), lstm_params.input_gate_bias());
+        ARM_COMPUTE_RETURN_ERROR_ON(lstm_params.input_to_input_weights()->num_dimensions() != 2);
+        ARM_COMPUTE_RETURN_ERROR_ON(lstm_params.recurrent_to_input_weights()->num_dimensions() != 2);
+        ARM_COMPUTE_RETURN_ERROR_ON(lstm_params.cell_to_input_weights()->num_dimensions() != 1);
+        ARM_COMPUTE_RETURN_ERROR_ON(lstm_params.input_gate_bias()->num_dimensions() != 1);
+        ARM_COMPUTE_RETURN_ON_ERROR(CLFullyConnectedLayer::validate(input, lstm_params.input_to_input_weights(), lstm_params.input_gate_bias(), cell_state, true, false));
+        ARM_COMPUTE_RETURN_ON_ERROR(CLGEMM::validate(cell_state, &num_units_transposed_info, nullptr, &gemmv_shape_info, 1.f, 0.f, GEMMInfo()));
+        ARM_COMPUTE_RETURN_ON_ERROR(CLArithmeticAddition::validate(cell_state, &gemmv_shape_info, cell_state, ConvertPolicy::SATURATE));
+        ARM_COMPUTE_RETURN_ON_ERROR(CLActivationLayerKernel::validate(cell_state, nullptr, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LOGISTIC)));
+    }
+    else
+    {
+        ARM_COMPUTE_RETURN_ON_ERROR(CLArithmeticSubtractionKernel::validate(cell_state, cell_state, cell_state, ConvertPolicy::SATURATE));
+    }
+
+    // Validate cell state
+    ARM_COMPUTE_RETURN_ON_ERROR(CLFullyConnectedLayer::validate(input, input_to_cell_weights, cell_bias, cell_state, true, false));
+    ARM_COMPUTE_RETURN_ON_ERROR(CLActivationLayerKernel::validate(cell_state, nullptr, activation_info));
+    ARM_COMPUTE_RETURN_ON_ERROR(CLPixelWiseMultiplicationKernel::validate(cell_state, cell_state, cell_state, 1, ConvertPolicy::SATURATE, RoundingPolicy::TO_NEAREST_EVEN));
+
+    if(cell_threshold != 0.f)
+    {
+        ARM_COMPUTE_RETURN_ON_ERROR(CLActivationLayerKernel::validate(cell_state, nullptr, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LU_BOUNDED_RELU, -cell_threshold, cell_threshold)));
+    }
+
+    ARM_COMPUTE_RETURN_ON_ERROR(CLFullyConnectedLayer::validate(input, input_to_output_weights, output_gate_bias, cell_state, true, false));
+    if(lstm_params.has_peephole_opt())
+    {
+        ARM_COMPUTE_RETURN_ON_ERROR(CLArithmeticAddition::validate(cell_state, cell_state, cell_state, ConvertPolicy::SATURATE));
+    }
+    ARM_COMPUTE_RETURN_ON_ERROR(CLActivationLayerKernel::validate(cell_state, output, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LOGISTIC)));
+
+    // Validate output state
+    ARM_COMPUTE_RETURN_ON_ERROR(CLActivationLayerKernel::validate(cell_state, cell_state, activation_info));
+    ARM_COMPUTE_RETURN_ON_ERROR(CLPixelWiseMultiplicationKernel::validate(cell_state, output, output_state, 1, ConvertPolicy::SATURATE, RoundingPolicy::TO_NEAREST_EVEN));
+    if(lstm_params.has_projection())
+    {
+        ARM_COMPUTE_RETURN_ON_ERROR(CLFullyConnectedLayer::validate(output_state, lstm_params.projection_weights(), lstm_params.projection_bias(), cell_state, true, false));
+        if(projection_threshold != 0.f)
+        {
+            ARM_COMPUTE_RETURN_ON_ERROR(CLActivationLayerKernel::validate(cell_state, output_state, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LU_BOUNDED_RELU, -projection_threshold,
+                                                                                                                        projection_threshold)));
+        }
+    }
+
+    std::vector<TensorInfo> inputs_vector_info;
+    if(lstm_params.has_cifg_opt())
+    {
+        inputs_vector_info.emplace_back(*cell_state);
+    }
+    inputs_vector_info.emplace_back(*cell_state);
+    inputs_vector_info.emplace_back(*cell_state);
+    inputs_vector_info.emplace_back(*cell_state);
+
+    std::vector<ITensorInfo *> inputs_vector_info_raw;
+    for(auto &input : inputs_vector_info)
+    {
+        inputs_vector_info_raw.emplace_back(&input);
+    }
+
+    ARM_COMPUTE_RETURN_ON_ERROR(CLWidthConcatenateLayer::validate(inputs_vector_info_raw, scratch_buffer));
+    return Status{};
+}
+
+void CLLSTMLayer::run()
+{
+    _memory_group.acquire();
+
+    _fully_connected_forget_gate.run();
+    CLScheduler::get().enqueue(_transpose_forget_gate1);
+    _gemm_forget_gate1.run();
+    CLScheduler::get().enqueue(_accum_forget_gate1);
+
+    if(_run_peephole_opt)
+    {
+        CLScheduler::get().enqueue(_transpose_forget_gate2);
+        _gemm_forget_gate2.run();
+        _accum_forget_gate2.run();
+    }
+    CLScheduler::get().enqueue(_activation_forget_gate);
+
+    if(_run_cifg_opt)
+    {
+        _ones.map(true);
+        std::fill_n(_ones.buffer(), _ones.info()->total_size(), 1);
+        _ones.unmap();
+        CLScheduler::get().enqueue(_subtract_input_gate);
+    }
+    else
+    {
+        _fully_connected_input_gate.run();
+        CLScheduler::get().enqueue(_transpose_input_gate1);
+        _gemm_input_gate1.run();
+        CLScheduler::get().enqueue(_transpose_input_gate2);
+        _gemm_input_gate2.run();
+        CLScheduler::get().enqueue(_accum_input_gate1);
+        _accum_input_gate2.run();
+        CLScheduler::get().enqueue(_activation_input_gate);
+    }
+
+    _fully_connected_cell_state.run();
+    CLScheduler::get().enqueue(_transpose_cell_state1);
+    _gemm_cell_state1.run();
+    CLScheduler::get().enqueue(_accum_cell_state1);
+    CLScheduler::get().enqueue(_activation_cell_state);
+    CLScheduler::get().enqueue(_pixelwise_mul_cell_state1);
+    CLScheduler::get().enqueue(_pixelwise_mul_cell_state2);
+    CLScheduler::get().enqueue(_accum_cell_state2);
+
+    if(_perform_cell_clipping)
+    {
+        CLScheduler::get().enqueue(_cell_clip);
+    }
+
+    _fully_connected_output.run();
+    CLScheduler::get().enqueue(_transpose_output1);
+    _gemm_output1.run();
+    CLScheduler::get().enqueue(_accum_output1);
+    CLScheduler::get().enqueue(_pixelwise_mul_output_state);
+
+    if(_run_peephole_opt)
+    {
+        CLScheduler::get().enqueue(_transpose_output2);
+        _gemm_output2.run();
+        _accum_output2.run();
+    }
+    CLScheduler::get().enqueue(_activation_output);
+
+    CLScheduler::get().enqueue(_activation_output_state);
+    CLScheduler::get().enqueue(_pixelwise_mul_output_state);
+
+    if(_has_projection_weights)
+    {
+        _fully_connected_output_state.run();
+        if(_perform_projection_clipping)
+        {
+            CLScheduler::get().enqueue(_projection_clip);
+        }
+    }
+
+    CLScheduler::get().enqueue(_copy_cell_state);
+    CLScheduler::get().enqueue(_copy_output);
+
+    _concat_scratch_buffer.run();
+
+    _memory_group.release();
+}
\ No newline at end of file
diff --git a/tests/datasets/LSTMLayerDataset.h b/tests/datasets/LSTMLayerDataset.h
new file mode 100644
index 0000000..51802fd
--- /dev/null
+++ b/tests/datasets/LSTMLayerDataset.h
@@ -0,0 +1,171 @@
+/*
+ * Copyright (c) 2018 ARM Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#ifndef ARM_COMPUTE_TEST_LSTM_LAYER_DATASET
+#define ARM_COMPUTE_TEST_LSTM_LAYER_DATASET
+
+#include "utils/TypePrinter.h"
+
+#include "arm_compute/core/TensorShape.h"
+#include "arm_compute/core/Types.h"
+
+namespace arm_compute
+{
+namespace test
+{
+namespace datasets
+{
+class LSTMLayerDataset
+{
+public:
+    using type = std::tuple<TensorShape, TensorShape, TensorShape, TensorShape, TensorShape, TensorShape, TensorShape, ActivationLayerInfo, float, float>;
+
+    struct iterator
+    {
+        iterator(std::vector<TensorShape>::const_iterator         src_it,
+                 std::vector<TensorShape>::const_iterator         input_weights_it,
+                 std::vector<TensorShape>::const_iterator         recurrent_weights_it,
+                 std::vector<TensorShape>::const_iterator         cells_bias_it,
+                 std::vector<TensorShape>::const_iterator         output_cell_it,
+                 std::vector<TensorShape>::const_iterator         dst_it,
+                 std::vector<TensorShape>::const_iterator         scratch_it,
+                 std::vector<ActivationLayerInfo>::const_iterator infos_it,
+                 std::vector<float>::const_iterator               cell_threshold_it,
+                 std::vector<float>::const_iterator               projection_threshold_it)
+            : _src_it{ std::move(src_it) },
+              _input_weights_it{ std::move(input_weights_it) },
+              _recurrent_weights_it{ std::move(recurrent_weights_it) },
+              _cells_bias_it{ std::move(cells_bias_it) },
+              _output_cell_it{ std::move(output_cell_it) },
+              _dst_it{ std::move(dst_it) },
+              _scratch_it{ std::move(scratch_it) },
+              _infos_it{ std::move(infos_it) },
+              _cell_threshold_it{ std::move(cell_threshold_it) },
+              _projection_threshold_it{ std::move(projection_threshold_it) }
+        {
+        }
+
+        std::string description() const
+        {
+            std::stringstream description;
+            description << "In=" << *_src_it << ":";
+            description << "InputWeights=" << *_input_weights_it << ":";
+            description << "RecurrentWeights=" << *_recurrent_weights_it << ":";
+            description << "Biases=" << *_cells_bias_it << ":";
+            description << "Scratch=" << *_scratch_it << ":";
+            description << "Out=" << *_dst_it;
+            return description.str();
+        }
+
+        LSTMLayerDataset::type operator*() const
+        {
+            return std::make_tuple(*_src_it, *_input_weights_it, *_recurrent_weights_it, *_cells_bias_it, *_output_cell_it, *_dst_it, *_scratch_it, *_infos_it, *_cell_threshold_it, *_projection_threshold_it);
+        }
+
+        iterator &operator++()
+        {
+            ++_src_it;
+            ++_input_weights_it;
+            ++_recurrent_weights_it;
+            ++_cells_bias_it;
+            ++_output_cell_it;
+            ++_dst_it;
+            ++_scratch_it;
+            ++_infos_it;
+            ++_cell_threshold_it;
+            ++_projection_threshold_it;
+
+            return *this;
+        }
+
+    private:
+        std::vector<TensorShape>::const_iterator         _src_it;
+        std::vector<TensorShape>::const_iterator         _input_weights_it;
+        std::vector<TensorShape>::const_iterator         _recurrent_weights_it;
+        std::vector<TensorShape>::const_iterator         _cells_bias_it;
+        std::vector<TensorShape>::const_iterator         _output_cell_it;
+        std::vector<TensorShape>::const_iterator         _dst_it;
+        std::vector<TensorShape>::const_iterator         _scratch_it;
+        std::vector<ActivationLayerInfo>::const_iterator _infos_it;
+        std::vector<float>::const_iterator               _cell_threshold_it;
+        std::vector<float>::const_iterator               _projection_threshold_it;
+    };
+
+    iterator begin() const
+    {
+        return iterator(_src_shapes.begin(), _input_weights_shapes.begin(), _recurrent_weights_shapes.begin(), _cell_bias_shapes.begin(), _output_cell_shapes.begin(), _dst_shapes.begin(),
+                        _scratch_shapes.begin(), _infos.begin(), _cell_threshold.begin(), _projection_threshold.begin());
+    }
+
+    int size() const
+    {
+        return std::min(_src_shapes.size(), std::min(_input_weights_shapes.size(), std::min(_recurrent_weights_shapes.size(), std::min(_cell_bias_shapes.size(), std::min(_output_cell_shapes.size(),
+                                                                                            std::min(_dst_shapes.size(), std::min(_scratch_shapes.size(), std::min(_cell_threshold.size(), std::min(_projection_threshold.size(), _infos.size())))))))));
+    }
+
+    void add_config(TensorShape src, TensorShape input_weights, TensorShape recurrent_weights, TensorShape cell_bias_weights, TensorShape output_cell_state, TensorShape dst, TensorShape scratch,
+                    ActivationLayerInfo info, float cell_threshold, float projection_threshold)
+    {
+        _src_shapes.emplace_back(std::move(src));
+        _input_weights_shapes.emplace_back(std::move(input_weights));
+        _recurrent_weights_shapes.emplace_back(std::move(recurrent_weights));
+        _cell_bias_shapes.emplace_back(std::move(cell_bias_weights));
+        _output_cell_shapes.emplace_back(std::move(output_cell_state));
+        _dst_shapes.emplace_back(std::move(dst));
+        _scratch_shapes.emplace_back(std::move(scratch));
+        _infos.emplace_back(std::move(info));
+        _cell_threshold.emplace_back(std::move(cell_threshold));
+        _projection_threshold.emplace_back(std::move(projection_threshold));
+    }
+
+protected:
+    LSTMLayerDataset()                    = default;
+    LSTMLayerDataset(LSTMLayerDataset &&) = default;
+
+private:
+    std::vector<TensorShape>         _src_shapes{};
+    std::vector<TensorShape>         _input_weights_shapes{};
+    std::vector<TensorShape>         _recurrent_weights_shapes{};
+    std::vector<TensorShape>         _cell_bias_shapes{};
+    std::vector<TensorShape>         _output_cell_shapes{};
+    std::vector<TensorShape>         _dst_shapes{};
+    std::vector<TensorShape>         _scratch_shapes{};
+    std::vector<ActivationLayerInfo> _infos{};
+    std::vector<float>               _cell_threshold{};
+    std::vector<float>               _projection_threshold{};
+};
+
+class SmallLSTMLayerDataset final : public LSTMLayerDataset
+{
+public:
+    SmallLSTMLayerDataset()
+    {
+        add_config(TensorShape(8U, 2U), TensorShape(8U, 16U), TensorShape(16U, 16U), TensorShape(16U), TensorShape(16U, 2U), TensorShape(16U, 2U), TensorShape(64U, 2U), ActivationLayerInfo(), 0.05f, 0.93f);
+        add_config(TensorShape(8U, 2U), TensorShape(8U, 16U), TensorShape(16U, 16U), TensorShape(16U), TensorShape(16U, 2U), TensorShape(16U, 2U), TensorShape(48U, 2U), ActivationLayerInfo(), 0.05f, 0.93f);
+    }
+};
+
+} // namespace datasets
+} // namespace test
+} // namespace arm_compute
+#endif /* ARM_COMPUTE_TEST_LSTM_LAYER_DATASET */
diff --git a/tests/validation/CL/LSTMLayer.cpp b/tests/validation/CL/LSTMLayer.cpp
new file mode 100644
index 0000000..bd43678
--- /dev/null
+++ b/tests/validation/CL/LSTMLayer.cpp
@@ -0,0 +1,177 @@
+/*
+ * Copyright (c) 2018 ARM Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#include "arm_compute/runtime/CL/functions/CLLSTMLayer.h"
+#include "tests/CL/CLAccessor.h"
+#include "tests/PaddingCalculator.h"
+#include "tests/datasets/LSTMLayerDataset.h"
+#include "tests/framework/Asserts.h"
+#include "tests/framework/Macros.h"
+#include "tests/framework/datasets/Datasets.h"
+#include "tests/validation/Validation.h"
+#include "tests/validation/fixtures/LSTMLayerFixture.h"
+
+namespace arm_compute
+{
+namespace test
+{
+namespace validation
+{
+namespace
+{
+RelativeTolerance<float> tolerance_f32(0.001f);
+RelativeTolerance<half>  tolerance_f16(half(0.1));
+} // namespace
+
+TEST_SUITE(CL)
+TEST_SUITE(LSTMLayer)
+
+// *INDENT-OFF*
+// clang-format off
+DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip(zip(zip(zip(zip(zip(zip(zip(
+               framework::dataset::make("InputInfo", { TensorInfo(TensorShape(8U, 2U), 1, DataType::U8, 0),      // Wrong data type
+                                                       TensorInfo(TensorShape(27U, 13U, 2U), 1, DataType::F32, 0), // Wrong input size
+                                                       TensorInfo(TensorShape(8U, 2U), 1, DataType::F32, 0),     // Wrong input weights size
+                                                       TensorInfo(TensorShape(8U, 2U), 1, DataType::F32, 0),     // Wrong recurrent weights size
+                                                       TensorInfo(TensorShape(8U, 2U), 1, DataType::F32, 0),     // Wrong cell bias size
+                                                       TensorInfo(TensorShape(8U, 2U), 1, DataType::F32, 0),     // Wrong cell state size
+                                                       TensorInfo(TensorShape(8U, 2U), 1, DataType::F32, 0),     // Wrong output size
+                                                       TensorInfo(TensorShape(8U, 2U), 1, DataType::F32, 0),     // Wrong scratch size
+               }),
+               framework::dataset::make("InputWeightsInfo", { TensorInfo(TensorShape(8U, 16U), 1, DataType::F32, 0),
+                                                       TensorInfo(TensorShape(8U, 16U), 1, DataType::F32, 0),
+                                                       TensorInfo(TensorShape(27U, 11U, 2U), 1, DataType::F32, 0),
+                                                       TensorInfo(TensorShape(8U, 16U), 1, DataType::F32, 0),
+                                                       TensorInfo(TensorShape(8U, 16U), 1, DataType::F32, 0),
+                                                       TensorInfo(TensorShape(8U, 16U), 1, DataType::F32, 0),
+                                                       TensorInfo(TensorShape(8U, 16U), 1, DataType::F32, 0),
+                                                       TensorInfo(TensorShape(8U, 16U), 1, DataType::F32, 0),
+               })),
+               framework::dataset::make("RecurrentWeightsInfo", { TensorInfo(TensorShape(16U, 16U), 1, DataType::F32, 0),
+                                                                  TensorInfo(TensorShape(16U, 16U), 1, DataType::F32, 0),
+                                                                  TensorInfo(TensorShape(16U, 16U), 1, DataType::F32, 0),
+                                                                  TensorInfo(TensorShape(25U, 11U, 2U), 1, DataType::F32, 0),
+                                                                  TensorInfo(TensorShape(16U, 16U), 1, DataType::F32, 0),
+                                                                  TensorInfo(TensorShape(16U, 16U), 1, DataType::F32, 0),
+                                                                  TensorInfo(TensorShape(16U, 16U), 1, DataType::F32, 0),
+                                                                  TensorInfo(TensorShape(16U, 16U), 1, DataType::F32, 0),
+               })),
+               framework::dataset::make("CellBiasInfo", { TensorInfo(TensorShape(16U), 1, DataType::F32, 0),
+                                                      TensorInfo(TensorShape(16U), 1, DataType::F32, 0),
+                                                      TensorInfo(TensorShape(16U), 1, DataType::F32, 0),
+                                                      TensorInfo(TensorShape(16U), 1, DataType::F32, 0),
+                                                      TensorInfo(TensorShape(30U), 1, DataType::F32, 0),
+                                                      TensorInfo(TensorShape(16U), 1, DataType::F32, 0),
+                                                      TensorInfo(TensorShape(16U), 1, DataType::F32, 0),
+                                                      TensorInfo(TensorShape(16U), 1, DataType::F32, 0),
+               })),
+               framework::dataset::make("ProjectionBiasInfo", { TensorInfo(TensorShape(16U), 1, DataType::F32, 0),
+                                                      TensorInfo(TensorShape(16U), 1, DataType::F32, 0),
+                                                      TensorInfo(TensorShape(16U), 1, DataType::F32, 0),
+                                                      TensorInfo(TensorShape(16U), 1, DataType::F32, 0),
+                                                      TensorInfo(TensorShape(16U), 1, DataType::F32, 0),
+                                                      TensorInfo(TensorShape(16U), 1, DataType::F32, 0),
+                                                      TensorInfo(TensorShape(16U), 1, DataType::F32, 0),
+                                                      TensorInfo(TensorShape(16U), 1, DataType::F32, 0),
+               })),
+               framework::dataset::make("CellStateInfo", { TensorInfo(TensorShape(16U, 2U), 1, DataType::F32, 0),
+                                                        TensorInfo(TensorShape(16U, 2U), 1, DataType::F32, 0),
+                                                        TensorInfo(TensorShape(16U, 2U), 1, DataType::F32, 0),
+                                                        TensorInfo(TensorShape(16U, 2U), 1, DataType::F32, 0),
+                                                        TensorInfo(TensorShape(16U, 2U), 1, DataType::F32, 0),
+                                                        TensorInfo(TensorShape(11U), 1, DataType::F32, 0),
+                                                        TensorInfo(TensorShape(16U, 2U), 1, DataType::F32, 0),
+                                                        TensorInfo(TensorShape(16U, 2U), 1, DataType::F32, 0),
+               })),
+               framework::dataset::make("OutputInfo", { TensorInfo(TensorShape(16U, 2U), 1, DataType::F32, 0),
+                                                        TensorInfo(TensorShape(16U, 2U), 1, DataType::F32, 0),
+                                                        TensorInfo(TensorShape(16U, 2U), 1, DataType::F32, 0),
+                                                        TensorInfo(TensorShape(16U, 2U), 1, DataType::F32, 0),
+                                                        TensorInfo(TensorShape(16U, 2U), 1, DataType::F32, 0),
+                                                        TensorInfo(TensorShape(16U, 2U), 1, DataType::F32, 0),
+                                                        TensorInfo(TensorShape(11U, 13U), 1, DataType::F32, 0),
+                                                        TensorInfo(TensorShape(16U, 2U), 1, DataType::F32, 0),
+               })),
+               framework::dataset::make("ScratchInfo", { TensorInfo(TensorShape(64U, 2U), 1, DataType::F32, 0),
+                                                             TensorInfo(TensorShape(64U, 2U), 1, DataType::F32, 0),
+                                                             TensorInfo(TensorShape(64U, 2U), 1, DataType::F32, 0),
+                                                             TensorInfo(TensorShape(64U, 2U), 1, DataType::F32, 0),
+                                                             TensorInfo(TensorShape(64U, 2U), 1, DataType::F32, 0),
+                                                             TensorInfo(TensorShape(64U, 2U), 1, DataType::F32, 0),
+                                                             TensorInfo(TensorShape(64U, 2U), 1, DataType::F32, 0),
+                                                             TensorInfo(TensorShape(12U, 2U), 1, DataType::F32, 0),
+               })),
+               framework::dataset::make("ActivationInfo", { ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU),
+                                                            ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU),
+                                                            ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU),
+                                                            ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU),
+                                                            ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU),
+                                                            ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU),
+                                                            ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU),
+                                                            ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU),
+               })),
+               framework::dataset::make("Expected", { false, false, false, false, false, false, false, false })),
+               input_info, input_weights_info, recurrent_weights_info, cell_bias_info, projection_bias_info, cell_state_info, output_info, scratch_info, info, expected)
+{
+    LSTMParams<ITensorInfo> lstm_params_info;
+    lstm_params_info.set_peephole_params(&cell_bias_info, &cell_bias_info, &cell_bias_info)
+                    .set_projection_params(&recurrent_weights_info, &projection_bias_info)
+                    .set_cifg_params(&input_weights_info, &recurrent_weights_info, &cell_bias_info, &cell_bias_info);
+
+    ARM_COMPUTE_EXPECT(bool(CLLSTMLayer::validate(&input_info.clone()->set_is_resizable(false), &input_weights_info.clone()->set_is_resizable(false), &input_weights_info.clone()->set_is_resizable(false),
+                                                  &input_weights_info.clone()->set_is_resizable(false), &recurrent_weights_info.clone()->set_is_resizable(false), &recurrent_weights_info.clone()->set_is_resizable(false),
+                                                  &recurrent_weights_info.clone()->set_is_resizable(false), &cell_bias_info.clone()->set_is_resizable(false), &cell_bias_info.clone()->set_is_resizable(false),
+                                                  &cell_bias_info.clone()->set_is_resizable(false), &output_info.clone()->set_is_resizable(false), &cell_state_info.clone()->set_is_resizable(false),
+                                                  &scratch_info.clone()->set_is_resizable(false), &output_info.clone()->set_is_resizable(false), lstm_params_info, info, 0.05, 0.9)) == expected, framework::LogLevel::ERRORS);
+}
+// clang-format on
+// *INDENT-ON*
+
+template <typename T>
+using CLLSTMLayerFixture = LSTMLayerValidationFixture<CLTensor, CLAccessor, CLLSTMLayer, LSTMParams<ICLTensor>, T>;
+
+TEST_SUITE(FP32)
+FIXTURE_DATA_TEST_CASE(RunSmall, CLLSTMLayerFixture<float>, framework::DatasetMode::ALL, combine(combine(combine(datasets::SmallLSTMLayerDataset(), framework::dataset::make("DataType",
+                                                                                                                 DataType::F32)),
+                                                                                                         framework::dataset::make("ProjectionOpt", { true, false })),
+                                                                                                 framework::dataset::make("PeepholeOpt", { true, false })))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference, tolerance_f32);
+}
+TEST_SUITE_END() // FP32
+
+TEST_SUITE(FP16)
+FIXTURE_DATA_TEST_CASE(RunSmall, CLLSTMLayerFixture<half>, framework::DatasetMode::ALL, combine(combine(combine(datasets::SmallLSTMLayerDataset(), framework::dataset::make("DataType", DataType::F16)),
+                                                                                                        framework::dataset::make("ProjectionOpt", { true, false })),
+                                                                                                framework::dataset::make("PeepholeOpt", { true, false })))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference, tolerance_f16);
+}
+TEST_SUITE_END() // FP16
+TEST_SUITE_END() // LSTMLayer
+TEST_SUITE_END() // CL
+} // namespace validation
+} // namespace test
+} // namespace arm_compute
diff --git a/tests/validation/fixtures/LSTMLayerFixture.h b/tests/validation/fixtures/LSTMLayerFixture.h
new file mode 100644
index 0000000..b7e43b3
--- /dev/null
+++ b/tests/validation/fixtures/LSTMLayerFixture.h
@@ -0,0 +1,404 @@
+/*
+ * Copyright (c) 2018 ARM Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#ifndef ARM_COMPUTE_TEST_LSTM_LAYER_FIXTURE
+#define ARM_COMPUTE_TEST_LSTM_LAYER_FIXTURE
+
+#include "tests/Globals.h"
+#include "tests/framework/Asserts.h"
+#include "tests/framework/Fixture.h"
+#include "tests/validation/reference/ActivationLayer.h"
+#include "tests/validation/reference/ArithmeticAddition.h"
+#include "tests/validation/reference/ArithmeticSubtraction.h"
+#include "tests/validation/reference/FullyConnectedLayer.h"
+#include "tests/validation/reference/GEMM.h"
+#include "tests/validation/reference/PixelWiseMultiplication.h"
+#include "tests/validation/reference/Transpose.h"
+
+namespace arm_compute
+{
+namespace test
+{
+namespace validation
+{
+template <typename TensorType, typename AccessorType, typename FunctionType, typename FunctionParams, typename T>
+class LSTMLayerValidationFixture : public framework::Fixture
+{
+public:
+    template <typename...>
+    void setup(TensorShape input_shape, TensorShape input_weights_shape, TensorShape recurrent_weights_shape, TensorShape cell_bias_shape, TensorShape output_cell_shape, TensorShape output_shape,
+               TensorShape scratch_shape, ActivationLayerInfo info, float cell_threshold, float projection_threshold, DataType data_type, bool projection_opt, bool peephole_opt)
+    {
+        _target = compute_target(input_shape, input_weights_shape, recurrent_weights_shape, cell_bias_shape, output_cell_shape, output_shape, scratch_shape, info, cell_threshold, projection_threshold,
+                                 data_type, projection_opt, peephole_opt);
+        _reference = compute_reference(input_shape, input_weights_shape, recurrent_weights_shape, cell_bias_shape, output_cell_shape, output_shape, scratch_shape, info, cell_threshold, projection_threshold,
+                                       data_type, projection_opt, peephole_opt);
+    }
+
+protected:
+    template <typename U>
+    void fill(U &&tensor, int i)
+    {
+        std::uniform_real_distribution<> distribution(-1.0f, 1.0f);
+        library->fill(tensor, distribution, i);
+    }
+    template <typename U>
+    void fill_custom_val(U &&tensor, float num, int i)
+    {
+        std::uniform_real_distribution<> distribution(num, num);
+        library->fill(tensor, distribution, i);
+    }
+    TensorType compute_target(const TensorShape &input_shape, const TensorShape &input_weights_shape, const TensorShape &recurrent_weights_shape, const TensorShape &cell_bias_shape,
+                              const TensorShape &output_cell_shape, const TensorShape &output_shape, const TensorShape &scratch_shape, ActivationLayerInfo info, float cell_threshold,
+                              float projection_threshold, DataType data_type, bool projection_opt, bool peephole_opt)
+    {
+        // Create projection bias shape
+        TensorShape projection_bias_shape{};
+        projection_bias_shape.set(0, output_shape.x());
+
+        // Create tensors
+        TensorType input                 = create_tensor<TensorType>(input_shape, data_type);
+        TensorType input_to_forget_w     = create_tensor<TensorType>(input_weights_shape, data_type);
+        TensorType input_to_cell_w       = create_tensor<TensorType>(input_weights_shape, data_type);
+        TensorType input_to_output_w     = create_tensor<TensorType>(input_weights_shape, data_type);
+        TensorType recurrent_to_forget_w = create_tensor<TensorType>(recurrent_weights_shape, data_type);
+        TensorType recurrent_to_cell_w   = create_tensor<TensorType>(recurrent_weights_shape, data_type);
+        TensorType recurrent_to_output_w = create_tensor<TensorType>(recurrent_weights_shape, data_type);
+        TensorType forget_gate_bias      = create_tensor<TensorType>(cell_bias_shape, data_type);
+        TensorType cell_bias             = create_tensor<TensorType>(cell_bias_shape, data_type);
+        TensorType output_gate_bias      = create_tensor<TensorType>(cell_bias_shape, data_type);
+        TensorType output_state          = create_tensor<TensorType>(output_shape, data_type);
+        TensorType cell_state            = create_tensor<TensorType>(output_cell_shape, data_type);
+        TensorType scratch               = create_tensor<TensorType>(scratch_shape, data_type);
+        TensorType output                = create_tensor<TensorType>(output_shape, data_type);
+        TensorType input_to_input_w;
+        TensorType recurrent_to_input_w;
+        TensorType cell_to_input_w;
+        TensorType cell_to_forget_w;
+        TensorType input_gate_bias;
+        TensorType cell_to_output_w;
+        TensorType projection_w;
+        TensorType projection_bias;
+
+        bool cifg_opt = scratch_shape.x() == cell_bias_shape.x() * 4 ? true : false;
+
+        FunctionParams lstm_params;
+
+        if(!cifg_opt)
+        {
+            input_to_input_w     = create_tensor<TensorType>(input_weights_shape, data_type);
+            recurrent_to_input_w = create_tensor<TensorType>(recurrent_weights_shape, data_type);
+            cell_to_input_w      = create_tensor<TensorType>(cell_bias_shape, data_type);
+            input_gate_bias      = create_tensor<TensorType>(cell_bias_shape, data_type);
+            lstm_params.set_cifg_params(&input_to_input_w, &recurrent_to_input_w, &cell_to_input_w, &input_gate_bias);
+        }
+
+        if(peephole_opt)
+        {
+            if(cifg_opt)
+            {
+                cell_to_input_w = create_tensor<TensorType>(cell_bias_shape, data_type);
+            }
+            cell_to_forget_w = create_tensor<TensorType>(cell_bias_shape, data_type);
+            cell_to_output_w = create_tensor<TensorType>(cell_bias_shape, data_type);
+            lstm_params.set_peephole_params(&cell_to_input_w, &cell_to_forget_w, &cell_to_output_w);
+        }
+
+        if(projection_opt)
+        {
+            projection_w    = create_tensor<TensorType>(recurrent_weights_shape, data_type);
+            projection_bias = create_tensor<TensorType>(projection_bias_shape, data_type);
+            lstm_params.set_projection_params(&projection_w, &projection_bias);
+        }
+
+        // Create and configure function
+        FunctionType lstm;
+        lstm.configure(&input, &input_to_forget_w, &input_to_cell_w, &input_to_output_w, &recurrent_to_forget_w,
+                       &recurrent_to_cell_w, &recurrent_to_output_w, &forget_gate_bias, &cell_bias, &output_gate_bias, &output_state, &cell_state,
+                       &scratch, &output, lstm_params, info, cell_threshold, projection_threshold);
+
+        ARM_COMPUTE_EXPECT(input.info()->is_resizable(), framework::LogLevel::ERRORS);
+        ARM_COMPUTE_EXPECT(input_to_forget_w.info()->is_resizable(), framework::LogLevel::ERRORS);
+        ARM_COMPUTE_EXPECT(input_to_cell_w.info()->is_resizable(), framework::LogLevel::ERRORS);
+        ARM_COMPUTE_EXPECT(input_to_output_w.info()->is_resizable(), framework::LogLevel::ERRORS);
+        ARM_COMPUTE_EXPECT(recurrent_to_forget_w.info()->is_resizable(), framework::LogLevel::ERRORS);
+        ARM_COMPUTE_EXPECT(recurrent_to_cell_w.info()->is_resizable(), framework::LogLevel::ERRORS);
+        ARM_COMPUTE_EXPECT(recurrent_to_output_w.info()->is_resizable(), framework::LogLevel::ERRORS);
+        ARM_COMPUTE_EXPECT(forget_gate_bias.info()->is_resizable(), framework::LogLevel::ERRORS);
+        ARM_COMPUTE_EXPECT(cell_bias.info()->is_resizable(), framework::LogLevel::ERRORS);
+        ARM_COMPUTE_EXPECT(output_gate_bias.info()->is_resizable(), framework::LogLevel::ERRORS);
+        ARM_COMPUTE_EXPECT(output_state.info()->is_resizable(), framework::LogLevel::ERRORS);
+        ARM_COMPUTE_EXPECT(cell_state.info()->is_resizable(), framework::LogLevel::ERRORS);
+        ARM_COMPUTE_EXPECT(scratch.info()->is_resizable(), framework::LogLevel::ERRORS);
+        ARM_COMPUTE_EXPECT(output.info()->is_resizable(), framework::LogLevel::ERRORS);
+
+        // Allocate tensors
+        input.allocator()->allocate();
+        input_to_forget_w.allocator()->allocate();
+        input_to_cell_w.allocator()->allocate();
+        input_to_output_w.allocator()->allocate();
+        recurrent_to_forget_w.allocator()->allocate();
+        recurrent_to_cell_w.allocator()->allocate();
+        recurrent_to_output_w.allocator()->allocate();
+        forget_gate_bias.allocator()->allocate();
+        cell_bias.allocator()->allocate();
+        output_gate_bias.allocator()->allocate();
+        output_state.allocator()->allocate();
+        cell_state.allocator()->allocate();
+        scratch.allocator()->allocate();
+        output.allocator()->allocate();
+
+        ARM_COMPUTE_EXPECT(!input.info()->is_resizable(), framework::LogLevel::ERRORS);
+        ARM_COMPUTE_EXPECT(!input_to_forget_w.info()->is_resizable(), framework::LogLevel::ERRORS);
+        ARM_COMPUTE_EXPECT(!input_to_cell_w.info()->is_resizable(), framework::LogLevel::ERRORS);
+        ARM_COMPUTE_EXPECT(!input_to_output_w.info()->is_resizable(), framework::LogLevel::ERRORS);
+        ARM_COMPUTE_EXPECT(!recurrent_to_forget_w.info()->is_resizable(), framework::LogLevel::ERRORS);
+        ARM_COMPUTE_EXPECT(!recurrent_to_cell_w.info()->is_resizable(), framework::LogLevel::ERRORS);
+        ARM_COMPUTE_EXPECT(!recurrent_to_output_w.info()->is_resizable(), framework::LogLevel::ERRORS);
+        ARM_COMPUTE_EXPECT(!forget_gate_bias.info()->is_resizable(), framework::LogLevel::ERRORS);
+        ARM_COMPUTE_EXPECT(!cell_bias.info()->is_resizable(), framework::LogLevel::ERRORS);
+        ARM_COMPUTE_EXPECT(!output_gate_bias.info()->is_resizable(), framework::LogLevel::ERRORS);
+        ARM_COMPUTE_EXPECT(!output_state.info()->is_resizable(), framework::LogLevel::ERRORS);
+        ARM_COMPUTE_EXPECT(!cell_state.info()->is_resizable(), framework::LogLevel::ERRORS);
+        ARM_COMPUTE_EXPECT(!scratch.info()->is_resizable(), framework::LogLevel::ERRORS);
+        ARM_COMPUTE_EXPECT(!output.info()->is_resizable(), framework::LogLevel::ERRORS);
+
+        // Fill tensors
+        fill(AccessorType(input), 0);
+        fill(AccessorType(input_to_forget_w), 1);
+        fill(AccessorType(input_to_cell_w), 2);
+        fill(AccessorType(input_to_output_w), 3);
+        fill(AccessorType(recurrent_to_forget_w), 4);
+        fill(AccessorType(recurrent_to_cell_w), 5);
+        fill(AccessorType(recurrent_to_output_w), 6);
+        fill(AccessorType(forget_gate_bias), 7);
+        fill(AccessorType(cell_bias), 8);
+        fill(AccessorType(output_gate_bias), 9);
+        fill(AccessorType(output_state), 10);
+        fill(AccessorType(cell_state), 11);
+        fill(AccessorType(scratch), 12);
+
+        if(!cifg_opt)
+        {
+            ARM_COMPUTE_EXPECT(input_to_input_w.info()->is_resizable(), framework::LogLevel::ERRORS);
+            ARM_COMPUTE_EXPECT(recurrent_to_input_w.info()->is_resizable(), framework::LogLevel::ERRORS);
+            ARM_COMPUTE_EXPECT(cell_to_input_w.info()->is_resizable(), framework::LogLevel::ERRORS);
+            ARM_COMPUTE_EXPECT(input_gate_bias.info()->is_resizable(), framework::LogLevel::ERRORS);
+            input_to_input_w.allocator()->allocate();
+            recurrent_to_input_w.allocator()->allocate();
+            cell_to_input_w.allocator()->allocate();
+            input_gate_bias.allocator()->allocate();
+            ARM_COMPUTE_EXPECT(!input_to_input_w.info()->is_resizable(), framework::LogLevel::ERRORS);
+            ARM_COMPUTE_EXPECT(!recurrent_to_input_w.info()->is_resizable(), framework::LogLevel::ERRORS);
+            ARM_COMPUTE_EXPECT(!cell_to_input_w.info()->is_resizable(), framework::LogLevel::ERRORS);
+            ARM_COMPUTE_EXPECT(!input_gate_bias.info()->is_resizable(), framework::LogLevel::ERRORS);
+            fill(AccessorType(input_to_input_w), 13);
+            fill(AccessorType(recurrent_to_input_w), 14);
+            fill(AccessorType(cell_to_input_w), 15);
+            fill(AccessorType(recurrent_to_input_w), 16);
+            fill(AccessorType(input_gate_bias), 17);
+        }
+
+        if(peephole_opt)
+        {
+            if(cifg_opt)
+            {
+                ARM_COMPUTE_EXPECT(cell_to_input_w.info()->is_resizable(), framework::LogLevel::ERRORS);
+                cell_to_input_w.allocator()->allocate();
+                ARM_COMPUTE_EXPECT(!cell_to_input_w.info()->is_resizable(), framework::LogLevel::ERRORS);
+                fill(AccessorType(cell_to_input_w), 15);
+            }
+            ARM_COMPUTE_EXPECT(cell_to_forget_w.info()->is_resizable(), framework::LogLevel::ERRORS);
+            ARM_COMPUTE_EXPECT(cell_to_output_w.info()->is_resizable(), framework::LogLevel::ERRORS);
+            cell_to_forget_w.allocator()->allocate();
+            cell_to_output_w.allocator()->allocate();
+            ARM_COMPUTE_EXPECT(!cell_to_forget_w.info()->is_resizable(), framework::LogLevel::ERRORS);
+            ARM_COMPUTE_EXPECT(!cell_to_output_w.info()->is_resizable(), framework::LogLevel::ERRORS);
+            fill(AccessorType(cell_to_output_w), 18);
+        }
+
+        if(projection_opt)
+        {
+            ARM_COMPUTE_EXPECT(projection_w.info()->is_resizable(), framework::LogLevel::ERRORS);
+            ARM_COMPUTE_EXPECT(projection_bias.info()->is_resizable(), framework::LogLevel::ERRORS);
+
+            projection_w.allocator()->allocate();
+            projection_bias.allocator()->allocate();
+
+            ARM_COMPUTE_EXPECT(!projection_w.info()->is_resizable(), framework::LogLevel::ERRORS);
+            ARM_COMPUTE_EXPECT(!projection_bias.info()->is_resizable(), framework::LogLevel::ERRORS);
+
+            fill(AccessorType(projection_w), 19);
+            fill(AccessorType(projection_bias), 20);
+        }
+
+        // Compute function
+        lstm.run();
+
+        return output;
+    }
+
+    SimpleTensor<T> compute_reference(const TensorShape &input_shape, const TensorShape &input_weights_shape, const TensorShape &recurrent_weights_shape, const TensorShape &cell_bias_shape,
+                                      const TensorShape &output_cell_shape, const TensorShape &output_shape, const TensorShape &scratch_shape, ActivationLayerInfo info, float cell_threshold,
+                                      float projection_threshold, DataType data_type, bool projection_opt, bool peephole_opt)
+    {
+        // Create projection bias shape
+        TensorShape projection_bias_shape{};
+        projection_bias_shape.set(0, output_shape.x());
+
+        TensorShape     gemm_shape{ 1, output_shape.y() };
+        SimpleTensor<T> gemm_out{ gemm_shape, data_type };
+
+        // Create reference
+        SimpleTensor<T> input{ input_shape, data_type };
+        SimpleTensor<T> input_to_input_w{ input_weights_shape, data_type };
+        SimpleTensor<T> input_to_forget_w{ input_weights_shape, data_type };
+        SimpleTensor<T> input_to_cell_w{ input_weights_shape, data_type };
+        SimpleTensor<T> input_to_output_w{ input_weights_shape, data_type };
+        SimpleTensor<T> recurrent_to_input_w{ recurrent_weights_shape, data_type };
+        SimpleTensor<T> recurrent_to_forget_w{ recurrent_weights_shape, data_type };
+        SimpleTensor<T> recurrent_to_cell_w{ recurrent_weights_shape, data_type };
+        SimpleTensor<T> recurrent_to_output_w{ recurrent_weights_shape, data_type };
+        SimpleTensor<T> cell_to_input_w{ cell_bias_shape, data_type };
+        SimpleTensor<T> cell_to_forget_w{ cell_bias_shape, data_type };
+        SimpleTensor<T> cell_to_output_w{ cell_bias_shape, data_type };
+        SimpleTensor<T> input_gate_bias{ cell_bias_shape, data_type };
+        SimpleTensor<T> forget_gate_bias{ cell_bias_shape, data_type };
+        SimpleTensor<T> cell_bias{ cell_bias_shape, data_type };
+        SimpleTensor<T> output_gate_bias{ cell_bias_shape, data_type };
+        SimpleTensor<T> projection_w{ recurrent_weights_shape, data_type };
+        SimpleTensor<T> projection_bias{ projection_bias_shape, data_type };
+        SimpleTensor<T> output_state{ output_shape, data_type };
+        SimpleTensor<T> cell_state{ output_cell_shape, data_type };
+        SimpleTensor<T> scratch{ scratch_shape, data_type };
+        SimpleTensor<T> output{ output_shape, data_type };
+
+        // Fill reference
+        fill(input, 0);
+        fill(input_to_forget_w, 1);
+        fill(input_to_cell_w, 2);
+        fill(input_to_output_w, 3);
+        fill(recurrent_to_forget_w, 4);
+        fill(recurrent_to_cell_w, 5);
+        fill(recurrent_to_output_w, 6);
+        fill(forget_gate_bias, 7);
+        fill(cell_bias, 8);
+        fill(output_gate_bias, 9);
+        fill(output_state, 10);
+        fill(cell_state, 11);
+        fill(scratch, 12);
+        fill(input_to_input_w, 13);
+        fill(recurrent_to_input_w, 14);
+        fill(cell_to_input_w, 15);
+        fill(recurrent_to_input_w, 16);
+        fill(input_gate_bias, 17);
+        fill(cell_to_output_w, 18);
+        fill(projection_w, 19);
+        fill(projection_bias, 20);
+
+        bool cifg_opt = scratch_shape.x() == cell_bias_shape.x() * 4 ? true : false;
+
+        // Compute forget_gate
+        SimpleTensor<T> fully_connected_forget = reference::fully_connected_layer(input, input_to_forget_w, forget_gate_bias, output_cell_shape);
+        SimpleTensor<T> transposed_weights     = reference::transpose(recurrent_to_forget_w);
+        SimpleTensor<T> gemm                   = reference::gemm(output_state, transposed_weights, cell_state, 1.f, 0.f);
+        SimpleTensor<T> forget_gate            = reference::arithmetic_addition(fully_connected_forget, gemm, data_type, ConvertPolicy::SATURATE);
+
+        if(peephole_opt)
+        {
+            transposed_weights = reference::transpose(cell_to_forget_w);
+            gemm               = reference::gemm(cell_state, transposed_weights, gemm_out, 1.f, 0.f);
+            forget_gate        = reference::arithmetic_addition(forget_gate, gemm, data_type, ConvertPolicy::SATURATE);
+        }
+
+        forget_gate = reference::activation_layer(forget_gate, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LOGISTIC));
+
+        // Compute input_gate
+        SimpleTensor<T> input_gate;
+        if(cifg_opt)
+        {
+            SimpleTensor<T> ones{ cell_bias_shape, data_type };
+            fill_custom_val(ones, 1.f, 0);
+            input_gate = reference::arithmetic_subtraction<T, T, T>(ones, forget_gate, data_type, ConvertPolicy::SATURATE);
+        }
+        else
+        {
+            SimpleTensor<T> fully_connected_input = reference::fully_connected_layer(input, input_to_input_w, input_gate_bias, output_cell_shape);
+            transposed_weights                    = reference::transpose(recurrent_to_input_w);
+            gemm                                  = reference::gemm(output_state, transposed_weights, cell_state, 1.f, 0.f);
+            input_gate                            = reference::arithmetic_addition(fully_connected_input, gemm, data_type, ConvertPolicy::SATURATE);
+            transposed_weights                    = reference::transpose(cell_to_input_w);
+            gemm                                  = reference::gemm(cell_state, transposed_weights, gemm_out, 1.f, 0.f);
+            input_gate                            = reference::arithmetic_addition(input_gate, gemm, data_type, ConvertPolicy::SATURATE);
+            input_gate                            = reference::activation_layer(input_gate, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LOGISTIC));
+        }
+
+        // Compute cell_state
+        SimpleTensor<T> fully_connected_cell_state = reference::fully_connected_layer(input, input_to_cell_w, cell_bias, output_cell_shape);
+        transposed_weights                         = reference::transpose(recurrent_to_cell_w);
+        gemm                                       = reference::gemm(output_state, transposed_weights, cell_state, 1.f, 0.f);
+        SimpleTensor<T> pixelwise_mul              = reference::pixel_wise_multiplication(cell_state, forget_gate, 1, ConvertPolicy::SATURATE, RoundingPolicy::TO_NEAREST_EVEN);
+        cell_state                                 = reference::arithmetic_addition(fully_connected_cell_state, gemm, data_type, ConvertPolicy::SATURATE);
+        cell_state                                 = reference::activation_layer(cell_state, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LOGISTIC));
+        cell_state                                 = reference::pixel_wise_multiplication(cell_state, input_gate, 1, ConvertPolicy::SATURATE, RoundingPolicy::TO_NEAREST_EVEN);
+        cell_state                                 = reference::arithmetic_addition(cell_state, pixelwise_mul, data_type, ConvertPolicy::SATURATE);
+        if(cell_threshold != 0.f)
+        {
+            cell_state = reference::activation_layer(cell_state, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LU_BOUNDED_RELU, -cell_threshold, cell_threshold));
+        }
+
+        // Compute output
+        SimpleTensor<T> fully_connected_output = reference::fully_connected_layer(input, input_to_output_w, output_gate_bias, output_cell_shape);
+        transposed_weights                     = reference::transpose(recurrent_to_output_w);
+        gemm                                   = reference::gemm(output_state, transposed_weights, cell_state, 1.f, 0.f);
+        output                                 = reference::arithmetic_addition(fully_connected_output, gemm, data_type, ConvertPolicy::SATURATE);
+        if(peephole_opt)
+        {
+            transposed_weights = reference::transpose(cell_to_output_w);
+            gemm               = reference::gemm(cell_state, transposed_weights, gemm_out, 1.f, 0.f);
+            output             = reference::arithmetic_addition(output, gemm, data_type, ConvertPolicy::SATURATE);
+        }
+        output = reference::activation_layer(output, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LOGISTIC));
+
+        // Compute output state
+        SimpleTensor<T> cell_state_activation = reference::activation_layer(cell_state, info);
+        output_state                          = reference::pixel_wise_multiplication(output, cell_state_activation, 1, ConvertPolicy::SATURATE, RoundingPolicy::TO_NEAREST_EVEN);
+
+        if(projection_opt)
+        {
+            SimpleTensor<T> fully_connected_projection = reference::fully_connected_layer(output_state, projection_w, projection_bias, output_cell_shape);
+            if(projection_threshold != 0.f)
+            {
+                output_state = reference::activation_layer(fully_connected_projection, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LU_BOUNDED_RELU, -projection_threshold, projection_threshold));
+            }
+        }
+        return output_state;
+    }
+
+    TensorType      _target{};
+    SimpleTensor<T> _reference{};
+};
+} // namespace validation
+} // namespace test
+} // namespace arm_compute
+#endif /* ARM_COMPUTE_TEST_LSTM_LAYER_FIXTURE */