COMPMID-1067 NEON RNN FP32 / FP16
Change-Id: I440df2b2af512fd874651baf28428caa6f8e0b41
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/134433
Tested-by: Jenkins <bsgcomp@arm.com>
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
diff --git a/arm_compute/runtime/NEON/NEFunctions.h b/arm_compute/runtime/NEON/NEFunctions.h
index bd4097c..b83675b 100644
--- a/arm_compute/runtime/NEON/NEFunctions.h
+++ b/arm_compute/runtime/NEON/NEFunctions.h
@@ -95,6 +95,7 @@
#include "arm_compute/runtime/NEON/functions/NEPixelWiseMultiplication.h"
#include "arm_compute/runtime/NEON/functions/NEPoolingLayer.h"
#include "arm_compute/runtime/NEON/functions/NEQuantizationLayer.h"
+#include "arm_compute/runtime/NEON/functions/NERNNLayer.h"
#include "arm_compute/runtime/NEON/functions/NEROIPoolingLayer.h"
#include "arm_compute/runtime/NEON/functions/NEReductionOperation.h"
#include "arm_compute/runtime/NEON/functions/NERemap.h"
diff --git a/arm_compute/runtime/NEON/functions/NEFullyConnectedLayer.h b/arm_compute/runtime/NEON/functions/NEFullyConnectedLayer.h
index 2739f5e..42c9e2d 100644
--- a/arm_compute/runtime/NEON/functions/NEFullyConnectedLayer.h
+++ b/arm_compute/runtime/NEON/functions/NEFullyConnectedLayer.h
@@ -104,7 +104,7 @@
NEFullyConnectedLayer &operator=(NEFullyConnectedLayer &&) = default;
/** Set the input and output tensors.
*
- * @param[in] input Source tensor. Data type supported: QS8/QS16/F32.
+ * @param[in] input Source tensor. Data type supported: QS8/QS16/F16/F32.
* @param[in] weights Weights tensor. The weights must be 2 dimensional. Data type supported: Same as @p input.
* @param[in] biases Bias tensor. Can be nullptr. Data type supported:Same as @p input.
* @param[out] output Destination tensor. Data type supported: Same as @p input.
diff --git a/arm_compute/runtime/NEON/functions/NERNNLayer.h b/arm_compute/runtime/NEON/functions/NERNNLayer.h
new file mode 100644
index 0000000..f1398eb
--- /dev/null
+++ b/arm_compute/runtime/NEON/functions/NERNNLayer.h
@@ -0,0 +1,96 @@
+/*
+ * 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_NERNNLAYER_H__
+#define __ARM_COMPUTE_NERNNLAYER_H__
+
+#include "arm_compute/core/NEON/kernels/NEActivationLayerKernel.h"
+#include "arm_compute/core/NEON/kernels/NEArithmeticAdditionKernel.h"
+#include "arm_compute/runtime/NEON/INESimpleFunction.h"
+
+#include "arm_compute/core/Types.h"
+#include "arm_compute/runtime/NEON/functions/NEFullyConnectedLayer.h"
+#include "arm_compute/runtime/NEON/functions/NEGEMM.h"
+
+namespace arm_compute
+{
+// Forward declarations
+class ITensor;
+
+/** Basic function to run @ref NERNNLayer */
+class NERNNLayer : public IFunction
+{
+public:
+ /** Default constructor */
+ NERNNLayer(std::shared_ptr<IMemoryManager> memory_manager = nullptr);
+ /** Prevent instances of this class from being copied (As this class contains pointers) */
+ NERNNLayer(const NERNNLayer &) = delete;
+ /** Default move constructor */
+ NERNNLayer(NERNNLayer &&) = default;
+ /** Prevent instances of this class from being copied (As this class contains pointers) */
+ NERNNLayer &operator=(const NERNNLayer &) = delete;
+ /** Default move assignment operator */
+ NERNNLayer &operator=(NERNNLayer &&) = default;
+ /** Initialize the function
+ *
+ * @param[in] input Input is a 2-D tensor of shape [input_size, batch_size]. Data types supported: F16/F32
+ * @param[in] weights Weights tensor of shape [input_size, num_units] that multiplies the input. Data types supported: Same as @p input
+ * @param[in] recurrent_weights Weights tensor of shape [num_units, num_units] that multiplies the current 'state'. Data types supported: Same as @p input
+ * @param[in] bias Bias vector of shape [num_units]. Data types supported: Same as @p input
+ * @param[out] output Output tensor of shape [num_units, batch_size]. Data types supported: Same as @p input
+ * @param[in,out] hidden_state Output tensor of shape [num_units, batch_size]. Data types supported: Same as @p input
+ * @param[in] info Activation layer parameter.
+ */
+ void configure(const ITensor *input, const ITensor *weights, const ITensor *recurrent_weights, const ITensor *bias, ITensor *hidden_state, ITensor *output, ActivationLayerInfo &info);
+ /** Initialize the function
+ *
+ * @param[in] input Input is a 2-D tensor of shape [input_size, batch_size]. Data types supported: F16/F32
+ * @param[in] weights Weights tensor of shape [input_size, num_units] that multiplies the input. Data types supported: Same as @p input
+ * @param[in] recurrent_weights Weights tensor of shape [num_units, num_units] that multiplies the current 'state'. Data types supported: Same as @p input
+ * @param[in] bias Bias vector of shape [num_units]. Data types supported: Same as @p input
+ * @param[in] output Output tensor of shape [num_units, batch_size]. Data types supported: Same as @p input
+ * @param[in] hidden_state Output tensor of shape [num_units, batch_size]. Data types supported: Same as @p input
+ * @param[in] info Activation layer parameter.
+ *
+ * @return a status
+ */
+ static Status validate(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *recurrent_weights, const ITensorInfo *bias, const ITensorInfo *hidden_state, const ITensorInfo *output,
+ const ActivationLayerInfo &info);
+
+ // Inherited methods overridden:
+ void run() override;
+
+private:
+ MemoryGroup _memory_group;
+ NEGEMM _gemm_state_f;
+ NEArithmeticAdditionKernel _add_kernel;
+ NEActivationLayerKernel _activation_kernel;
+ NEFullyConnectedLayer _fully_connected_kernel;
+ Tensor _fully_connected_out;
+ Tensor _gemm_output;
+ Tensor _add_output;
+ ITensor *_hidden_state;
+ ITensor *_output;
+};
+} // namespace arm_compute
+#endif /* __ARM_COMPUTE_NERNNLAYER_H__ */