COMPMID-1756: NEON: Implement RSqrt, Exp
Change-Id: I6b140b8868b04f7d3032a51831a80829e8e1560e
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/165590
Tested-by: bsgcomp <bsgcomp@arm.com>
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
Reviewed-on: https://review.mlplatform.org/470
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
diff --git a/arm_compute/core/NEON/NEKernels.h b/arm_compute/core/NEON/NEKernels.h
index 95a3e05..6c804e7 100644
--- a/arm_compute/core/NEON/NEKernels.h
+++ b/arm_compute/core/NEON/NEKernels.h
@@ -58,6 +58,7 @@
#include "arm_compute/core/NEON/kernels/NEDirectConvolutionLayerKernel.h"
#include "arm_compute/core/NEON/kernels/NEDirectConvolutionLayerOutputStageKernel.h"
#include "arm_compute/core/NEON/kernels/NEElementwiseOperationKernel.h"
+#include "arm_compute/core/NEON/kernels/NEElementwiseUnaryKernel.h"
#include "arm_compute/core/NEON/kernels/NEErodeKernel.h"
#include "arm_compute/core/NEON/kernels/NEFastCornersKernel.h"
#include "arm_compute/core/NEON/kernels/NEFillArrayKernel.h"
diff --git a/arm_compute/core/NEON/kernels/NEElementwiseUnaryKernel.h b/arm_compute/core/NEON/kernels/NEElementwiseUnaryKernel.h
new file mode 100644
index 0000000..f632b1a
--- /dev/null
+++ b/arm_compute/core/NEON/kernels/NEElementwiseUnaryKernel.h
@@ -0,0 +1,100 @@
+/*
+ * Copyright (c) 2018-2019 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_NEELEMENTWISEUNARYKERNEL_H__
+#define __ARM_COMPUTE_NEELEMENTWISEUNARYKERNEL_H__
+
+#include "arm_compute/core/NEON/INEKernel.h"
+#include "arm_compute/core/Types.h"
+
+namespace arm_compute
+{
+class ITensor;
+
+/** Interface for an element-wise unary operation kernel
+ *
+ * Element-wise operation is computed by:
+ * @f[ output(x) = OP(input(x))@f]
+ *
+ */
+class NEElementwiseUnaryKernel : public INEKernel
+{
+public:
+ const char *name() const override
+ {
+ return "NEElementwiseUnaryKernel";
+ }
+ /** Default constructor */
+ NEElementwiseUnaryKernel();
+ /** Prevent instances of this class from being copied (As this class contains pointers) */
+ NEElementwiseUnaryKernel(const NEElementwiseUnaryKernel &) = delete;
+ /** Prevent instances of this class from being copied (As this class contains pointers) */
+ NEElementwiseUnaryKernel &operator=(const NEElementwiseUnaryKernel &) = delete;
+ /** Allow instances of this class to be moved */
+ NEElementwiseUnaryKernel(NEElementwiseUnaryKernel &&) = default;
+ /** Allow instances of this class to be moved */
+ NEElementwiseUnaryKernel &operator=(NEElementwiseUnaryKernel &&) = default;
+ /** Default destructor */
+ ~NEElementwiseUnaryKernel() = default;
+
+ /** Static function to check if given info will lead to a valid configuration of @ref NEElementwiseUnaryKernel
+ *
+ * @param[in] op Arithmetic operation to be executed.
+ * @param[in] input First tensor input. Data types supported: F16/F32.
+ * @param[in] output Output tensor. Data types supported: Same as @p input.
+ */
+ void configure(ElementWiseUnary op, const ITensor *input, ITensor *output);
+
+ /** Static function to check if given info will lead to a valid configuration of @ref NEElementwiseUnaryKernel
+ *
+ * @param[in] op Arithmetic operation to be executed.
+ * @param[in] input First tensor input info. Data types supported: F16/F32.
+ * @param[in] output Output tensor info. Data types supported: Same as @p input.
+ *
+ * @return a Status
+ */
+ static Status validate(ElementWiseUnary op, const ITensorInfo *input, const ITensorInfo *output);
+
+ // Inherited methods overridden:
+ void run(const Window &window, const ThreadInfo &info) override;
+
+ /** Common signature for all the specialised arithmetic functions
+ *
+ * @param[in] input An input tensor. Data types supported: F16/F32.
+ * @param[out] output The output tensor. Data types supported: F16/F32.
+ * @param[in] window Region on which to execute the kernel.
+ */
+ using ElementwiseUnaryFunction = void(const ITensor *input, ITensor *output, const Window &window);
+
+protected:
+ // Inherited methods overridden:
+ static Status validate_arguments(const ITensorInfo &input, const ITensorInfo &output);
+
+ /** Function to use for the particular tensor types passed to configure() */
+ std::function<void(const ITensor *input, ITensor *output, const Window &window)> _function;
+
+ const ITensor *_input;
+ ITensor *_output;
+};
+} // namespace arm_compute
+#endif /* __ARM_COMPUTE_NEELEMENTWISEUNARYKERNEL_H__ */
diff --git a/arm_compute/core/NEON/kernels/NEFuseBatchNormalizationKernel.h b/arm_compute/core/NEON/kernels/NEFuseBatchNormalizationKernel.h
index 4d5855e..97d6a9f 100644
--- a/arm_compute/core/NEON/kernels/NEFuseBatchNormalizationKernel.h
+++ b/arm_compute/core/NEON/kernels/NEFuseBatchNormalizationKernel.h
@@ -1,5 +1,5 @@
/*
- * Copyright (c) 2018 ARM Limited.
+ * Copyright (c) 2018-2019 ARM Limited.
*
* SPDX-License-Identifier: MIT
*
@@ -14,9 +14,9 @@
* copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INNEUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * 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 NEAIM, DAMAGES OR OTHER
+ * 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.
diff --git a/arm_compute/core/NEON/wrapper/intrinsics/exp.h b/arm_compute/core/NEON/wrapper/intrinsics/exp.h
index 85aa9c7..c538ba8 100644
--- a/arm_compute/core/NEON/wrapper/intrinsics/exp.h
+++ b/arm_compute/core/NEON/wrapper/intrinsics/exp.h
@@ -1,5 +1,5 @@
/*
- * Copyright (c) 2018 ARM Limited.
+ * Copyright (c) 2018-2019 ARM Limited.
*
* SPDX-License-Identifier: MIT
*
@@ -42,6 +42,7 @@
VEXPQ_IMPL(float16x8_t, f16)
#endif // __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
#undef VEXPQ_IMPL
+
} // namespace wrapper
} // namespace arm_compute
#endif /* __ARM_COMPUTE_WRAPPER_EXP_H__ */
diff --git a/arm_compute/core/NEON/wrapper/intrinsics/invsqrt.h b/arm_compute/core/NEON/wrapper/intrinsics/invsqrt.h
index 0bbf49b..1068cf5 100644
--- a/arm_compute/core/NEON/wrapper/intrinsics/invsqrt.h
+++ b/arm_compute/core/NEON/wrapper/intrinsics/invsqrt.h
@@ -1,5 +1,5 @@
/*
- * Copyright (c) 2018 ARM Limited.
+ * Copyright (c) 2018-2019 ARM Limited.
*
* SPDX-License-Identifier: MIT
*
diff --git a/arm_compute/runtime/CL/functions/CLElementwiseOperations.h b/arm_compute/runtime/CL/functions/CLElementwiseOperations.h
index 4a0911e..1740bb0 100644
--- a/arm_compute/runtime/CL/functions/CLElementwiseOperations.h
+++ b/arm_compute/runtime/CL/functions/CLElementwiseOperations.h
@@ -1,5 +1,5 @@
/*
- * Copyright (c) 2018 ARM Limited.
+ * Copyright (c) 2018-2019 ARM Limited.
*
* SPDX-License-Identifier: MIT
*
@@ -17,7 +17,7 @@
* 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, ARI SING FROM,
+ * 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.
*/
diff --git a/arm_compute/runtime/NEON/NEFunctions.h b/arm_compute/runtime/NEON/NEFunctions.h
index 53afb98..4e198c5 100644
--- a/arm_compute/runtime/NEON/NEFunctions.h
+++ b/arm_compute/runtime/NEON/NEFunctions.h
@@ -57,6 +57,7 @@
#include "arm_compute/runtime/NEON/functions/NEDilate.h"
#include "arm_compute/runtime/NEON/functions/NEDirectConvolutionLayer.h"
#include "arm_compute/runtime/NEON/functions/NEElementwiseOperations.h"
+#include "arm_compute/runtime/NEON/functions/NEElementwiseUnaryLayer.h"
#include "arm_compute/runtime/NEON/functions/NEEqualizeHistogram.h"
#include "arm_compute/runtime/NEON/functions/NEErode.h"
#include "arm_compute/runtime/NEON/functions/NEFastCorners.h"
diff --git a/arm_compute/runtime/NEON/functions/NEElementwiseUnaryLayer.h b/arm_compute/runtime/NEON/functions/NEElementwiseUnaryLayer.h
new file mode 100644
index 0000000..7682790
--- /dev/null
+++ b/arm_compute/runtime/NEON/functions/NEElementwiseUnaryLayer.h
@@ -0,0 +1,73 @@
+/*
+ * Copyright (c) 2018-2019 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_NEELEMENTWISEUNARYLAYER_H__
+#define __ARM_COMPUTE_NEELEMENTWISEUNARYLAYER_H__
+
+#include "arm_compute/runtime/NEON/INESimpleFunction.h"
+
+namespace arm_compute
+{
+class ITensor;
+
+/** Basic function to perform inverse square root on an input tensor. */
+class NERsqrtLayer : public INESimpleFunction
+{
+public:
+ /** Initialize the function
+ *
+ * @param[in] input Input tensor. Data types supported: F16/F32.
+ * @param[out] output Output tensor. Data types supported: same as @p input.
+ */
+ void configure(const ITensor *input, ITensor *output);
+ /** Static function to check if given info will lead to a valid configuration of @ref NERsqrtLayer
+ *
+ * @param[in] input First tensor input info. Data types supported: F16/F32.
+ * @param[in] output Output tensor info. Data types supported: Same as @p input.
+ *
+ * @return a status
+ */
+ static Status validate(const ITensorInfo *input, const ITensorInfo *output);
+};
+
+/** Basic function to perform exponential on an input tensor. */
+class NEExpLayer : public INESimpleFunction
+{
+public:
+ /** Initialize the function
+ *
+ * @param[in] input Input tensor. Data types supported: F16/F32.
+ * @param[out] output Output tensor. Data types supported: same as @p input.
+ */
+ void configure(const ITensor *input, ITensor *output);
+ /** Static function to check if given info will lead to a valid configuration of @ref NEExpLayer
+ *
+ * @param[in] input First tensor input info. Data types supported: F16/F32.
+ * @param[in] output Output tensor info. Data types supported: Same as @p input.
+ *
+ * @return a status
+ */
+ static Status validate(const ITensorInfo *input, const ITensorInfo *output);
+};
+} // namespace arm_compute
+#endif /* __ARM_COMPUTE_NEELEMENTWISEUNARYLAYER_H__ */
diff --git a/arm_compute/runtime/NEON/functions/NEFuseBatchNormalization.h b/arm_compute/runtime/NEON/functions/NEFuseBatchNormalization.h
index 82ed329..5e62869 100644
--- a/arm_compute/runtime/NEON/functions/NEFuseBatchNormalization.h
+++ b/arm_compute/runtime/NEON/functions/NEFuseBatchNormalization.h
@@ -1,5 +1,5 @@
/*
- * Copyright (c) 2018 ARM Limited.
+ * Copyright (c) 2018-2019 ARM Limited.
*
* SPDX-License-Identifier: MIT
*
@@ -14,9 +14,9 @@
* copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INNEUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * 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 NEAIM, DAMAGES OR OTHER
+ * 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.