COMPMID-1973: Implement FFTConvolutionLayer on NEON
Change-Id: I2e667c0411bda0164a616ffe44473a78de6752c9
Signed-off-by: giuros01 <giuseppe.rossini@arm.com>
Reviewed-on: https://review.mlplatform.org/c/1066
Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
diff --git a/arm_compute/core/NEON/kernels/NEFFTDigitReverseKernel.h b/arm_compute/core/NEON/kernels/NEFFTDigitReverseKernel.h
index 84d55fd..0090c8c 100644
--- a/arm_compute/core/NEON/kernels/NEFFTDigitReverseKernel.h
+++ b/arm_compute/core/NEON/kernels/NEFFTDigitReverseKernel.h
@@ -24,6 +24,7 @@
#ifndef __ARM_COMPUTE_NEFFTDIGITREVERSEKERNEL_H__
#define __ARM_COMPUTE_NEFFTDIGITREVERSEKERNEL_H__
+#include "arm_compute/core/KernelDescriptors.h"
#include "arm_compute/core/NEON/INEKernel.h"
namespace arm_compute
@@ -53,31 +54,40 @@
~NEFFTDigitReverseKernel() = default;
/** Set the input and output tensors.
*
- * @param[in] input Source tensor. Data types supported: F32.
- * @param[out] output Destination tensor. Data type supported: same as @p input
+ * @param[in] input Source tensor. Data types supported: F32. Number of channels supported: 1 (real tensor) or 2 (complex tensor).
+ * @param[out] output Destination tensor. Data type supported: same as @p input. Number of channels supported: 2 (complex tensor).
* @param[in] idx Digit reverse index tensor. Data type supported: U32
- * @param[in] axis Axis to perform digit reverse on.
+ * @param[in] config Kernel configuration.
*/
- void configure(const ITensor *input, ITensor *output, const ITensor *idx, unsigned int axis);
+ void configure(const ITensor *input, ITensor *output, const ITensor *idx, const FFTDigitReverseKernelInfo &config);
+
/** Static function to check if given info will lead to a valid configuration of @ref NEFFTDigitReverseKernel
*
- * @param[in] input Source tensor info. Data types supported: F32.
- * @param[in] output Destination tensor info. Data type supported: same as @p input
+ * @param[in] input Source tensor info. Data types supported: F32. Number of channels supported: 1 (real tensor) or 2 (complex tensor).
+ * @param[in] output Destination tensor info. Data type supported: same as @p input. Number of channels supported: 2 (complex tensor).
* @param[in] idx Digit reverse index tensor info. Data type supported: U32
- * @param[in] axis Axis to perform digit reverse on.
+ * @param[in] config Kernel configuration
*
* @return a status
*/
- static Status validate(const ITensorInfo *input, const ITensorInfo *output, const ITensorInfo *idx, unsigned int axis);
+ static Status validate(const ITensorInfo *input, const ITensorInfo *output, const ITensorInfo *idx, const FFTDigitReverseKernelInfo &config);
// Inherited methods overridden:
void run(const Window &window, const ThreadInfo &info) override;
private:
- const ITensor *_input;
- ITensor *_output;
- const ITensor *_idx;
- unsigned int _axis;
+ using NEFFTDigitReverseKernelFunctionPtr = void (NEFFTDigitReverseKernel::*)(const Window &window);
+
+ template <bool is_input_complex, bool is_conj>
+ void digit_reverse_kernel_axis_0(const Window &window);
+
+ template <bool is_input_complex, bool is_conj>
+ void digit_reverse_kernel_axis_1(const Window &window);
+
+ NEFFTDigitReverseKernelFunctionPtr _func;
+ const ITensor *_input;
+ ITensor *_output;
+ const ITensor *_idx;
};
} // namespace arm_compute
#endif /*__ARM_COMPUTE_NEFFTDIGITREVERSEKERNEL_H__ */
diff --git a/arm_compute/core/NEON/kernels/NEFFTRadixStageKernel.h b/arm_compute/core/NEON/kernels/NEFFTRadixStageKernel.h
index 8498d3c..2210980 100644
--- a/arm_compute/core/NEON/kernels/NEFFTRadixStageKernel.h
+++ b/arm_compute/core/NEON/kernels/NEFFTRadixStageKernel.h
@@ -59,15 +59,15 @@
*
* @note If the output tensor is nullptr, the FFT will be performed in-place
*
- * @param[in,out] input Source tensor. Data types supported: F32.
- * @param[out] output Destination tensor. Data type supported: same as @p input
+ * @param[in,out] input Source tensor. Data types supported: F32. Number of channels supported: 2 (complex tensor).
+ * @param[out] output Destination tensor. Data type supported: same as @p input. Number of channels supported: same as @p input.
* @param[in] config FFT descriptor metadata.
*/
void configure(ITensor *input, ITensor *output, const FFTRadixStageKernelInfo &config);
/** Static function to check if given info will lead to a valid configuration of @ref NEFFTRadixStageKernel
*
- * @param[in] input Source tensor info. Data types supported: F32.
- * @param[in] output Destination tensor info. Data type supported: same as @p input
+ * @param[in] input Source tensor info. Data types supported: F32. Number of channels supported: 2 (complex tensor).
+ * @param[in] output Destination tensor info. Data type supported: same as @p input. Number of channels supported: same as @p input.
* @param[in] config FFT descriptor metadata.
*
* @return a status
diff --git a/arm_compute/core/NEON/kernels/NEFFTScaleKernel.h b/arm_compute/core/NEON/kernels/NEFFTScaleKernel.h
index 5a19af7..51e6d5a 100644
--- a/arm_compute/core/NEON/kernels/NEFFTScaleKernel.h
+++ b/arm_compute/core/NEON/kernels/NEFFTScaleKernel.h
@@ -55,15 +55,15 @@
~NEFFTScaleKernel() = default;
/** Set the input and output tensors.
*
- * @param[in,out] input Source tensor. Data types supported: F32.
- * @param[out] output Destination tensor. Data type supported: same as @p input
+ * @param[in,out] input Source tensor. Data types supported: F32. Number of channels supported: 2 (complex tensor).
+ * @param[out] output Destination tensor. Data type supported: same as @p input. Number of channels supported: 1 (real tensor) or 2 (complex tensor).
* @param[in] config Kernel configuration
*/
void configure(ITensor *input, ITensor *output, const FFTScaleKernelInfo &config);
/** Static function to check if given info will lead to a valid configuration of @ref NEFFTScaleKernel
*
- * @param[in] input Source tensor info. Data types supported: F32.
- * @param[in] output Destination tensor info. Data type supported: same as @p input
+ * @param[in] input Source tensor info. Data types supported: F32. Number of channels supported: 2 (complex tensor).
+ * @param[in] output Destination tensor info. Data type supported: same as @p input. Number of channels supported: 1 (real tensor) or 2 (complex tensor).
* @param[in] config Kernel configuration
*
* @return a status
diff --git a/arm_compute/core/NEON/kernels/NEPixelWiseMultiplicationKernel.h b/arm_compute/core/NEON/kernels/NEPixelWiseMultiplicationKernel.h
index 2a8e36b..cbb961f 100644
--- a/arm_compute/core/NEON/kernels/NEPixelWiseMultiplicationKernel.h
+++ b/arm_compute/core/NEON/kernels/NEPixelWiseMultiplicationKernel.h
@@ -128,5 +128,51 @@
float _scale;
int _scale_exponent;
};
+
+/** Interface for the complex pixelwise multiplication kernel. */
+class NEComplexPixelWiseMultiplicationKernel : public INEKernel
+{
+public:
+ const char *name() const override
+ {
+ return "NEComplexPixelWiseMultiplicationKernel";
+ }
+ /** Default constructor.*/
+ NEComplexPixelWiseMultiplicationKernel();
+ /** Prevent instances of this class from being copied (As this class contains pointers) */
+ NEComplexPixelWiseMultiplicationKernel(const NEComplexPixelWiseMultiplicationKernel &) = delete;
+ /** Prevent instances of this class from being copied (As this class contains pointers) */
+ NEComplexPixelWiseMultiplicationKernel &operator=(const NEComplexPixelWiseMultiplicationKernel &) = delete;
+ /** Allow instances of this class to be moved */
+ NEComplexPixelWiseMultiplicationKernel(NEComplexPixelWiseMultiplicationKernel &&) = default;
+ /** Allow instances of this class to be moved */
+ NEComplexPixelWiseMultiplicationKernel &operator=(NEComplexPixelWiseMultiplicationKernel &&) = default;
+ /** Initialise the kernel's input, output and border mode.
+ *
+ * @param[in] input1 An input tensor. Data types supported: F32. Number of channels supported: 2 (complex tensor).
+ * @param[in] input2 An input tensor. Data types supported: same as @p input1. Number of channels supported: same as @p input1.
+ * @param[out] output The output tensor, Data types supported: same as @p input1. Number of channels supported: same as @p input1.
+ */
+ void configure(const ITensor *input1, const ITensor *input2, ITensor *output);
+ /** Static function to check if given info will lead to a valid configuration of @ref NEComplexPixelWiseMultiplicationKernel
+ *
+ * @param[in] input1 An input tensor info. Data types supported: F32. Number of channels supported: 2 (complex tensor).
+ * @param[in] input2 An input tensor info. Data types supported: same as @p input1. Number of channels supported: same as @p input1.
+ * @param[in] output The output tensor info. Data types supported: same as @p input1. Number of channels supported: same as @p input1.
+ *
+ * @return a status
+ */
+ static Status validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output);
+
+ // Inherited methods overridden:
+ void run(const Window &window, const ThreadInfo &info) override;
+ BorderSize border_size() const override;
+
+private:
+ const ITensor *_input1;
+ const ITensor *_input2;
+ ITensor *_output;
+};
+
} // namespace arm_compute
#endif /*__ARM_COMPUTE_NEPIXELWISEMULTIPLICATIONKERNEL_H__ */