Add the HIIR downsampler

src/external/hiir/Downsampler2xFpu.h

@@ -0,0 +1,115 @@
+/*****************************************************************************
+
+        Downsampler2xFpu.h
+        Author: Laurent de Soras, 2005
+
+Downsamples by a factor 2 the input signal, using FPU.
+
+Template parameters:
+	- NC: number of coefficients, > 0
+
+--- Legal stuff ---
+
+This program is free software. It comes without any warranty, to
+the extent permitted by applicable law. You can redistribute it
+and/or modify it under the terms of the Do What The Fuck You Want
+To Public License, Version 2, as published by Sam Hocevar. See
+http://sam.zoy.org/wtfpl/COPYING for more details.
+
+*Tab=3***********************************************************************/
+
+
+
+#if ! defined (hiir_Downsampler2xFpu_HEADER_INCLUDED)
+#define hiir_Downsampler2xFpu_HEADER_INCLUDED
+
+#if defined (_MSC_VER)
+	#pragma once
+	#pragma warning (4 : 4250) // "Inherits via dominance."
+#endif
+
+
+
+/*\\\ INCLUDE FILES \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/
+
+#include "hiir/def.h"
+
+#include <array>
+
+
+
+namespace hiir
+{
+
+
+
+template <int NC>
+class Downsampler2xFpu
+{
+
+	static_assert ((NC > 0), "Number of coefficient must be positive.");
+
+/*\\\ PUBLIC \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/
+
+public:
+
+	enum {         NBR_COEFS = NC };
+
+	               Downsampler2xFpu ();
+
+	void           set_coefs (const double coef_arr []);
+
+	hiir_FORCEINLINE float
+	               process_sample (const float in_ptr [2]);
+	void           process_block (float out_ptr [], const float in_ptr [], long nbr_spl);
+
+	hiir_FORCEINLINE void
+	               process_sample_split (float &low, float &high, const float in_ptr [2]);
+	void           process_block_split (float out_l_ptr [], float out_h_ptr [], const float in_ptr [], long nbr_spl);
+
+	void           clear_buffers ();
+
+
+
+/*\\\ PROTECTED \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/
+
+protected:
+
+
+
+/*\\\ PRIVATE \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/
+
+private:
+
+	typedef	std::array <float, NBR_COEFS>	HyperGluar;
+
+	HyperGluar     _coef;
+	HyperGluar     _x;
+	HyperGluar     _y;
+
+
+
+/*\\\ FORBIDDEN MEMBER FUNCTIONS \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/
+
+private:
+
+	bool           operator == (const Downsampler2xFpu <NC> &other);
+	bool           operator != (const Downsampler2xFpu <NC> &other);
+
+}; // class Downsampler2xFpu
+
+
+
+}  // namespace hiir
+
+
+
+#include "hiir/Downsampler2xFpu.hpp"
+
+
+
+#endif   // hiir_Downsampler2xFpu_HEADER_INCLUDED
+
+
+
+/*\\\ EOF \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/

src/external/hiir/Downsampler2xFpu.hpp

@@ -0,0 +1,303 @@
+/*****************************************************************************
+
+        Downsampler2xFpu.hpp
+        Author: Laurent de Soras, 2005
+
+--- Legal stuff ---
+
+This program is free software. It comes without any warranty, to
+the extent permitted by applicable law. You can redistribute it
+and/or modify it under the terms of the Do What The Fuck You Want
+To Public License, Version 2, as published by Sam Hocevar. See
+http://sam.zoy.org/wtfpl/COPYING for more details.
+
+*Tab=3***********************************************************************/
+
+
+
+#if defined (hiir_Downsampler2xFpu_CURRENT_CODEHEADER)
+	#error Recursive inclusion of Downsampler2xFpu code header.
+#endif
+#define hiir_Downsampler2xFpu_CURRENT_CODEHEADER
+
+#if ! defined (hiir_Downsampler2xFpu_CODEHEADER_INCLUDED)
+#define hiir_Downsampler2xFpu_CODEHEADER_INCLUDED
+
+
+
+/*\\\ INCLUDE FILES \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/
+
+#include "hiir/StageProcFpu.h"
+
+#include <cassert>
+
+
+
+namespace hiir
+{
+
+
+
+/*\\\ PUBLIC \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/
+
+
+
+/*
+==============================================================================
+Name: ctor
+Throws: Nothing
+==============================================================================
+*/
+
+template <int NC>
+Downsampler2xFpu <NC>::Downsampler2xFpu ()
+:	_coef ()
+,	_x ()
+,	_y ()
+{
+	for (int i = 0; i < NBR_COEFS; ++i)
+	{
+		_coef [i] = 0;
+	}
+	clear_buffers ();
+}
+
+
+
+/*
+==============================================================================
+Name: set_coefs
+Description:
+   Sets filter coefficients. Generate them with the PolyphaseIir2Designer
+   class.
+   Call this function before doing any processing.
+Input parameters:
+	- coef_arr: Array of coefficients. There should be as many coefficients as
+      mentioned in the class template parameter.
+Throws: Nothing
+==============================================================================
+*/
+
+template <int NC>
+void	Downsampler2xFpu <NC>::set_coefs (const double coef_arr [])
+{
+	assert (coef_arr != 0);
+
+	for (int i = 0; i < NBR_COEFS; ++i)
+	{
+		_coef [i] = float (coef_arr [i]);
+	}
+}
+
+
+
+/*
+==============================================================================
+Name: process_sample
+Description:
+   Downsamples (x2) one pair of samples, to generate one output sample.
+Input parameters:
+	- in_ptr: pointer on the two samples to decimate
+Returns: Samplerate-reduced sample.
+Throws: Nothing
+==============================================================================
+*/
+
+template <int NC>
+float	Downsampler2xFpu <NC>::process_sample (const float in_ptr [2])
+{
+	assert (in_ptr != 0);
+
+	float          spl_0 (in_ptr [1]);
+	float          spl_1 (in_ptr [0]);
+
+	#if defined (_MSC_VER)
+		#pragma inline_depth (255)
+	#endif   // _MSC_VER
+
+	StageProcFpu <NBR_COEFS>::process_sample_pos (
+		NBR_COEFS,
+		spl_0,
+		spl_1,
+		&_coef [0],
+		&_x [0],
+		&_y [0]
+	);
+
+	return 0.5f * (spl_0 + spl_1);
+}
+
+
+
+/*
+==============================================================================
+Name: process_block
+Description:
+   Downsamples (x2) a block of samples.
+	Input and output blocks may overlap, see assert() for details.
+Input parameters:
+	- in_ptr: Input array, containing nbr_spl * 2 samples.
+	- nbr_spl: Number of samples to output, > 0
+Output parameters:
+	- out_ptr: Array for the output samples, capacity: nbr_spl samples.
+Throws: Nothing
+==============================================================================
+*/
+
+template <int NC>
+void	Downsampler2xFpu <NC>::process_block (float out_ptr [], const float in_ptr [], long nbr_spl)
+{
+	assert (in_ptr != 0);
+	assert (out_ptr != 0);
+	assert (out_ptr <= in_ptr || out_ptr >= in_ptr + nbr_spl * 2);
+	assert (nbr_spl > 0);
+
+	long           pos = 0;
+	do
+	{
+		out_ptr [pos] = process_sample (&in_ptr [pos * 2]);
+		++pos;
+	}
+	while (pos < nbr_spl);
+}
+
+
+
+/*
+==============================================================================
+Name: process_sample_split
+Description:
+   Split (spectrum-wise) in half a pair of samples. The lower part of the
+   spectrum is a classic downsampling, equivalent to the output of
+   process_sample().
+   The higher part is the complementary signal: original filter response
+   is flipped from left to right, becoming a high-pass filter with the same
+   cutoff frequency. This signal is then critically sampled (decimation by 2),
+   flipping the spectrum: Fs/4...Fs/2 becomes Fs/4...0.
+Input parameters:
+	- in_ptr: pointer on the pair of input samples
+Output parameters:
+	- low: output sample, lower part of the spectrum (downsampling)
+	- high: output sample, higher part of the spectrum.
+Throws: Nothing
+==============================================================================
+*/
+
+template <int NC>
+void	Downsampler2xFpu <NC>::process_sample_split (float &low, float &high, const float in_ptr [2])
+{
+	assert (in_ptr != 0);
+
+	float          spl_0 = in_ptr [1];
+	float          spl_1 = in_ptr [0];
+
+	#if defined (_MSC_VER)
+		#pragma inline_depth (255)
+	#endif	// _MSC_VER
+
+	StageProcFpu <NBR_COEFS>::process_sample_pos (
+		NBR_COEFS,
+		spl_0,
+		spl_1,
+		&_coef [0],
+		&_x [0],
+		&_y [0]
+	);
+
+	low  = (spl_0 + spl_1) * 0.5f;
+	high =  spl_0 - low; // (spl_0 - spl_1) * 0.5f;
+}
+
+
+
+/*
+==============================================================================
+Name: process_block_split
+Description:
+   Split (spectrum-wise) in half a block of samples. The lower part of the
+   spectrum is a classic downsampling, equivalent to the output of
+   process_block().
+   The higher part is the complementary signal: original filter response
+   is flipped from left to right, becoming a high-pass filter with the same
+   cutoff frequency. This signal is then critically sampled (decimation by 2),
+   flipping the spectrum: Fs/4...Fs/2 becomes Fs/4...0.
+	Input and output blocks may overlap, see assert() for details.
+Input parameters:
+	- in_ptr: Input array, containing nbr_spl * 2 samples.
+	- nbr_spl: Number of samples for each output, > 0
+Output parameters:
+	- out_l_ptr: Array for the output samples, lower part of the spectrum
+      (downsampling). Capacity: nbr_spl samples.
+	- out_h_ptr: Array for the output samples, higher part of the spectrum.
+      Capacity: nbr_spl samples.
+Throws: Nothing
+==============================================================================
+*/
+
+template <int NC>
+void	Downsampler2xFpu <NC>::process_block_split (float out_l_ptr [], float out_h_ptr [], const float in_ptr [], long nbr_spl)
+{
+	assert (in_ptr != 0);
+	assert (out_l_ptr != 0);
+	assert (out_l_ptr <= in_ptr || out_l_ptr >= in_ptr + nbr_spl * 2);
+	assert (out_h_ptr != 0);
+	assert (out_h_ptr <= in_ptr || out_h_ptr >= in_ptr + nbr_spl * 2);
+	assert (out_h_ptr != out_l_ptr);
+	assert (nbr_spl > 0);
+
+	long           pos = 0;
+	do
+	{
+		process_sample_split (
+			out_l_ptr [pos],
+			out_h_ptr [pos],
+			&in_ptr [pos * 2]
+		);
+		++pos;
+	}
+	while (pos < nbr_spl);
+}
+
+
+
+/*
+==============================================================================
+Name: clear_buffers
+Description:
+	Clears filter memory, as if it processed silence since an infinite amount
+	of time.
+Throws: Nothing
+==============================================================================
+*/
+
+template <int NC>
+void	Downsampler2xFpu <NC>::clear_buffers ()
+{
+	for (int i = 0; i < NBR_COEFS; ++i)
+	{
+		_x [i] = 0;
+		_y [i] = 0;
+	}
+}
+
+
+
+/*\\\ PROTECTED \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/
+
+
+
+/*\\\ PRIVATE \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/
+
+
+
+}  // namespace hiir
+
+
+
+#endif   // hiir_Downsampler2xFpu_CODEHEADER_INCLUDED
+
+#undef hiir_Downsampler2xFpu_CURRENT_CODEHEADER
+
+
+
+/*\\\ EOF \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/

src/external/hiir/Downsampler2xNeon.h

@@ -0,0 +1,126 @@
+/*****************************************************************************
+
+        Downsampler2xNeon.h
+        Author: Laurent de Soras, 2016
+
+Downsamples by a factor 2 the input signal, using NEON instruction set.
+
+This object must be aligned on a 16-byte boundary!
+
+If the number of coefficients is 2 or 3 modulo 4, the output is delayed from
+1 sample, compared to the theoretical formula (or FPU implementation).
+
+Template parameters:
+	- NC: number of coefficients, > 0
+
+--- Legal stuff ---
+
+This program is free software. It comes without any warranty, to
+the extent permitted by applicable law. You can redistribute it
+and/or modify it under the terms of the Do What The Fuck You Want
+To Public License, Version 2, as published by Sam Hocevar. See
+http://sam.zoy.org/wtfpl/COPYING for more details.
+
+*Tab=3***********************************************************************/
+
+
+
+#pragma once
+#if ! defined (hiir_Downsampler2xNeon_HEADER_INCLUDED)
+#define hiir_Downsampler2xNeon_HEADER_INCLUDED
+
+#if defined (_MSC_VER)
+	#pragma warning (4 : 4250)
+#endif
+
+
+
+/*\\\ INCLUDE FILES \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/
+
+#include "hiir/def.h"
+#include "hiir/StageDataNeon.h"
+
+#include <array>
+
+
+
+namespace hiir
+{
+
+
+
+template <int NC>
+class Downsampler2xNeon
+{
+
+	static_assert ((NC > 0), "Number of coefficient must be positive.");
+
+/*\\\ PUBLIC \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/
+
+public:
+
+	enum {         NBR_COEFS = NC	};
+
+	               Downsampler2xNeon ();
+	               Downsampler2xNeon (const Downsampler2xNeon &other) = default;
+
+	Downsampler2xNeon &
+	               operator = (const Downsampler2xNeon &other)        = default;
+
+	void           set_coefs (const double coef_arr []);
+
+	hiir_FORCEINLINE float
+	               process_sample (const float in_ptr [2]);
+	void           process_block (float out_ptr [], const float in_ptr [], long nbr_spl);
+
+	hiir_FORCEINLINE void
+	               process_sample_split (float &low, float &high, const float in_ptr [2]);
+	void           process_block_split (float out_l_ptr [], float out_h_ptr [], const float in_ptr [], long nbr_spl);
+
+	void           clear_buffers ();
+
+
+
+/*\\\ PROTECTED \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/
+
+protected:
+
+
+
+/*\\\ PRIVATE \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/
+
+private:
+
+	enum {         STAGE_WIDTH	= 4 };
+	enum {         NBR_STAGES  = (NBR_COEFS + STAGE_WIDTH - 1) / STAGE_WIDTH	 };
+
+	typedef	std::array <StageDataNeon, NBR_STAGES + 1>	Filter;  // Stage 0 contains only input memory
+
+	Filter         _filter; // Should be the first member (thus easier to align)
+
+
+
+/*\\\ FORBIDDEN MEMBER FUNCTIONS \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/
+
+private:
+
+	bool           operator == (const Downsampler2xNeon <NC> &other) const = delete;
+	bool           operator != (const Downsampler2xNeon <NC> &other) const = delete;
+
+}; // class Downsampler2xNeon
+
+
+
+}  // namespace hiir
+
+
+
+#include "hiir/Downsampler2xNeon.hpp"
+
+
+
+#endif   // hiir_Downsampler2xNeon_HEADER_INCLUDED
+
+
+
+/*\\\ EOF \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/

src/external/hiir/Downsampler2xNeon.hpp

@@ -0,0 +1,299 @@
+/*****************************************************************************
+
+        Downsampler2xNeon.hpp
+        Author: Laurent de Soras, 2016
+
+--- Legal stuff ---
+
+This program is free software. It comes without any warranty, to
+the extent permitted by applicable law. You can redistribute it
+and/or modify it under the terms of the Do What The Fuck You Want
+To Public License, Version 2, as published by Sam Hocevar. See
+http://sam.zoy.org/wtfpl/COPYING for more details.
+
+*Tab=3***********************************************************************/
+
+
+
+#if ! defined (hiir_Downsampler2xNeon_CODEHEADER_INCLUDED)
+#define hiir_Downsampler2xNeon_CODEHEADER_INCLUDED
+
+
+
+/*\\\ INCLUDE FILES \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/
+
+#include "hiir/StageProcNeon.h"
+
+#include <arm_neon.h>
+
+#include <cassert>
+
+
+
+namespace hiir
+{
+
+
+
+/*\\\ PUBLIC \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/
+
+
+
+/*
+==============================================================================
+Name: ctor
+Throws: Nothing
+==============================================================================
+*/
+
+template <int NC>
+Downsampler2xNeon <NC>::Downsampler2xNeon ()
+:	_filter ()
+{
+	for (int i = 0; i < NBR_STAGES + 1; ++i)
+	{
+		_filter [i]._mem4 = vdupq_n_f32 (0);
+	}
+	if ((NBR_COEFS & 1) != 0)
+	{
+		const int      pos = (NBR_COEFS ^ 1) & (STAGE_WIDTH - 1);
+		_filter [NBR_STAGES]._coef [pos] = 1;
+	}
+
+	clear_buffers ();
+}
+
+
+
+/*
+==============================================================================
+Name: set_coefs
+Description:
+   Sets filter coefficients. Generate them with the PolyphaseIir2Designer
+   class.
+   Call this function before doing any processing.
+Input parameters:
+	- coef_arr: Array of coefficients. There should be as many coefficients as
+      mentioned in the class template parameter.
+Throws: Nothing
+==============================================================================
+*/
+
+template <int NC>
+void	Downsampler2xNeon <NC>::set_coefs (const double coef_arr [])
+{
+	assert (coef_arr != 0);
+
+	for (int i = 0; i < NBR_COEFS; ++i)
+	{
+		const int      stage = (i / STAGE_WIDTH) + 1;
+		const int      pos   = (i ^ 1) & (STAGE_WIDTH - 1);
+		_filter [stage]._coef [pos] = float (coef_arr [i]);
+	}
+}
+
+
+
+/*
+==============================================================================
+Name: process_sample
+Description:
+   Downsamples (x2) one pair of samples, to generate one output sample.
+Input parameters:
+	- in_ptr: pointer on the two samples to decimate
+Returns: Samplerate-reduced sample.
+Throws: Nothing
+==============================================================================
+*/
+
+template <int NC>
+float	Downsampler2xNeon <NC>::process_sample (const float in_ptr [2])
+{
+	assert (in_ptr != 0);
+
+	// Combines two input samples and two mid-processing data
+	const float32x2_t spl_in  = vreinterpret_f32_u8 (
+		vld1_u8 (reinterpret_cast <const uint8_t *> (in_ptr))
+	);
+	const float32x2_t spl_mid = vget_low_f32 (_filter [NBR_STAGES]._mem4);
+	float32x4_t       y       = vcombine_f32 (spl_in, spl_mid);
+	float32x4_t       mem     = _filter [0]._mem4;
+
+	// Processes each stage
+	StageProcNeon <NBR_STAGES>::process_sample_pos (&_filter [0], y, mem);
+	_filter [NBR_STAGES]._mem4 = y;
+
+	// Averages both paths and outputs the result
+	const float       out_0  = vgetq_lane_f32 (y, 3);
+	const float       out_1  = vgetq_lane_f32 (y, 2);
+	const float       out    = (out_0 + out_1) * 0.5f;
+
+	return out;
+}
+
+
+
+/*
+==============================================================================
+Name: process_block
+Description:
+   Downsamples (x2) a block of samples.
+	Input and output blocks may overlap, see assert() for details.
+Input parameters:
+	- in_ptr: Input array, containing nbr_spl * 2 samples.
+	- nbr_spl: Number of samples to output, > 0
+Output parameters:
+	- out_ptr: Array for the output samples, capacity: nbr_spl samples.
+Throws: Nothing
+==============================================================================
+*/
+
+template <int NC>
+void	Downsampler2xNeon <NC>::process_block (float out_ptr [], const float in_ptr [], long nbr_spl)
+{
+	assert (in_ptr != 0);
+	assert (out_ptr != 0);
+	assert (out_ptr <= in_ptr || out_ptr >= in_ptr + nbr_spl * 2);
+	assert (nbr_spl > 0);
+
+	long           pos = 0;
+	do
+	{
+		out_ptr [pos] = process_sample (in_ptr + pos * 2);
+		++ pos;
+	}
+	while (pos < nbr_spl);
+}
+
+
+
+/*
+==============================================================================
+Name: process_sample_split
+Description:
+   Split (spectrum-wise) in half a pair of samples. The lower part of the
+   spectrum is a classic downsampling, equivalent to the output of
+   process_sample().
+   The higher part is the complementary signal: original filter response
+   is flipped from left to right, becoming a high-pass filter with the same
+   cutoff frequency. This signal is then critically sampled (decimation by 2),
+   flipping the spectrum: Fs/4...Fs/2 becomes Fs/4...0.
+Input parameters:
+	- in_ptr: pointer on the pair of input samples
+Output parameters:
+	- low: output sample, lower part of the spectrum (downsampling)
+	- high: output sample, higher part of the spectrum.
+Throws: Nothing
+==============================================================================
+*/
+
+template <int NC>
+void	Downsampler2xNeon <NC>::process_sample_split (float &low, float &high, const float in_ptr [2])
+{
+	assert (in_ptr != 0);
+
+	// Combines two input samples and two mid-processing data
+	const float32x2_t spl_in  = vreinterpret_f32_u8 (
+		vld1_u8 (reinterpret_cast <const uint8_t *> (in_ptr))
+	);
+	const float32x2_t spl_mid = vget_low_f32 (_filter [NBR_STAGES]._mem4);
+	float32x4_t       y       = vcombine_f32 (spl_in, spl_mid);
+	float32x4_t       mem     = _filter [0]._mem4;
+
+	// Processes each stage
+	StageProcNeon <NBR_STAGES>::process_sample_pos (&_filter [0], y, mem);
+	_filter [NBR_STAGES]._mem4 = y;
+
+	// Outputs the result
+	const float       out_0  = vgetq_lane_f32 (y, 3);
+	const float       out_1  = vgetq_lane_f32 (y, 2);
+	low  = (out_0 + out_1) * 0.5f;
+	high =  out_0 - low;
+}
+
+
+
+/*
+==============================================================================
+Name: process_block_split
+Description:
+   Split (spectrum-wise) in half a block of samples. The lower part of the
+   spectrum is a classic downsampling, equivalent to the output of
+   process_block().
+   The higher part is the complementary signal: original filter response
+   is flipped from left to right, becoming a high-pass filter with the same
+   cutoff frequency. This signal is then critically sampled (decimation by 2),
+   flipping the spectrum: Fs/4...Fs/2 becomes Fs/4...0.
+	Input and output blocks may overlap, see assert() for details.
+Input parameters:
+	- in_ptr: Input array, containing nbr_spl * 2 samples.
+	- nbr_spl: Number of samples for each output, > 0
+Output parameters:
+	- out_l_ptr: Array for the output samples, lower part of the spectrum
+      (downsampling). Capacity: nbr_spl samples.
+	- out_h_ptr: Array for the output samples, higher part of the spectrum.
+      Capacity: nbr_spl samples.
+Throws: Nothing
+==============================================================================
+*/
+
+template <int NC>
+void	Downsampler2xNeon <NC>::process_block_split (float out_l_ptr [], float out_h_ptr [], const float in_ptr [], long nbr_spl)
+{
+	assert (in_ptr != 0);
+	assert (out_l_ptr != 0);
+	assert (out_l_ptr <= in_ptr || out_l_ptr >= in_ptr + nbr_spl * 2);
+	assert (out_h_ptr != 0);
+	assert (out_h_ptr <= in_ptr || out_h_ptr >= in_ptr + nbr_spl * 2);
+	assert (out_h_ptr != out_l_ptr);
+	assert (nbr_spl > 0);
+
+	long           pos = 0;
+	do
+	{
+		process_sample_split (out_l_ptr [pos], out_h_ptr [pos], in_ptr + pos * 2);
+		++ pos;
+	}
+	while (pos < nbr_spl);
+}
+
+
+
+/*
+==============================================================================
+Name: clear_buffers
+Description:
+	Clears filter memory, as if it processed silence since an infinite amount
+	of time.
+Throws: Nothing
+==============================================================================
+*/
+
+template <int NC>
+void	Downsampler2xNeon <NC>::clear_buffers ()
+{
+	for (int i = 0; i < NBR_STAGES + 1; ++i)
+	{
+		_filter [i]._mem4 = vdupq_n_f32 (0);
+	}
+}
+
+
+
+/*\\\ PROTECTED \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/
+
+
+
+/*\\\ PRIVATE \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/
+
+
+
+}  // namespace hiir
+
+
+
+#endif   // hiir_Downsampler2xNeon_CODEHEADER_INCLUDED
+
+
+
+/*\\\ EOF \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/

src/external/hiir/Downsampler2xSse.h

@@ -0,0 +1,126 @@
+/*****************************************************************************
+
+        Downsampler2xSse.h
+        Author: Laurent de Soras, 2005
+
+Downsamples by a factor 2 the input signal, using SSE instruction set.
+
+This object must be aligned on a 16-byte boundary!
+
+If the number of coefficients is 2 or 3 modulo 4, the output is delayed from
+1 sample, compared to the theoretical formula (or FPU implementation).
+
+Template parameters:
+	- NC: number of coefficients, > 0
+
+--- Legal stuff ---
+
+This program is free software. It comes without any warranty, to
+the extent permitted by applicable law. You can redistribute it
+and/or modify it under the terms of the Do What The Fuck You Want
+To Public License, Version 2, as published by Sam Hocevar. See
+http://sam.zoy.org/wtfpl/COPYING for more details.
+
+*Tab=3***********************************************************************/
+
+
+
+#if ! defined (hiir_Downsampler2xSse_HEADER_INCLUDED)
+#define hiir_Downsampler2xSse_HEADER_INCLUDED
+
+#if defined (_MSC_VER)
+	#pragma once
+	#pragma warning (4 : 4250) // "Inherits via dominance."
+#endif
+
+
+
+/*\\\ INCLUDE FILES \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/
+
+#include "hiir/def.h"
+#include "hiir/StageDataSse.h"
+
+#include <array>
+
+
+
+namespace hiir
+{
+
+
+
+template <int NC>
+class Downsampler2xSse
+{
+
+	static_assert ((NC > 0), "Number of coefficient must be positive.");
+
+/*\\\ PUBLIC \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/
+
+public:
+
+	enum {         NBR_COEFS = NC };
+
+	               Downsampler2xSse ();
+	               Downsampler2xSse (const Downsampler2xSse &other) = default;
+
+	Downsampler2xSse &
+	               operator = (const Downsampler2xSse &other)       = default;
+
+	void           set_coefs (const double coef_arr []);
+
+	hiir_FORCEINLINE float
+	               process_sample (const float in_ptr [2]);
+	void           process_block (float out_ptr [], const float in_ptr [], long nbr_spl);
+
+	hiir_FORCEINLINE void
+	               process_sample_split (float &low, float &high, const float in_ptr [2]);
+	void           process_block_split (float out_l_ptr [], float out_h_ptr [], const float in_ptr [], long nbr_spl);
+
+	void           clear_buffers ();
+
+
+
+/*\\\ PROTECTED \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/
+
+protected:
+
+
+
+/*\\\ PRIVATE \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/
+
+private:
+
+	enum {         STAGE_WIDTH	= 4 };
+	enum {         NBR_STAGES  = (NBR_COEFS + STAGE_WIDTH - 1) / STAGE_WIDTH	 };
+
+	typedef	std::array <StageDataSse, NBR_STAGES + 1>	Filter;  // Stage 0 contains only input memory
+
+	Filter         _filter; // Should be the first member (thus easier to align)
+
+
+
+/*\\\ FORBIDDEN MEMBER FUNCTIONS \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/
+
+private:
+
+	bool           operator == (const Downsampler2xSse <NC> &other) = delete;
+	bool           operator != (const Downsampler2xSse <NC> &other) = delete;
+
+}; // class Downsampler2xSse
+
+
+
+}  // namespace hiir
+
+
+
+#include "hiir/Downsampler2xSse.hpp"
+
+
+
+#endif   // hiir_Downsampler2xSse_HEADER_INCLUDED
+
+
+
+/*\\\ EOF \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/

src/external/hiir/Downsampler2xSse.hpp

@@ -0,0 +1,322 @@
+/*****************************************************************************
+
+        Downsampler2xSse.hpp
+        Author: Laurent de Soras, 2005
+
+--- Legal stuff ---
+
+This program is free software. It comes without any warranty, to
+the extent permitted by applicable law. You can redistribute it
+and/or modify it under the terms of the Do What The Fuck You Want
+To Public License, Version 2, as published by Sam Hocevar. See
+http://sam.zoy.org/wtfpl/COPYING for more details.
+
+*Tab=3***********************************************************************/
+
+
+
+#if defined (hiir_Downsampler2xSse_CURRENT_CODEHEADER)
+	#error Recursive inclusion of Downsampler2xSse code header.
+#endif
+#define hiir_Downsampler2xSse_CURRENT_CODEHEADER
+
+#if ! defined (hiir_Downsampler2xSse_CODEHEADER_INCLUDED)
+#define hiir_Downsampler2xSse_CODEHEADER_INCLUDED
+
+
+
+/*\\\ INCLUDE FILES \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/
+
+#include "hiir/StageProcSse.h"
+
+#include <xmmintrin.h>
+
+#include <cassert>
+
+
+
+namespace hiir
+{
+
+
+
+/*\\\ PUBLIC \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/
+
+
+
+/*
+==============================================================================
+Name: ctor
+Throws: Nothing
+==============================================================================
+*/
+
+template <int NC>
+Downsampler2xSse <NC>::Downsampler2xSse ()
+:	_filter ()
+{
+	for (int i = 0; i < NBR_STAGES + 1; ++i)
+	{
+		_filter [i]._coef [0] = 0;
+		_filter [i]._coef [1] = 0;
+		_filter [i]._coef [2] = 0;
+		_filter [i]._coef [3] = 0;
+	}
+	if ((NBR_COEFS & 1) != 0)
+	{
+		const int      pos = (NBR_COEFS ^ 1) & (STAGE_WIDTH - 1);
+		_filter [NBR_STAGES]._coef [pos] = 1;
+	}
+
+	clear_buffers ();
+}
+
+
+
+/*
+==============================================================================
+Name: set_coefs
+Description:
+   Sets filter coefficients. Generate them with the PolyphaseIir2Designer
+   class.
+   Call this function before doing any processing.
+Input parameters:
+	- coef_arr: Array of coefficients. There should be as many coefficients as
+      mentioned in the class template parameter.
+Throws: Nothing
+==============================================================================
+*/
+
+template <int NC>
+void	Downsampler2xSse <NC>::set_coefs (const double coef_arr [])
+{
+	assert (coef_arr != 0);
+
+	for (int i = 0; i < NBR_COEFS; ++i)
+	{
+		const int      stage = (i / STAGE_WIDTH) + 1;
+		const int      pos   = (i ^ 1) & (STAGE_WIDTH - 1);
+		_filter [stage]._coef [pos] = float (coef_arr [i]);
+	}
+}
+
+
+
+/*
+==============================================================================
+Name: process_sample
+Description:
+   Downsamples (x2) one pair of samples, to generate one output sample.
+Input parameters:
+	- in_ptr: pointer on the two samples to decimate
+Returns: Samplerate-reduced sample.
+Throws: Nothing
+==============================================================================
+*/
+
+template <int NC>
+float	Downsampler2xSse <NC>::process_sample (const float in_ptr [2])
+{
+	assert (in_ptr != 0);
+
+	// Combines two input samples and two mid-processing data
+	const __m128   spl_in  = _mm_loadu_ps (in_ptr);
+	const __m128   spl_mid = _mm_load_ps (_filter [NBR_STAGES]._mem);
+	__m128         y       = _mm_shuffle_ps (spl_in, spl_mid, 0x44);
+
+	__m128         mem     = _mm_load_ps (_filter [0]._mem);
+
+	// Processes each stage
+	StageProcSse <NBR_STAGES>::process_sample_pos (&_filter [0], y, mem);
+
+	_mm_store_ps (_filter [NBR_STAGES]._mem, y);
+
+	// Averages both paths and outputs the result
+	const __m128   dup_y = y;
+	y = _mm_shuffle_ps (y, y, 0x80);
+	y = _mm_add_ps (y, dup_y);
+	y = _mm_shuffle_ps (y, y, 3);
+	y = _mm_mul_ss (y, _mm_set_ss (0.5f));
+	float          result;
+	_mm_store_ss (&result, y);
+
+	return (result);
+}
+
+
+
+/*
+==============================================================================
+Name: process_block
+Description:
+   Downsamples (x2) a block of samples.
+	Input and output blocks may overlap, see assert() for details.
+Input parameters:
+	- in_ptr: Input array, containing nbr_spl * 2 samples.
+	- nbr_spl: Number of samples to output, > 0
+Output parameters:
+	- out_ptr: Array for the output samples, capacity: nbr_spl samples.
+Throws: Nothing
+==============================================================================
+*/
+
+template <int NC>
+void	Downsampler2xSse <NC>::process_block (float out_ptr [], const float in_ptr [], long nbr_spl)
+{
+	assert (in_ptr != 0);
+	assert (out_ptr != 0);
+	assert (out_ptr <= in_ptr || out_ptr >= in_ptr + nbr_spl * 2);
+	assert (nbr_spl > 0);
+
+	long           pos = 0;
+	do
+	{
+		out_ptr [pos] = process_sample (in_ptr + pos * 2);
+		++ pos;
+	}
+	while (pos < nbr_spl);
+}
+
+
+
+/*
+==============================================================================
+Name: process_sample_split
+Description:
+   Split (spectrum-wise) in half a pair of samples. The lower part of the
+   spectrum is a classic downsampling, equivalent to the output of
+   process_sample().
+   The higher part is the complementary signal: original filter response
+   is flipped from left to right, becoming a high-pass filter with the same
+   cutoff frequency. This signal is then critically sampled (decimation by 2),
+   flipping the spectrum: Fs/4...Fs/2 becomes Fs/4...0.
+Input parameters:
+	- in_ptr: pointer on the pair of input samples
+Output parameters:
+	- low: output sample, lower part of the spectrum (downsampling)
+	- high: output sample, higher part of the spectrum.
+Throws: Nothing
+==============================================================================
+*/
+
+template <int NC>
+void	Downsampler2xSse <NC>::process_sample_split (float &low, float &high, const float in_ptr [2])
+{
+	assert (in_ptr != 0);
+
+	// Combines two input samples and two mid-processing data
+	const __m128   spl_in  = _mm_loadu_ps (in_ptr);
+	const __m128   spl_mid = _mm_load_ps (_filter [NBR_STAGES]._mem);
+	__m128         y       = _mm_shuffle_ps (spl_in, spl_mid, 0x44);
+
+	__m128         mem     = _mm_load_ps (_filter [0]._mem);
+
+	// Processes each stage
+	StageProcSse <NBR_STAGES>::process_sample_pos (&_filter [0], y, mem);
+
+	_mm_store_ps (_filter [NBR_STAGES]._mem, y);
+
+	//Outputs the result
+	__m128         dup_y = y;
+	y = _mm_shuffle_ps (y, y, 0x80);
+	y = _mm_add_ps (y, dup_y);
+	y = _mm_shuffle_ps (y, y, 3);
+	y = _mm_mul_ss (y, _mm_set_ss (0.5f));
+	_mm_store_ss (&low, y);
+
+	dup_y = _mm_shuffle_ps (dup_y, dup_y, 3);
+	dup_y = _mm_sub_ps (dup_y, y);
+	_mm_store_ss (&high, dup_y);
+}
+
+
+
+/*
+==============================================================================
+Name: process_block_split
+Description:
+   Split (spectrum-wise) in half a block of samples. The lower part of the
+   spectrum is a classic downsampling, equivalent to the output of
+   process_block().
+   The higher part is the complementary signal: original filter response
+   is flipped from left to right, becoming a high-pass filter with the same
+   cutoff frequency. This signal is then critically sampled (decimation by 2),
+   flipping the spectrum: Fs/4...Fs/2 becomes Fs/4...0.
+	Input and output blocks may overlap, see assert() for details.
+Input parameters:
+	- in_ptr: Input array, containing nbr_spl * 2 samples.
+	- nbr_spl: Number of samples for each output, > 0
+Output parameters:
+	- out_l_ptr: Array for the output samples, lower part of the spectrum
+      (downsampling). Capacity: nbr_spl samples.
+	- out_h_ptr: Array for the output samples, higher part of the spectrum.
+      Capacity: nbr_spl samples.
+Throws: Nothing
+==============================================================================
+*/
+
+template <int NC>
+void	Downsampler2xSse <NC>::process_block_split (float out_l_ptr [], float out_h_ptr [], const float in_ptr [], long nbr_spl)
+{
+	assert (in_ptr != 0);
+	assert (out_l_ptr != 0);
+	assert (out_l_ptr <= in_ptr || out_l_ptr >= in_ptr + nbr_spl * 2);
+	assert (out_h_ptr != 0);
+	assert (out_h_ptr <= in_ptr || out_h_ptr >= in_ptr + nbr_spl * 2);
+	assert (out_h_ptr != out_l_ptr);
+	assert (nbr_spl > 0);
+
+	long           pos = 0;
+	do
+	{
+		process_sample_split (out_l_ptr [pos], out_h_ptr [pos], in_ptr + pos * 2);
+		++ pos;
+	}
+	while (pos < nbr_spl);
+}
+
+
+
+/*
+==============================================================================
+Name: clear_buffers
+Description:
+	Clears filter memory, as if it processed silence since an infinite amount
+	of time.
+Throws: Nothing
+==============================================================================
+*/
+
+template <int NC>
+void	Downsampler2xSse <NC>::clear_buffers ()
+{
+	for (int i = 0; i < NBR_STAGES + 1; ++i)
+	{
+		_filter [i]._mem [0] = 0;
+		_filter [i]._mem [1] = 0;
+		_filter [i]._mem [2] = 0;
+		_filter [i]._mem [3] = 0;
+	}
+}
+
+
+
+/*\\\ PROTECTED \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/
+
+
+
+/*\\\ PRIVATE \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/
+
+
+
+}  // namespace hiir
+
+
+
+#endif   // hiir_Downsampler2xSse_CODEHEADER_INCLUDED
+
+#undef hiir_Downsampler2xSse_CURRENT_CODEHEADER
+
+
+
+/*\\\ EOF \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\*/