Update Files

lib/aura/Backends/common_c/math/fft.c

@@ -0,0 +1,123 @@
+#include <stdint.h>
+
+#include "fft.h"
+
+#define PI_FLOAT 3.14159265358979323846f
+
+static uint32_t bitReverseUint32(uint32_t value, uint32_t log2N);
+
+static uint32_t log2Unsigned(uint32_t n) {
+	uint32_t res = 0;
+
+	while (n >>= 1) {
+		res++;
+	}
+
+	return res;
+}
+
+void aura_ditfft2(const aura_complex_t * times, int t, aura_complex_t * freqs, int f, int n, int step, bool inverse) {
+	if (n == 1) {
+		aura_copy_complex_elem(freqs, f, times, t);
+	}
+	else {
+		const int halfLen = n >> 1;
+
+		aura_ditfft2(times, t, freqs, f, halfLen, step << 1, inverse);
+		aura_ditfft2(times, t + step, freqs, f + halfLen, halfLen, step << 1, inverse);
+
+		const float t_exp = ((inverse ? 1.0f : -1.0f) * 2.0f * PI_FLOAT) / n;
+		for (int k = 0; k < halfLen; k++) {
+			aura_complex_t even = { 0 };
+			aura_complex_t odd = { 0 };
+			aura_copy_complex(&even, freqs[f + k]);
+			aura_copy_complex(&odd, freqs[f + k + halfLen]);
+
+			const aura_complex_t twiddle = aura_cmult(aura_cexp(t_exp * k), odd);
+
+			aura_copy_complex(&(freqs[f + k]), aura_cadd(even, twiddle));
+			aura_copy_complex(&(freqs[f + k + halfLen]), aura_csub(even, twiddle));
+		}
+	}
+}
+
+void aura_ditfft2_iterative(const aura_complex_t * times, aura_complex_t * freqs, int n, bool inverse, const aura_complex_t * exp_lut) {
+	// Decimate
+	const uint32_t log2N = log2Unsigned(n);
+	for (uint32_t i = 0; i < ((uint32_t) n); i++) {
+		uint32_t reversedI = bitReverseUint32(i, log2N);
+
+		if (reversedI >= i) {
+			aura_copy_complex(&freqs[i], times[reversedI]);
+			aura_copy_complex(&freqs[reversedI], times[i]);
+		}
+		else if (reversedI == i) {
+			aura_copy_complex(&freqs[reversedI], times[i]);
+		}
+	}
+
+	int halfLayerIdx = 0;
+	for (int layerSize = 2; layerSize <= n; layerSize <<= 1) {
+		const int halfLayerSize = layerSize >> 1;
+
+		aura_complex_t expRotationStep;
+		aura_copy_complex(&expRotationStep, exp_lut[halfLayerIdx]);
+		if (inverse) {
+			expRotationStep = aura_cconj(expRotationStep);
+		}
+
+		for (int sectionOffset = 0; sectionOffset < n; sectionOffset += layerSize) {
+			aura_complex_t currentExpRotation = {.real = 1.0, .imag = 0.0};
+
+			for (int i = 0; i < halfLayerSize; i++) {
+				aura_complex_t even;
+				aura_complex_t odd;
+				aura_copy_complex(&even, freqs[sectionOffset + i]);
+				aura_copy_complex(&odd, freqs[sectionOffset + i + halfLayerSize]);
+
+				const aura_complex_t twiddle = aura_cmult(currentExpRotation, odd);
+
+				aura_copy_complex(&(freqs[sectionOffset + i]), aura_cadd(even, twiddle));
+				aura_copy_complex(&(freqs[sectionOffset + i + halfLayerSize]), aura_csub(even, twiddle));
+
+				aura_copy_complex(&currentExpRotation, aura_cmult(currentExpRotation, expRotationStep));
+			}
+		}
+		halfLayerIdx++;
+	}
+}
+
+/**
+	The following bit reversal code is taken (and slightly changed) from
+	https://graphics.stanford.edu/~seander/bithacks.html#BitReverseTable.
+	The original sources are released in the public domain.
+**/
+
+// Lookup table for bit reversal where each entry is one possible byte
+static const uint8_t bitReverseTable[] = {
+	0x00, 0x80, 0x40, 0xC0, 0x20, 0xA0, 0x60, 0xE0, 0x10, 0x90, 0x50, 0xD0, 0x30, 0xB0, 0x70, 0xF0,
+	0x08, 0x88, 0x48, 0xC8, 0x28, 0xA8, 0x68, 0xE8, 0x18, 0x98, 0x58, 0xD8, 0x38, 0xB8, 0x78, 0xF8,
+	0x04, 0x84, 0x44, 0xC4, 0x24, 0xA4, 0x64, 0xE4, 0x14, 0x94, 0x54, 0xD4, 0x34, 0xB4, 0x74, 0xF4,
+	0x0C, 0x8C, 0x4C, 0xCC, 0x2C, 0xAC, 0x6C, 0xEC, 0x1C, 0x9C, 0x5C, 0xDC, 0x3C, 0xBC, 0x7C, 0xFC,
+	0x02, 0x82, 0x42, 0xC2, 0x22, 0xA2, 0x62, 0xE2, 0x12, 0x92, 0x52, 0xD2, 0x32, 0xB2, 0x72, 0xF2,
+	0x0A, 0x8A, 0x4A, 0xCA, 0x2A, 0xAA, 0x6A, 0xEA, 0x1A, 0x9A, 0x5A, 0xDA, 0x3A, 0xBA, 0x7A, 0xFA,
+	0x06, 0x86, 0x46, 0xC6, 0x26, 0xA6, 0x66, 0xE6, 0x16, 0x96, 0x56, 0xD6, 0x36, 0xB6, 0x76, 0xF6,
+	0x0E, 0x8E, 0x4E, 0xCE, 0x2E, 0xAE, 0x6E, 0xEE, 0x1E, 0x9E, 0x5E, 0xDE, 0x3E, 0xBE, 0x7E, 0xFE,
+	0x01, 0x81, 0x41, 0xC1, 0x21, 0xA1, 0x61, 0xE1, 0x11, 0x91, 0x51, 0xD1, 0x31, 0xB1, 0x71, 0xF1,
+	0x09, 0x89, 0x49, 0xC9, 0x29, 0xA9, 0x69, 0xE9, 0x19, 0x99, 0x59, 0xD9, 0x39, 0xB9, 0x79, 0xF9,
+	0x05, 0x85, 0x45, 0xC5, 0x25, 0xA5, 0x65, 0xE5, 0x15, 0x95, 0x55, 0xD5, 0x35, 0xB5, 0x75, 0xF5,
+	0x0D, 0x8D, 0x4D, 0xCD, 0x2D, 0xAD, 0x6D, 0xED, 0x1D, 0x9D, 0x5D, 0xDD, 0x3D, 0xBD, 0x7D, 0xFD,
+	0x03, 0x83, 0x43, 0xC3, 0x23, 0xA3, 0x63, 0xE3, 0x13, 0x93, 0x53, 0xD3, 0x33, 0xB3, 0x73, 0xF3,
+	0x0B, 0x8B, 0x4B, 0xCB, 0x2B, 0xAB, 0x6B, 0xEB, 0x1B, 0x9B, 0x5B, 0xDB, 0x3B, 0xBB, 0x7B, 0xFB,
+	0x07, 0x87, 0x47, 0xC7, 0x27, 0xA7, 0x67, 0xE7, 0x17, 0x97, 0x57, 0xD7, 0x37, 0xB7, 0x77, 0xF7,
+	0x0F, 0x8F, 0x4F, 0xCF, 0x2F, 0xAF, 0x6F, 0xEF, 0x1F, 0x9F, 0x5F, 0xDF, 0x3F, 0xBF, 0x7F, 0xFF
+};
+
+static uint32_t bitReverseUint32(uint32_t value, uint32_t log2N) {
+	return ((uint32_t)(
+		(bitReverseTable[ value        & 0xff] << 24) |
+		(bitReverseTable[(value >> 8 ) & 0xff] << 16) |
+		(bitReverseTable[(value >> 16) & 0xff] << 8 ) |
+		(bitReverseTable[(value >> 24) & 0xff]      )
+	) >> (32 - log2N));
+}

lib/aura/Backends/common_c/math/fft.h

@@ -0,0 +1,16 @@
+#pragma once
+
+#include <stdbool.h>
+
+#include "common_c/types/complex_t.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+void aura_ditfft2(const aura_complex_t * times, int t, aura_complex_t * freqs, int f, int n, int step, bool inverse);
+void aura_ditfft2_iterative(const aura_complex_t * times, aura_complex_t * freqs, int n, bool inverse, const aura_complex_t * exp_lut);
+
+#ifdef __cplusplus
+}
+#endif

lib/aura/Backends/common_c/types/complex_array.c

@@ -0,0 +1,21 @@
+#include <stdlib.h>
+
+#include "complex_array.h"
+
+aura_complex_t* aura_complex_array_alloc(int length) {
+	return calloc(length, sizeof(aura_complex_t));
+}
+
+void aura_complex_array_free(aura_complex_t* complex_array) {
+	free(complex_array);
+}
+
+aura_complex_t* aura_complex_array_set(aura_complex_t* complex_array, int index, float real, float imag) {
+	complex_array[index].real = real;
+	complex_array[index].imag = imag;
+	return &(complex_array[index]);
+}
+
+aura_complex_t* aura_complex_array_get(aura_complex_t* complex_array, int index) {
+	return &(complex_array[index]);
+}

lib/aura/Backends/common_c/types/complex_array.h

@@ -0,0 +1,17 @@
+#pragma once
+
+#include "common_c/types/complex_t.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+aura_complex_t* aura_complex_array_alloc(int length);
+void aura_complex_array_free(aura_complex_t* complex_array);
+
+aura_complex_t* aura_complex_array_set(aura_complex_t* complex_array, int index, float real, float imag);
+aura_complex_t* aura_complex_array_get(aura_complex_t* complex_array, int index);
+
+#ifdef __cplusplus
+}
+#endif

lib/aura/Backends/common_c/types/complex_t.c

@@ -0,0 +1,41 @@
+#include <math.h>
+
+#include "complex_t.h"
+
+void aura_copy_complex_elem(aura_complex_t *to, const int toIndex, const aura_complex_t *from, const int fromIndex) {
+	to[toIndex].real = from[fromIndex].real;
+	to[toIndex].imag = from[fromIndex].imag;
+}
+
+void aura_copy_complex(aura_complex_t *to, const aura_complex_t from) {
+	to->real = from.real;
+	to->imag = from.imag;
+}
+
+aura_complex_t aura_cexp(const float w) {
+	const aura_complex_t out = {.real = cosf(w), .imag = sinf(w)};
+	return out;
+}
+
+aura_complex_t aura_cadd(const aura_complex_t a, const aura_complex_t b) {
+	const aura_complex_t out = {.real = a.real + b.real, .imag = a.imag + b.imag};
+	return out;
+}
+
+aura_complex_t aura_csub(const aura_complex_t a, const aura_complex_t b) {
+	const aura_complex_t out = {.real = a.real - b.real, .imag = a.imag - b.imag};
+	return out;
+}
+
+aura_complex_t aura_cmult(const aura_complex_t a, const aura_complex_t b) {
+	const aura_complex_t out = {
+ 		.real = a.real * b.real - a.imag * b.imag,
+ 		.imag = a.real * b.imag + a.imag * b.real
+	};
+	return out;
+}
+
+aura_complex_t aura_cconj(const aura_complex_t val) {
+	const aura_complex_t out = {.real = val.real, .imag = -val.imag};
+	return out;
+}

lib/aura/Backends/common_c/types/complex_t.h

@@ -0,0 +1,24 @@
+#pragma once
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+typedef struct aura_complex_t {
+	float real;
+	float imag;
+} aura_complex_t;
+
+void aura_copy_complex_elem(aura_complex_t *to, const int toIndex, const aura_complex_t *from, const int fromIndex);
+void aura_copy_complex(aura_complex_t *to, const aura_complex_t from);
+
+aura_complex_t aura_cexp(const float w);
+
+aura_complex_t aura_cadd(const aura_complex_t a, const aura_complex_t b);
+aura_complex_t aura_csub(const aura_complex_t a, const aura_complex_t b);
+aura_complex_t aura_cmult(const aura_complex_t a, const aura_complex_t b);
+aura_complex_t aura_cconj(const aura_complex_t val);
+
+#ifdef __cplusplus
+}
+#endif

lib/aura/Backends/hl/aura/aurahl.c

@@ -0,0 +1,3 @@
+// Keep this file so that the headers are included in the compilation
+#include "hl/aura/math/fft.h"
+#include "hl/aura/types/complex_array.h"

lib/aura/Backends/hl/aura/aurahl.h

@@ -0,0 +1,3 @@
+#pragma once
+
+#define AURA_HL_FUNC(n) aura_hl_##n

lib/aura/Backends/hl/aura/math/FFT.h

@@ -0,0 +1,28 @@
+#pragma once
+
+#include <hl.h>
+
+#include <aura/types/_ComplexArray/HL_ComplexArrayImpl.h>
+
+#include "hl/aura/aurahl.h"
+#include "common_c/math/fft.h"
+#include "common_c/types/complex_t.h"
+
+HL_PRIM void AURA_HL_FUNC(ditfft2)(aura__types___ComplexArray__HL_ComplexArrayImpl time_array, int t, aura__types___ComplexArray__HL_ComplexArrayImpl freq_array, int f, int n, int step, bool inverse) {
+	const aura_complex_t *times = (aura_complex_t*) time_array->self;
+	aura_complex_t *freqs = (aura_complex_t*) freq_array->self;
+
+	aura_ditfft2(times, t, freqs, f, n, step, inverse);
+}
+
+HL_PRIM void AURA_HL_FUNC(ditfft2_iterative)(aura__types___ComplexArray__HL_ComplexArrayImpl time_array, aura__types___ComplexArray__HL_ComplexArrayImpl freq_array, int n, bool inverse, aura__types___ComplexArray__HL_ComplexArrayImpl exp_rotation_step_table) {
+	const aura_complex_t *times = (aura_complex_t*) time_array->self;
+	aura_complex_t *freqs = (aura_complex_t*) freq_array->self;
+
+	const aura_complex_t *exp_lut = (aura_complex_t*) exp_rotation_step_table->self;
+
+	aura_ditfft2_iterative(times, freqs, n, inverse, exp_lut);
+}
+
+DEFINE_PRIM(_VOID, ditfft2, _BYTES _I32 _BYTES _I32 _I32 _I32 _BOOL)
+DEFINE_PRIM(_VOID, ditfft2_iterative, _BYTES _BYTES _I32 _BOOL _BYTES)

lib/aura/Backends/hl/aura/types/complex_array.h

@@ -0,0 +1,29 @@
+#pragma once
+
+#include <hl.h>
+
+#include "hl/aura/aurahl.h"
+
+#include "common_c/types/complex_array.h"
+#include "common_c/types/complex_t.h"
+
+HL_PRIM vbyte* AURA_HL_FUNC(complex_array_alloc)(int length) {
+	return (vbyte*) aura_complex_array_alloc(length);
+}
+
+HL_PRIM void AURA_HL_FUNC(complex_array_free)(vbyte* complex_array) {
+	aura_complex_array_free((aura_complex_t*) complex_array);
+}
+
+HL_PRIM aura_complex_t* AURA_HL_FUNC(complex_array_set)(vbyte* complex_array, int index, float real, float imag) {
+	return aura_complex_array_set((aura_complex_t *) complex_array, index, real, imag);
+}
+HL_PRIM aura_complex_t* AURA_HL_FUNC(complex_array_get)(vbyte* complex_array, int index) {
+	return aura_complex_array_get((aura_complex_t*) complex_array, index);
+}
+
+DEFINE_PRIM(_BYTES, complex_array_alloc, _I32)
+DEFINE_PRIM(_VOID, complex_array_free, _BYTES)
+
+DEFINE_PRIM(_REF(_aura__types___Complex__ComplexImpl), complex_array_set, _BYTES _I32 _F32 _F32)
+DEFINE_PRIM(_REF(_aura__types___Complex__ComplexImpl), complex_array_get, _BYTES _I32)

lib/aura/Backends/hl/kincfile.js

@@ -0,0 +1,8 @@
+const project = new Project("aura-hl");
+
+project.addIncludeDir("../");
+
+project.addFile("../common_c/**");
+project.addFile("aura/**");
+
+resolve(project);

lib/aura/Backends/readme.md

@@ -0,0 +1,41 @@
+# Aura Backends
+
+The `/Backends` directory contains target-specific code that may be used on some
+targets to improve performance.
+
+The backends are enabled by default, but if you want to only use the generic
+Haxe sources (for performance comparisons e.g.), compile your Kha project with
+the command line flag `--aura-no-backend`.
+
+## Folder Structure
+
+- `/common_c`:
+  Pure C code that can be used by multiple backends.
+
+- `/hl`:
+  Sources/headers for the Hashlink/C backend. The header files are mostly
+  Hashlink API wrappers around the code in `/common_c`.
+
+Most of the backend sources mirror the respective Haxe code, so please don't
+expect much documentation for the individual functions.
+
+The Haxe implementation/glue code for the backends is in Aura's source files in
+`/Sources`. There are no Haxe files in the `/backend` folder that shadow
+original sources to reduce redundancy and ensure completeness of the API.
+
+Instead, there is usually a static class per backend implementation at the
+bottom of a Haxe source module, whose methods are then called and inlined from
+the original class if the [backend specific define is set](#defines). This way
+all the Haxe functionality for a module stays inside a module and is not
+distributed in many per-backend Haxe files, which also keeps the API consistent
+for each target.
+
+## Defines
+
+If the backends are enabled, Aura sets some defines before compilation which are
+based on the Haxe target to which the project is compiled. They should only be
+used internally, but for the sake of completeness they are documented here:
+
+- `AURA_NO_BACKEND`: Defined if backends are disabled.
+- `AURA_BACKEND_HL`: Defined if backends are enabled and the project is compiled
+  to a Hashlink target.