Project-wide: Start separating Android logic from DSP logic

This commit is contained in:
Iscle 2022-08-28 19:02:27 +02:00
parent 6d204bd433
commit 3870fb45d6
127 changed files with 5858 additions and 332 deletions

View File

@ -11,52 +11,52 @@ set(KISSFFT_PKGCONFIG OFF)
set(KISSFFT_STATIC ON)
set(KISSFFT_TEST OFF)
set(KISSFFT_TOOLS OFF)
add_subdirectory(src/cpp/kissfft)
add_subdirectory(src/cpp/viper/kissfft)
# ViPERFX
include_directories(src/include)
set(FILES
# Main
src/cpp/Effect.cpp
src/cpp/ProcessUnit_FX.cpp
src/cpp/viper.cpp
src/cpp/viper/Effect.cpp
src/cpp/viper/ViPER.cpp
src/cpp/ViPER4Android.cpp
# Effects
src/cpp/effects/AnalogX.cpp
src/cpp/effects/Cure.cpp
src/cpp/effects/DiffSurround.cpp
src/cpp/effects/DynamicSystem.cpp
src/cpp/effects/Reverberation.cpp
src/cpp/effects/SpeakerCorrection.cpp
src/cpp/effects/SpectrumExtend.cpp
src/cpp/effects/TubeSimulator.cpp
src/cpp/effects/VHE.cpp
src/cpp/effects/ViPERClarity.cpp
src/cpp/viper/effects/AnalogX.cpp
src/cpp/viper/effects/Cure.cpp
src/cpp/viper/effects/DiffSurround.cpp
src/cpp/viper/effects/DynamicSystem.cpp
src/cpp/viper/effects/Reverberation.cpp
src/cpp/viper/effects/SpeakerCorrection.cpp
src/cpp/viper/effects/SpectrumExtend.cpp
src/cpp/viper/effects/TubeSimulator.cpp
src/cpp/viper/effects/VHE.cpp
src/cpp/viper/effects/ViPERClarity.cpp
# Utils
src/cpp/utils/AdaptiveBuffer_F32.cpp
src/cpp/utils/CAllpassFilter.cpp
src/cpp/utils/CCombFilter.cpp
src/cpp/utils/CRevModel.cpp
src/cpp/utils/Crossfeed.cpp
src/cpp/utils/DepthSurround.cpp
src/cpp/utils/DynamicBass.cpp
src/cpp/utils/FixedBiquad.cpp
src/cpp/utils/Harmonic.cpp
src/cpp/utils/HiFi.cpp
src/cpp/utils/HighShelf.cpp
src/cpp/utils/IIR_1st.cpp
src/cpp/utils/IIR_NOrder_BW_BP.cpp
src/cpp/utils/IIR_NOrder_BW_LH.cpp
src/cpp/utils/MultiBiquad.cpp
src/cpp/utils/NoiseSharpening.cpp
src/cpp/utils/PassFilter.cpp
src/cpp/utils/PConvSingle_F32.cpp
src/cpp/utils/PolesFilter.cpp
src/cpp/utils/Subwoofer.cpp
src/cpp/utils/TimeConstDelay.cpp
src/cpp/utils/WaveBuffer_I32.cpp)
src/cpp/viper/utils/AdaptiveBuffer_F32.cpp
src/cpp/viper/utils/CAllpassFilter.cpp
src/cpp/viper/utils/CCombFilter.cpp
src/cpp/viper/utils/CRevModel.cpp
src/cpp/viper/utils/Crossfeed.cpp
src/cpp/viper/utils/DepthSurround.cpp
src/cpp/viper/utils/DynamicBass.cpp
src/cpp/viper/utils/FixedBiquad.cpp
src/cpp/viper/utils/Harmonic.cpp
src/cpp/viper/utils/HiFi.cpp
src/cpp/viper/utils/HighShelf.cpp
src/cpp/viper/utils/IIR_1st.cpp
src/cpp/viper/utils/IIR_NOrder_BW_BP.cpp
src/cpp/viper/utils/IIR_NOrder_BW_LH.cpp
src/cpp/viper/utils/MultiBiquad.cpp
src/cpp/viper/utils/NoiseSharpening.cpp
src/cpp/viper/utils/PassFilter.cpp
src/cpp/viper/utils/PConvSingle_F32.cpp
src/cpp/viper/utils/PolesFilter.cpp
src/cpp/viper/utils/Subwoofer.cpp
src/cpp/viper/utils/TimeConstDelay.cpp
src/cpp/viper/utils/WaveBuffer_I32.cpp)
message(${CMAKE_BUILD_TYPE})

148
src/cpp/ViPER4Android.cpp Normal file
View File

@ -0,0 +1,148 @@
#include <cstring>
#include <cerrno>
#include "viper/ViPER.h"
#include "log.h"
#include "viper/constants.h"
#define VIPER_EFFECT_NAME "ViPER4Android"
static effect_descriptor_t viper_descriptor = {
// Identical type/uuid to original ViPER4Android
.type = {0x00000000, 0x0000, 0x0000, 0x0000, {0x00, 0x00, 0x00, 0x00, 0x00, 0x00}},
.uuid = {0x41d3c987, 0xe6cf, 0x11e3, 0xa88a, {0x11, 0xab, 0xa5, 0xd5, 0xc5, 0x1b}},
.apiVersion = EFFECT_CONTROL_API_VERSION,
.flags = EFFECT_FLAG_OUTPUT_BOTH | EFFECT_FLAG_INPUT_BOTH | EFFECT_FLAG_INSERT_LAST | EFFECT_FLAG_TYPE_INSERT,
.cpuLoad = 8, // In 0.1 MIPS units as estimated on an ARM9E core (ARMv5TE) with 0 WS
.memoryUsage = 1, // In KB and includes only dynamically allocated memory
.name = VIPER_EFFECT_NAME,
.implementor = VIPER_AUTHORS
};
extern "C" {
struct ViperContext {
const struct effect_interface_s *interface; // Should always be the first struct member
ViPER *viper;
};
static int32_t Viper_IProcess(effect_handle_t self, audio_buffer_t *inBuffer, audio_buffer_t *outBuffer) {
auto pContext = (ViperContext *) self;
if (pContext == nullptr || inBuffer == nullptr || outBuffer == nullptr) {
VIPER_LOGE("Viper_IProcess: pContext, inBuffer or outBuffer is null!");
return -EINVAL;
}
return pContext->viper->process(inBuffer, outBuffer);
}
static int32_t Viper_ICommand(effect_handle_t self,
uint32_t cmdCode, uint32_t cmdSize, void *pCmdData,
uint32_t *replySize, void *pReplyData) {
auto pContext = (ViperContext *) self;
if (pContext == nullptr) {
VIPER_LOGE("Viper_ICommand: pContext is null!");
return -EINVAL;
}
return pContext->viper->command(cmdCode, cmdSize, pCmdData, replySize, pReplyData);
}
static int32_t Viper_IGetDescriptor(effect_handle_t self, effect_descriptor_t *pDescriptor) {
auto *pContext = (ViperContext *) self;
if (pContext == nullptr || pDescriptor == nullptr) {
VIPER_LOGE("Viper_IGetDescriptor: pContext or pDescriptor is null!");
return -EINVAL;
}
*pDescriptor = viper_descriptor;
return 0;
}
static const effect_interface_s viper_interface = {
.process = Viper_IProcess,
.command = Viper_ICommand,
.get_descriptor = Viper_IGetDescriptor
};
static int32_t
Viper_Create(const effect_uuid_t *uuid, int32_t sessionId __unused, int32_t ioId __unused, effect_handle_t *pHandle) {
VIPER_LOGI("Enter Viper_Create()");
if (uuid == nullptr || pHandle == nullptr) {
VIPER_LOGE("Viper_Create: uuid or pHandle is null!");
return -EINVAL;
}
if (memcmp(uuid, &viper_descriptor.uuid, sizeof(effect_uuid_t)) != 0) {
VIPER_LOGE("Viper_Create: uuid is not viper_descriptor.uuid!");
return -EINVAL;
}
VIPER_LOGI("Viper_Create: uuid matches, creating viper...");
auto *pContext = new ViperContext();
pContext->interface = &viper_interface;
pContext->viper = new ViPER();
*pHandle = (effect_handle_t) pContext;
return 0;
}
static int32_t Viper_Release(effect_handle_t handle) {
auto *pContext = (ViperContext *) handle;
VIPER_LOGI("Enter Viper_Release()");
if (pContext == nullptr) {
VIPER_LOGE("Viper_Release: pContext is null!");
return -EINVAL;
}
VIPER_LOGI("Viper_Release: deleting viper...");
if (pContext->viper != nullptr) {
delete pContext->viper;
pContext->viper = nullptr;
}
delete pContext;
return 0;
}
static int32_t Viper_GetDescriptor(const effect_uuid_t *uuid, effect_descriptor_t *pDescriptor) {
VIPER_LOGI("Enter Viper_GetDescriptor()");
if (uuid == nullptr || pDescriptor == nullptr) {
VIPER_LOGE("Viper_GetDescriptor: uuid or pDescriptor is null!");
return -EINVAL;
}
if (memcmp(uuid, &viper_descriptor.uuid, sizeof(effect_uuid_t)) != 0) {
VIPER_LOGE("Viper_GetDescriptor: uuid is not viper_descriptor.uuid!");
return -EINVAL;
}
VIPER_LOGI("Viper_GetDescriptor: uuid matches, returning descriptor...");
*pDescriptor = viper_descriptor;
return 0;
}
__attribute__ ((visibility ("default")))
audio_effect_library_t AUDIO_EFFECT_LIBRARY_INFO_SYM = {
.tag = AUDIO_EFFECT_LIBRARY_TAG,
.version = EFFECT_LIBRARY_API_VERSION,
.name = VIPER_EFFECT_NAME,
.implementor = VIPER_AUTHORS,
.create_effect = Viper_Create,
.release_effect = Viper_Release,
.get_descriptor = Viper_GetDescriptor,
};
} // extern "C"

@ -1 +0,0 @@
Subproject commit 8f47a67f595a6641c566087bf5277034be64f24d

View File

@ -1,12 +1,9 @@
#pragma once
#ifdef ANDROID_TOOLCHAIN
#include <android/log.h>
#else
#define __android_log_write(...) do {} while (0)
#define __android_log_print(...) do {} while (0)
#define ANDROID_LOG_INFO 1
#endif
#define v4a_print(status, message) __android_log_write(status, "ViPER4Android_Reworked", message)
#define v4a_printf(status, format, ...) __android_log_print(status, "ViPER4Android_Reworked", format, __VA_ARGS__)
#define TAG "ViPER4Android"
#define VIPER_LOGD(...) __android_log_print(ANDROID_LOG_DEBUG, TAG, __VA_ARGS__)
#define VIPER_LOGI(...) __android_log_print(ANDROID_LOG_INFO, TAG, __VA_ARGS__)
#define VIPER_LOGE(...) __android_log_print(ANDROID_LOG_ERROR, TAG, __VA_ARGS__)

View File

@ -1,175 +0,0 @@
#include <cstring>
#include "ProcessUnit_FX.h"
#include "constants.h"
#include "log.h"
#define EFFECT_NAME "ViPER4Android Reworked " VERSION_STRING
#define EFFECT_IMPLEMENTOR "ViPER.WYF, Martmists, Iscle"
static effect_descriptor_t viper_descriptor = {
// Identical type/uuid to original ViPER4Android
.type = {0x00000000, 0x0000, 0x0000, 0x0000, {0x00, 0x00, 0x00, 0x00, 0x00, 0x00}},
.uuid = {0x41d3c987, 0xe6cf, 0x11e3, 0xa88a, {0x11, 0xab, 0xa5, 0xd5, 0xc5, 0x1b}},
.apiVersion = EFFECT_CONTROL_API_VERSION,
.flags = EFFECT_FLAG_OUTPUT_BOTH | EFFECT_FLAG_INPUT_BOTH | EFFECT_FLAG_INSERT_LAST | EFFECT_FLAG_TYPE_INSERT,
.cpuLoad = 8, // In 0.1 MIPS units as estimated on an ARM9E core (ARMv5TE) with 0 WS
.memoryUsage = 1, // In KB and includes only dynamically allocated memory
.name = EFFECT_NAME,
.implementor = EFFECT_IMPLEMENTOR
};
extern "C" {
struct ViperContext {
const struct effect_interface_s *interface; // Should always be the first struct member
ProcessUnit_FX *effect;
};
static int32_t ViperProcess(effect_handle_t self, audio_buffer_t *inBuffer, audio_buffer_t *outBuffer) {
auto pContext = (ViperContext *) self;
if (pContext == nullptr) {
v4a_printf(ANDROID_LOG_ERROR, "ViperProcess(), Error [pContext = NULL]");
return -EINVAL;
}
if (inBuffer == nullptr) {
v4a_printf(ANDROID_LOG_ERROR, "ViperProcess(), Error [inBuffer = NULL]");
return -EINVAL;
}
if (outBuffer == nullptr) {
v4a_printf(ANDROID_LOG_ERROR, "ViperProcess(), Error [outBuffer = NULL]");
return -EINVAL;
}
return pContext->effect->process(inBuffer, outBuffer);
}
static int32_t
ViperCommand(effect_handle_t self, uint32_t cmdCode, uint32_t cmdSize, void *pCmdData, uint32_t *replySize,
void *pReplyData) {
auto pContext = (ViperContext *) self;
if (pContext == nullptr) {
v4a_printf(ANDROID_LOG_ERROR, "ViperCommand(), Error [pContext = NULL]");
return -EINVAL;
}
return pContext->effect->command(cmdCode, cmdSize, pCmdData, replySize, pReplyData);
}
static int32_t ViperGetDescriptor(effect_handle_t self, effect_descriptor_t *pDescriptor) {
auto *pContext = (ViperContext *) self;
if (pContext == nullptr) {
v4a_printf(ANDROID_LOG_ERROR, "ViperGetDescriptor(), Error [pContext = NULL]");
return -EINVAL;
}
if (pDescriptor == nullptr) {
v4a_print(ANDROID_LOG_ERROR, "ViperGetDescriptor(), Error [pDescriptor = NULL]");
return -EINVAL;
}
*pDescriptor = viper_descriptor;
return 0;
}
const effect_interface_s viper_interface = {
.process = ViperProcess,
.command = ViperCommand,
.get_descriptor = ViperGetDescriptor
};
int32_t ViperEffectCreate(const effect_uuid_t *uuid, int32_t sessionId, int32_t ioId, effect_handle_t *pHandle) {
v4a_print(ANDROID_LOG_INFO, "Enter ViperEffectCreate()");
if (uuid == nullptr) {
v4a_printf(ANDROID_LOG_ERROR, "ViperEffectCreate(), Error [uuid = NULL]");
return -EINVAL;
}
if (pHandle == nullptr) {
v4a_printf(ANDROID_LOG_ERROR, "ViperEffectCreate(), Error [pHandle = NULL]");
return -EINVAL;
}
if (memcmp(uuid, &viper_descriptor.uuid, sizeof(effect_uuid_t)) != 0) {
v4a_print(ANDROID_LOG_ERROR, "ViperEffectCreate(), Error [effect not found]");
return -EINVAL;
}
v4a_printf(ANDROID_LOG_INFO, "ViperEffectCreate(), uuid = %08x-%04x-%04x-%04x-%02x%02x%02x%02x%02x%02x",
uuid->timeLow, uuid->timeMid, uuid->timeHiAndVersion, uuid->clockSeq, uuid->node[0], uuid->node[1],
uuid->node[2], uuid->node[3], uuid->node[4], uuid->node[5]);
v4a_print(ANDROID_LOG_INFO, "ViperEffectCreate(), Constructing ProcessUnit_FX");
auto *pContext = new ViperContext();
pContext->interface = &viper_interface;
pContext->effect = new ProcessUnit_FX();
*pHandle = (effect_handle_t) pContext;
return 0;
}
int32_t ViperEffectRelease(effect_handle_t handle) {
auto *pContext = (ViperContext *) handle;
v4a_print(ANDROID_LOG_INFO, "Enter ViperEffectRelease()");
if (pContext == nullptr) {
v4a_printf(ANDROID_LOG_ERROR, "ViperEffectRelease(), Error [pContext = NULL]");
return -EINVAL;
}
v4a_print(ANDROID_LOG_INFO, "ViperEffectRelease(), Deconstructing ProcessUnit");
if (pContext->effect != nullptr) {
delete pContext->effect;
pContext->effect = nullptr;
}
delete pContext;
return 0;
}
int32_t ViperEffectGetDescriptor(const effect_uuid_t *uuid, effect_descriptor_t *pDescriptor) {
v4a_print(ANDROID_LOG_INFO, "Enter ViperEffectGetDescriptor()");
if (uuid == nullptr) {
v4a_printf(ANDROID_LOG_ERROR, "ViperEffectGetDescriptor(), Error [uuid = NULL]");
return -EINVAL;
}
if (pDescriptor == nullptr) {
v4a_printf(ANDROID_LOG_ERROR, "ViperEffectGetDescriptor(), Error [pDescriptor = NULL]");
return -EINVAL;
}
if (memcmp(uuid, &viper_descriptor.uuid, sizeof(effect_uuid_t)) != 0) {
v4a_print(ANDROID_LOG_ERROR, "ViperEffectGetDescriptor(), Error [effect not found]");
return -EINVAL;
}
v4a_printf(ANDROID_LOG_INFO, "ViperEffectGetDescriptor(), uuid = %08x-%04x-%04x-%04x-%02x%02x%02x%02x%02x%02x",
uuid->timeLow, uuid->timeMid, uuid->timeHiAndVersion, uuid->clockSeq, uuid->node[0], uuid->node[1],
uuid->node[2], uuid->node[3], uuid->node[4], uuid->node[5]);
*pDescriptor = viper_descriptor;
return 0;
}
audio_effect_library_t AUDIO_EFFECT_LIBRARY_INFO_SYM = {
.tag = AUDIO_EFFECT_LIBRARY_TAG,
.version = EFFECT_LIBRARY_API_VERSION,
.name = EFFECT_NAME,
.implementor = EFFECT_IMPLEMENTOR,
.create_effect = ViperEffectCreate,
.release_effect = ViperEffectRelease,
.get_descriptor = ViperEffectGetDescriptor,
};
}

View File

@ -2,7 +2,6 @@
// Created by mart on 7/25/21.
//
#include <cstdlib>
#include <cstring>
#include "Effect.h"
#include "constants.h"
@ -66,51 +65,51 @@ int32_t Effect::command(uint32_t cmdCode, uint32_t cmdSize, void *pCmdData, uint
return -EINVAL
int32_t Effect::configure(effect_config_t *newConfig) {
v4a_print(ANDROID_LOG_INFO, "Begin audio configure ...");
v4a_print(ANDROID_LOG_INFO, "Checking input and output configuration ...");
VIPER_LOGI("Begin audio configure ...");
VIPER_LOGI("Checking input and output configuration ...");
if (newConfig->inputCfg.samplingRate != newConfig->outputCfg.samplingRate) {
v4a_printf(ANDROID_LOG_ERROR, "ViPER4Android disabled, reason [in.SR = %d, out.SR = %d]",
VIPER_LOGE("ViPER4Android disabled, reason [in.SR = %d, out.SR = %d]",
newConfig->inputCfg.samplingRate, newConfig->outputCfg.samplingRate);
DO_ERROR();
}
if (newConfig->inputCfg.samplingRate > 48000) {
v4a_print(ANDROID_LOG_ERROR, "ViPER4Android disabled, reason [SR out of range]");
VIPER_LOGE("ViPER4Android disabled, reason [SR out of range]");
DO_ERROR();
}
if (newConfig->inputCfg.channels != newConfig->outputCfg.channels) {
v4a_printf(ANDROID_LOG_ERROR, "ViPER4Android disabled, reason [in.CH = %d, out.CH = %d]",
VIPER_LOGE("ViPER4Android disabled, reason [in.CH = %d, out.CH = %d]",
newConfig->inputCfg.channels, newConfig->outputCfg.channels);
DO_ERROR();
}
if (newConfig->inputCfg.channels != AUDIO_CHANNEL_OUT_STEREO) {
v4a_print(ANDROID_LOG_ERROR, "ViPER4Android disabled, reason [CH != 2]");
VIPER_LOGE("ViPER4Android disabled, reason [CH != 2]");
DO_ERROR();
}
// TODO: Allow multiple formats by converting before/after processing
if (newConfig->inputCfg.format != AUDIO_FORMAT_PCM_FLOAT) {
v4a_printf(ANDROID_LOG_ERROR, "ViPER4Android disabled, reason [in.FMT = %d]", newConfig->inputCfg.format);
v4a_print(ANDROID_LOG_ERROR, "We only accept f32 format");
VIPER_LOGE("ViPER4Android disabled, reason [in.FMT = %d]", newConfig->inputCfg.format);
VIPER_LOGE("We only accept f32 format");
DO_ERROR();
}
if (newConfig->outputCfg.format != AUDIO_FORMAT_PCM_FLOAT) {
v4a_printf(ANDROID_LOG_ERROR, "ViPER4Android disabled, reason [out.FMT = %d]", newConfig->outputCfg.format);
v4a_print(ANDROID_LOG_ERROR, "We only accept f32 format");
VIPER_LOGE("ViPER4Android disabled, reason [out.FMT = %d]", newConfig->outputCfg.format);
VIPER_LOGE("We only accept f32 format");
DO_ERROR();
}
v4a_print(ANDROID_LOG_INFO, "Input and output configuration checked.");
VIPER_LOGI("Input and output configuration checked.");
memcpy(&this->config, newConfig, sizeof(effect_config_t));
this->configureOk = true;
v4a_print(ANDROID_LOG_INFO, "Audio configure finished");
VIPER_LOGI("Audio configure finished");
return 0;
}

View File

@ -28,5 +28,5 @@ public:
// TODO: Figure out what buffer is used for
float *buffer;
uint32_t bufferSize;
void *instance; // type: ProcessUnit_FX
void *instance; // type: ViPER
};

View File

@ -2,13 +2,13 @@
// Created by mart on 7/25/21.
//
#include "ProcessUnit_FX.h"
#include "ViPER.h"
#include "Effect.h"
#include "constants.h"
ProcessUnit_FX::ProcessUnit_FX() {
v4a_print(ANDROID_LOG_INFO, "Welcome to ViPER4Android Reworked driver[SQ]");
v4a_printf(ANDROID_LOG_INFO, "Current version is %s %s", VERSION_STRING, VERSION_CODENAME);
ViPER::ViPER() {
VIPER_LOGI("Welcome to ViPER FX");
VIPER_LOGI("Current version is %s %s", VERSION_STRING, VERSION_CODENAME);
this->adaptiveBuffer = new AdaptiveBuffer_F32(2, 4096);
this->waveBuffer = new WaveBuffer_I32(2, 4096);
@ -113,7 +113,7 @@ ProcessUnit_FX::ProcessUnit_FX() {
this->mode = ViPER_FX_TYPE_NONE; // 0
}
ProcessUnit_FX::~ProcessUnit_FX() {
ViPER::~ViPER() {
delete this->adaptiveBuffer;
this->adaptiveBuffer = nullptr;
@ -178,21 +178,21 @@ ProcessUnit_FX::~ProcessUnit_FX() {
}
int32_t
ProcessUnit_FX::command(uint32_t cmdCode, uint32_t cmdSize, void *pCmdData, uint32_t *replySize, void *pReplyData) {
ViPER::command(uint32_t cmdCode, uint32_t cmdSize, void *pCmdData, uint32_t *replySize, void *pReplyData) {
// TODO
return -1;
}
void ProcessUnit_FX::processBuffer(float *buffer, int frameSize) {
void ViPER::processBuffer(float *buffer, int frameSize) {
// TODO
}
void ProcessUnit_FX::DispatchCommand(int param_1, int param_2, int param_3, int param_4, int param_5, int param_6,
int param_7) {
void ViPER::DispatchCommand(int param_1, int param_2, int param_3, int param_4, int param_5, int param_6,
int param_7) {
// TODO
}
void ProcessUnit_FX::ResetAllEffects() {
void ViPER::ResetAllEffects() {
if (this->adaptiveBuffer != nullptr) {
// this->adaptiveBuffer->FlushBuffer();
}

View File

@ -4,7 +4,6 @@
#pragma once
#include "Effect.h"
#include "viper.h"
#include "utils/WaveBuffer_I32.h"
#include "effects/SpectrumExtend.h"
#include "effects/Reverberation.h"
@ -25,12 +24,13 @@
#include "effects/ViPERBass.h"
#include "effects/SoftwareLimiter.h"
#include "effects/PlaybackGain.h"
#include "../ViPER4Android.h"
class ProcessUnit_FX : public Effect {
class ViPER : public Effect {
public:
ProcessUnit_FX();
ViPER();
~ProcessUnit_FX();
~ViPER();
int32_t command(uint32_t cmdCode, uint32_t cmdSize, void *pCmdData, uint32_t *replySize, void *pReplyData) override;

View File

@ -7,15 +7,14 @@
#ifdef ANDROID_TOOLCHAIN
#include <android/errno.h>
#else
#include <cerrno>
#endif
#include "log.h" // TODO: Remove this dependency
#include "../log.h" // TODO: Remove this dependency
#define STR_HELPER(x) #x
#define STR(x) STR_HELPER(x)
#define VERSION_STRING "v" STR(VERSION_MAJOR) "." STR(VERSION_MINOR)
#define DEFAULT_SAMPLERATE 44100
#define DEFAULT_SAMPLERATE 44100
#define VIPER_AUTHORS "viper.WYF, Martmists, Iscle"

49
src/cpp/viper/kissfft/.gitignore vendored Normal file
View File

@ -0,0 +1,49 @@
*.o
*.swp
*.so
*.so.*
*.a
*.dylib
test/testcpp
test/bm_fftw_double
test/bm_fftw_float
test/bm_fftw_int16_t
test/bm_fftw_int32_t
test/bm_fftw_simd
test/bm_kiss_double
test/bm_kiss_float
test/bm_kiss_int16_t
test/bm_kiss_int32_t
test/bm_kiss_simd
test/st_double
test/st_float
test/st_int16_t
test/st_int32_t
test/st_simd
test/tkfc_double
test/tkfc_float
test/tkfc_int16_t
test/tkfc_int32_t
test/tkfc_simd
test/tr_double
test/tr_float
test/tr_int16_t
test/tr_int32_t
test/tr_simd
tools/fastconv_double
tools/fastconv_float
tools/fastconv_int16_t
tools/fastconv_int32_t
tools/fastconv_simd
tools/fastconvr_double
tools/fastconvr_float
tools/fastconvr_int16_t
tools/fastconvr_int32_t
tools/fastconvr_simd
tools/fft_double
tools/fft_float
tools/fft_int16_t
tools/fft_int32_t
tools/fft_simd
test/test_simd
build

View File

@ -0,0 +1,27 @@
language: python
python:
- "3.7"
dist: focal
before_install:
- sudo apt-get install -y libfftw3-dev
addons:
apt:
update: true
install: true
jobs:
include:
- name: "build (make)"
script:
- make all
- make testall
- name: "build (cmake)"
script:
- mkdir build && cd build
- cmake ..
- make

View File

@ -0,0 +1,123 @@
1.3.0 2012-07-18
removed non-standard malloc.h from kiss_fft.h
moved -lm to end of link line
checked various return values
converted python Numeric code to NumPy
fixed test of int32_t on 64 bit OS
added padding in a couple of places to allow SIMD alignment of structs
1.2.9 2010-05-27
threadsafe ( including OpenMP )
first edition of kissfft.hh the C++ template fft engine
1.2.8
Changed memory.h to string.h -- apparently more standard
Added openmp extensions. This can have fairly linear speedups for larger FFT sizes.
1.2.7
Shrank the real-fft memory footprint. Thanks to Galen Seitz.
1.2.6 (Nov 14, 2006) The "thanks to GenArts" release.
Added multi-dimensional real-optimized FFT, see tools/kiss_fftndr
Thanks go to GenArts, Inc. for sponsoring the development.
1.2.5 (June 27, 2006) The "release for no good reason" release.
Changed some harmless code to make some compilers' warnings go away.
Added some more digits to pi -- why not.
Added kiss_fft_next_fast_size() function to help people decide how much to pad.
Changed multidimensional test from 8 dimensions to only 3 to avoid testing
problems with fixed point (sorry Buckaroo Banzai).
1.2.4 (Oct 27, 2005) The "oops, inverse fixed point real fft was borked" release.
Fixed scaling bug for inverse fixed point real fft -- also fixed test code that should've been failing.
Thanks to Jean-Marc Valin for bug report.
Use sys/types.h for more portable types than short,int,long => int16_t,int32_t,int64_t
If your system does not have these, you may need to define them -- but at least it breaks in a
loud and easily fixable way -- unlike silently using the wrong size type.
Hopefully tools/psdpng.c is fixed -- thanks to Steve Kellog for pointing out the weirdness.
1.2.3 (June 25, 2005) The "you want to use WHAT as a sample" release.
Added ability to use 32 bit fixed point samples -- requires a 64 bit intermediate result, a la 'long long'
Added ability to do 4 FFTs in parallel by using SSE SIMD instructions. This is accomplished by
using the __m128 (vector of 4 floats) as kiss_fft_scalar. Define USE_SIMD to use this.
I know, I know ... this is drifting a bit from the "kiss" principle, but the speed advantages
make it worth it for some. Also recent gcc makes it SOO easy to use vectors of 4 floats like a POD type.
1.2.2 (May 6, 2005) The Matthew release
Replaced fixed point division with multiply&shift. Thanks to Jean-Marc Valin for
discussions regarding. Considerable speedup for fixed-point.
Corrected overflow protection in real fft routines when using fixed point.
Finder's Credit goes to Robert Oschler of robodance for pointing me at the bug.
This also led to the CHECK_OVERFLOW_OP macro.
1.2.1 (April 4, 2004)
compiles cleanly with just about every -W warning flag under the sun
reorganized kiss_fft_state so it could be read-only/const. This may be useful for embedded systems
that are willing to predeclare twiddle factors, factorization.
Fixed C_MUL,S_MUL on 16-bit platforms.
tmpbuf will only be allocated if input & output buffers are same
scratchbuf will only be allocated for ffts that are not multiples of 2,3,5
NOTE: The tmpbuf,scratchbuf changes may require synchronization code for multi-threaded apps.
1.2 (Feb 23, 2004)
interface change -- cfg object is forward declaration of struct instead of void*
This maintains type saftey and lets the compiler warn/error about stupid mistakes.
(prompted by suggestion from Erik de Castro Lopo)
small speed improvements
added psdpng.c -- sample utility that will create png spectrum "waterfalls" from an input file
( not terribly useful yet)
1.1.1 (Feb 1, 2004 )
minor bug fix -- only affects odd rank, in-place, multi-dimensional FFTs
1.1 : (Jan 30,2004)
split sample_code/ into test/ and tools/
Removed 2-D fft and added N-D fft (arbitrary)
modified fftutil.c to allow multi-d FFTs
Modified core fft routine to allow an input stride via kiss_fft_stride()
(eased support of multi-D ffts)
Added fast convolution filtering (FIR filtering using overlap-scrap method, with tail scrap)
Add kfc.[ch]: the KISS FFT Cache. It takes care of allocs for you ( suggested by Oscar Lesta ).
1.0.1 (Dec 15, 2003)
fixed bug that occurred when nfft==1. Thanks to Steven Johnson.
1.0 : (Dec 14, 2003)
changed kiss_fft function from using a single buffer, to two buffers.
If the same buffer pointer is supplied for both in and out, kiss will
manage the buffer copies.
added kiss_fft2d and kiss_fftr as separate source files (declarations in kiss_fft.h )
0.4 :(Nov 4,2003) optimized for radix 2,3,4,5
0.3 :(Oct 28, 2003) woops, version 2 didn't actually factor out any radices other than 2.
Thanks to Steven Johnson for finding this one.
0.2 :(Oct 27, 2003) added mixed radix, only radix 2,4 optimized versions
0.1 :(May 19 2003) initial release, radix 2 only

View File

@ -0,0 +1,371 @@
# Directory for easier includes
# Anywhere you see include(...) you can check <root>/cmake for that file
set(CMAKE_MODULE_PATH cmake)
#
# Extract version from Makefile
#
file(READ Makefile _MAKEFILE_CONTENTS)
string(REGEX MATCH "KFVER_MAJOR = ([0-9]+)\n" KFVER_MAJOR_MATCH "${_MAKEFILE_CONTENTS}")
if(NOT KFVER_MAJOR_MATCH)
message(FATAL_ERROR "Cannot extract major (ABI) version from Makefile")
endif()
set(KFVER_MAJOR "${CMAKE_MATCH_1}")
string(REGEX MATCH "KFVER_MINOR = ([0-9]+)\n" KFVER_MINOR_MATCH "${_MAKEFILE_CONTENTS}")
if(NOT KFVER_MINOR_MATCH)
message(FATAL_ERROR "Cannot extract minor version from Makefile")
endif()
set(KFVER_MINOR "${CMAKE_MATCH_1}")
string(REGEX MATCH "KFVER_PATCH = ([0-9]+)\n" KFVER_PATCH_MATCH "${_MAKEFILE_CONTENTS}")
if(NOT KFVER_PATCH_MATCH)
message(FATAL_ERROR "Cannot extract patch version from Makefile")
endif()
set(KFVER_PATCH "${CMAKE_MATCH_1}")
set(MAKEFILE_EXTRACTED_VERSION "${KFVER_MAJOR}.${KFVER_MINOR}.${KFVER_PATCH}")
#
# Declare CMake project
#
cmake_minimum_required(VERSION 3.6)
project(kissfft VERSION "${MAKEFILE_EXTRACTED_VERSION}")
#
# CMake configuration options
#
# Principal datatype: double, float (default), int16_t, int32_t, simd
set(KISSFFT_DATATYPE "float" CACHE STRING "Principal datatype of kissfft: double, float (default), int16_t, int32_t, simd")
# Additional options
option(KISSFFT_OPENMP "Build kissfft with OpenMP support" OFF)
option(KISSFFT_PKGCONFIG "Build pkg-config files" ON)
option(KISSFFT_STATIC "Build kissfft as static (ON) or shared library (OFF)" OFF)
option(KISSFFT_TEST "Build kissfft tests" ON)
option(KISSFFT_TOOLS "Build kissfft command-line tools" ON)
option(KISSFFT_USE_ALLOCA "Use alloca instead of malloc" OFF)
#
# Validate datatype
#
if (NOT KISSFFT_DATATYPE MATCHES "^double$" AND
NOT KISSFFT_DATATYPE MATCHES "^float$" AND
NOT KISSFFT_DATATYPE MATCHES "^int16_t$" AND
NOT KISSFFT_DATATYPE MATCHES "^int32_t$" AND
NOT KISSFFT_DATATYPE MATCHES "^simd$")
message(FATAL_ERROR "Incorrect value of KISSFFT_DATATYPE! It can be one of: double, float, int16_t, int32_t, simd")
endif()
#
# Print principal datatype
#
message(STATUS "Building KissFFT with datatype: ${KISSFFT_DATATYPE}")
set(KISSFFT_OUTPUT_NAME "kissfft-${KISSFFT_DATATYPE}")
#
# Validate KISSFFT_STATIC
#
if (BUILD_SHARED_LIBS AND KISSFFT_STATIC)
message(FATAL_ERROR "Conflicting CMake configuration: -DBUILD_SHARED_LIBS=ON and -DKISSFFT_STATIC=ON")
endif()
#
# Enable BUILD_SHARED_LIBS for shared library build before
# kissfft library is declared
#
if (NOT KISSFFT_STATIC)
set(BUILD_SHARED_LIBS ON)
message(STATUS "Building shared library")
else()
message(STATUS "Building static library")
endif()
#
# Detect C compiler and pass appropriate flags
#
if(CMAKE_C_COMPILER_ID MATCHES "GNU|Clang|AppleClang")
add_compile_options(-ffast-math -fomit-frame-pointer
-W -Wall -Wcast-align -Wcast-qual -Wshadow -Wwrite-strings
"$<$<COMPILE_LANGUAGE:C>:-Wstrict-prototypes;-Wmissing-prototypes;-Wnested-externs;-Wbad-function-cast>")
endif()
#
# Add GNUInstallDirs for GNU infrastructure before target)include_directories
#
if(CMAKE_SYSTEM_NAME MATCHES "^(Linux|kFreeBSD|GNU)$" AND NOT CMAKE_CROSSCOMPILING)
include(GNUInstallDirs)
endif()
#
# Declare PKGINCLUDEDIR for kissfft include path
#
set(PKGINCLUDEDIR "${CMAKE_INSTALL_INCLUDEDIR}/kissfft")
message(STATUS "PKGINCLUDEDIR is ${PKGINCLUDEDIR}")
#
# Declare kissfft library ( libkissfft.a / libkissfft-${KISSFFT_DATATYPE}.so.${MAKEFILE_EXTRACTED_VERSION} )
#
add_library(kissfft
kiss_fft.c
kfc.c
kiss_fftnd.c
kiss_fftndr.c
kiss_fftr.c)
target_include_directories(kissfft PUBLIC
$<BUILD_INTERFACE:${PROJECT_SOURCE_DIR}>
$<INSTALL_INTERFACE:${PKGINCLUDEDIR}>)
#
# Set compile definitions based on datatype and additional support flags
#
set(KISSFFT_COMPILE_DEFINITIONS)
#
# double / float
#
if(KISSFFT_DATATYPE MATCHES "^float$" OR KISSFFT_DATATYPE MATCHES "^double$")
list(APPEND KISSFFT_COMPILE_DEFINITIONS kiss_fft_scalar=${KISSFFT_DATATYPE})
else()
#
# int16_t
#
if(KISSFFT_DATATYPE MATCHES "^int16_t$")
list(APPEND KISSFFT_COMPILE_DEFINITIONS FIXED_POINT=16)
else()
#
# int32_t
#
if(KISSFFT_DATATYPE MATCHES "^int32_t$")
list(APPEND KISSFFT_COMPILE_DEFINITIONS FIXED_POINT=32)
else()
#
# simd
#
if(KISSFFT_DATATYPE MATCHES "^simd$")
list(APPEND KISSFFT_COMPILE_DEFINITIONS USE_SIMD)
if (NOT MSVC)
target_compile_options(kissfft PRIVATE -msse)
else()
target_compile_options(kissfft PRIVATE "/arch:SSE")
endif()
endif()
endif()
endif()
endif()
#
# OpenMP support
#
if(KISSFFT_OPENMP)
if(CMAKE_C_COMPILER_ID MATCHES "GNU|Clang|AppleClang")
if (NOT MSVC)
target_compile_options(kissfft PRIVATE -fopenmp)
if(${CMAKE_VERSION} VERSION_LESS "3.13.0")
target_link_libraries(kissfft PRIVATE "-fopenmp")
else()
target_link_options(kissfft PRIVATE -fopenmp)
endif()
else()
target_compile_options(kissfft PRIVATE "/openmp")
if(${CMAKE_VERSION} VERSION_LESS "3.13.0")
target_link_libraries(kissfft PRIVATE "/openmp")
else()
target_link_options(kissfft PRIVATE "/openmp")
endif()
endif()
set(KISSFFT_EXPORT_SUFFIX "-openmp")
set(KISSFFT_OUTPUT_NAME "kissfft-${KISSFFT_DATATYPE}-openmp")
else()
message(FATAL_ERROR "Don't know how to enable OpenMP for this compiler")
endif()
endif()
#
# Shared / static library
#
if(NOT KISSFFT_STATIC)
list(APPEND KISSFFT_COMPILE_DEFINITIONS KISS_FFT_SHARED)
set_target_properties(kissfft PROPERTIES
C_VISIBILITY_PRESET hidden)
set(KISSFFT_EXPORT_SUFFIX "${KISSFFT_EXPORT_SUFFIX}-shared")
else()
set(KISSFFT_EXPORT_SUFFIX "${KISSFFT_EXPORT_SUFFIX}-static")
endif()
#
# Alloca support
#
if(KISSFFT_USE_ALLOCA)
list(APPEND KISSFFT_COMPILE_DEFINITIONS KISS_FFT_USE_ALLOCA)
endif()
# Set library name, version, soversion and aliases
target_compile_definitions(kissfft PUBLIC ${KISSFFT_COMPILE_DEFINITIONS})
set_target_properties(kissfft PROPERTIES
OUTPUT_NAME "${KISSFFT_OUTPUT_NAME}"
DEFINE_SYMBOL KISS_FFT_BUILD
EXPORT_NAME "${KISSFFT_OUTPUT_NAME}"
VERSION ${PROJECT_VERSION}
SOVERSION ${KFVER_MAJOR})
add_library(kissfft::kissfft ALIAS kissfft)
add_library(kissfft::kissfft-${KISSFFT_DATATYPE} ALIAS kissfft)
#
# Build with libm (-lm) on Linux and kFreeBSD
#
if(CMAKE_SYSTEM_NAME MATCHES "^(Linux|kFreeBSD|GNU)$" AND NOT CMAKE_CROSSCOMPILING)
target_link_libraries(kissfft PRIVATE m)
endif()
#
# Define a helper function to define executable file
#
function(add_kissfft_executable NAME)
add_executable(${NAME} ${ARGN})
target_link_libraries(${NAME} PRIVATE kissfft::kissfft)
#
# Build with libm (-lm) on Linux and kFreeBSD
#
if(CMAKE_SYSTEM_NAME MATCHES "^(Linux|kFreeBSD|GNU)$" AND NOT CMAKE_CROSSCOMPILING)
target_link_libraries(${NAME} PRIVATE m)
endif()
if (NOT KISSFFT_OPENMP)
set_target_properties(${NAME} PROPERTIES
OUTPUT_NAME "${NAME}-${KISSFFT_DATATYPE}")
else()
if (NOT MSVC)
target_compile_options(${NAME} PRIVATE -fopenmp)
if(${CMAKE_VERSION} VERSION_LESS "3.13.0")
target_link_libraries(${NAME} PRIVATE "-fopenmp")
else()
target_link_options(${NAME} PRIVATE -fopenmp)
endif()
else()
target_compile_options(${NAME} PRIVATE "/openmp")
if(${CMAKE_VERSION} VERSION_LESS "3.13.0")
target_link_libraries(${NAME} PRIVATE "/openmp")
else()
target_link_options(${NAME} PRIVATE "/openmp")
endif()
endif()
set_target_properties(${NAME} PROPERTIES
OUTPUT_NAME "${NAME}-${KISSFFT_DATATYPE}-openmp")
endif()
endfunction()
#
# Perform installation of kissfft library and development files
#
install(TARGETS kissfft EXPORT kissfft
ARCHIVE DESTINATION "${CMAKE_INSTALL_LIBDIR}"
LIBRARY DESTINATION "${CMAKE_INSTALL_LIBDIR}"
RUNTIME DESTINATION "${CMAKE_INSTALL_BINDIR}")
install(FILES kiss_fft.h
kissfft.hh
kiss_fftnd.h
kiss_fftndr.h
kiss_fftr.h
DESTINATION "${PKGINCLUDEDIR}")
set(KISSFFT_INSTALL_CMAKE "${CMAKE_INSTALL_LIBDIR}/cmake/${PROJECT_NAME}"
CACHE FILEPATH "Install destination of kissfft cmake modules")
mark_as_advanced(KISSFFT_INSTALL_CMAKE)
install(EXPORT kissfft DESTINATION "${KISSFFT_INSTALL_CMAKE}"
NAMESPACE "kissfft::"
FILE "${PROJECT_NAME}-${KISSFFT_DATATYPE}${KISSFFT_EXPORT_SUFFIX}-targets.cmake")
include(CMakePackageConfigHelpers)
configure_package_config_file(kissfft-config.cmake.in kissfft-config.cmake
INSTALL_DESTINATION "${KISSFFT_INSTALL_CMAKE}")
write_basic_package_version_file(kissfft-config-version.cmake COMPATIBILITY AnyNewerVersion)
install(FILES "${CMAKE_CURRENT_BINARY_DIR}/kissfft-config.cmake" "${CMAKE_CURRENT_BINARY_DIR}/kissfft-config-version.cmake"
DESTINATION "${KISSFFT_INSTALL_CMAKE}")
set(PKG_KISSFFT_DEFS)
foreach(_def ${KISSFFT_COMPILE_DEFINITIONS})
set(PKG_KISSFFT_DEFS "${PKG_KISSFFT_DEFS} -D${_def}")
endforeach()
if (KISSFFT_PKGCONFIG)
include(JoinPaths)
set(PKGCONFIG_KISSFFT_PKGINCLUDEDIR "\${includedir}/kissfft")
set(PKGCONFIG_KISSFFT_PREFIX "${CMAKE_INSTALL_PREFIX}")
set(PKGCONFIG_KISSFFT_VERSION "${kissfft_VERSION}")
join_paths(PKGCONFIG_KISSFFT_LIBDIR "\${prefix}" "${CMAKE_INSTALL_LIBDIR}")
join_paths(PKGCONFIG_KISSFFT_INCLUDEDIR "\${prefix}" "${CMAKE_INSTALL_INCLUDEDIR}")
if(KISSFFT_DATATYPE MATCHES "^simd$")
list(APPEND KISSFFT_COMPILE_DEFINITIONS USE_SIMD)
if (NOT MSVC)
set(PKG_KISSFFT_DEFS "${PKG_KISSFFT_DEFS} -msse")
else()
set(PKG_KISSFFT_DEFS "${PKG_KISSFFT_DEFS} /ARCH:SSE")
endif()
endif()
if (NOT KISSFFT_OPENMP)
configure_file(kissfft.pc.in "kissfft-${KISSFFT_DATATYPE}.pc" @ONLY)
install(FILES "${CMAKE_CURRENT_BINARY_DIR}/kissfft-${KISSFFT_DATATYPE}.pc"
DESTINATION "${CMAKE_INSTALL_LIBDIR}/pkgconfig")
else()
if (NOT MSVC)
set(PKG_OPENMP "-fopenmp")
set(PKG_KISSFFT_DEFS "${PKG_KISSFFT_DEFS} -fopenmp")
else()
set(PKG_KISSFFT_DEFS "${PKG_KISSFFT_DEFS} /openmp")
set(PKG_OPENMP "/openmp")
endif()
configure_file(kissfft.pc.in "kissfft-${KISSFFT_DATATYPE}-openmp.pc" @ONLY)
install(FILES "${CMAKE_CURRENT_BINARY_DIR}/kissfft-${KISSFFT_DATATYPE}-openmp.pc"
DESTINATION "${CMAKE_INSTALL_LIBDIR}/pkgconfig")
endif()
endif()
#
# Build and install tools if requested by user
#
if(KISSFFT_TOOLS)
add_subdirectory(tools)
endif()
#
# Build and run tests if requested by user
#
if(KISSFFT_TEST)
enable_testing()
add_subdirectory(test)
endif()

View File

@ -0,0 +1,11 @@
Copyright (c) 2003-2010 Mark Borgerding . All rights reserved.
KISS FFT is provided under:
SPDX-License-Identifier: BSD-3-Clause
Being under the terms of the BSD 3-clause "New" or "Revised" License,
according with:
LICENSES/BSD-3-Clause

View File

@ -0,0 +1,35 @@
Valid-License-Identifier: BSD-3-Clause
SPDX-URL: https://spdx.org/licenses/BSD-3-Clause.html
Usage-Guide:
To use the BSD 3-clause "New" or "Revised" License put the following SPDX
tag/value pair into a comment according to the placement guidelines in
the licensing rules documentation:
SPDX-License-Identifier: BSD-3-Clause
License-Text:
Copyright (c) <year> <owner> . All rights reserved.
Redistribution and use in source and binary forms, with or without modification,
are permitted provided that the following conditions are met:
1. Redistributions of source code must retain the above copyright notice,
this list of conditions and the following disclaimer.
2. Redistributions in binary form must reproduce the above copyright notice,
this list of conditions and the following disclaimer in the documentation
and/or other materials provided with the distribution.
3. Neither the name of the copyright holder nor the names of its contributors
may be used to endorse or promote products derived from this software without
specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

View File

@ -0,0 +1,30 @@
Valid-License-Identifier: Unlicense
SPDX-URL: https://spdx.org/licenses/Unlicense.html
Usage-Guide:
To use the Unlicense put the following SPDX tag/value pair into a
comment according to the placement guidelines in the licensing rules
documentation:
SPDX-License-Identifier: Unlicense
License-Text:
This is free and unencumbered software released into the public domain.
Anyone is free to copy, modify, publish, use, compile, sell, or distribute
this software, either in source code form or as a compiled binary, for any
purpose, commercial or non-commercial, and by any means.
In jurisdictions that recognize copyright laws, the author or authors of this
software dedicate any and all copyright interest in the software to the public
domain. We make this dedication for the benefit of the public at large and
to the detriment of our heirs and successors. We intend this dedication to be
an overt act of relinquishment in perpetuity of all present and future rights
to this software under copyright law.
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 VIPER_AUTHORS
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.
For more information, please refer to <http://unlicense.org/>

View File

@ -0,0 +1,245 @@
# KISS FFT [![Build Status](https://travis-ci.com/mborgerding/kissfft.svg?branch=master)](https://travis-ci.com/mborgerding/kissfft)
KISS FFT - A mixed-radix Fast Fourier Transform based up on the principle,
"Keep It Simple, Stupid."
There are many great fft libraries already around. Kiss FFT is not trying
to be better than any of them. It only attempts to be a reasonably efficient,
moderately useful FFT that can use fixed or floating data types and can be
incorporated into someone's C program in a few minutes with trivial licensing.
## USAGE:
The basic usage for 1-d complex FFT is:
```c
#include "kiss_fft.h"
kiss_fft_cfg cfg = kiss_fft_alloc( nfft ,is_inverse_fft ,0,0 );
while ...
... // put kth sample in cx_in[k].r and cx_in[k].i
kiss_fft( cfg , cx_in , cx_out );
... // transformed. DC is in cx_out[0].r and cx_out[0].i
kiss_fft_free(cfg);
```
- **Note**: frequency-domain data is stored from dc up to 2pi.
so cx_out[0] is the dc bin of the FFT
and cx_out[nfft/2] is the Nyquist bin (if exists)
Declarations are in "kiss_fft.h", along with a brief description of the
functions you'll need to use.
Code definitions for 1d complex FFTs are in kiss_fft.c.
You can do other cool stuff with the extras you'll find in tools/
> - multi-dimensional FFTs
> - real-optimized FFTs (returns the positive half-spectrum:
(nfft/2+1) complex frequency bins)
> - fast convolution FIR filtering (not available for fixed point)
> - spectrum image creation
The core fft and most tools/ code can be compiled to use float, double,
Q15 short or Q31 samples. The default is float.
## BUILDING:
There are two functionally-equivalent build systems supported by kissfft:
- Make (traditional Makefiles for Unix / Linux systems)
- CMake (more modern and feature-rich build system developed by Kitware)
To build kissfft, the following build environment can be used:
- GNU build environment with GCC, Clang and GNU Make or CMake (>= 3.6)
- Microsoft Visual C++ (MSVC) with CMake (>= 3.6)
Additional libraries required to build and test kissfft include:
- libpng for psdpng tool,
- libfftw3 to validate kissfft results against it,
- python 2/3 with Numpy to validate kissfft results against it.
- OpenMP supported by GCC, Clang or MSVC for multi-core FFT transformations
Environments like Cygwin and MinGW can be highly likely used to build kissfft
targeting Windows platform, but no tests were performed to the date.
Both Make and CMake builds are easily configurable:
- `KISSFFT_DATATYPE=<datatype>` (for Make) or `-DKISSFFT_DATATYPE=<datatype>`
(for CMake) denote the principal datatype used by kissfft. It can be one
of the following:
- float (default)
- double
- int16_t
- int32_t
- SIMD (requires SSE instruction set support on target CPU)
- `KISSFFT_OPENMP=1` (for Make) or `-DKISSFFT_OPENMP=ON` (for CMake) builds kissfft
with OpenMP support. Please note that a supported compiler is required and this
option is turned off by default.
- `KISSFFT_STATIC=1` (for Make) or `-DKISSFFT_STATIC=ON` (for CMake) instructs
the builder to create static library ('.lib' for Windows / '.a' for Unix or Linux).
By default, this option is turned off and the shared library is created
('.dll' for Windows, '.so' for Linux or Unix, '.dylib' for Mac OSX)
- `-DKISSFFT_TEST=OFF` (for CMake) disables building tests for kissfft. On Make,
building tests is done separately by 'make testall' or 'make testsingle', so
no specific setting is required.
- `KISSFFT_TOOLS=0` (for Make) or `-DKISSFFT_TOOLS=OFF` (for CMake) builds kissfft
without command-line tools like 'fastconv'. By default the tools are built.
- `KISSFFT_USE_ALLOCA=1` (for Make) or `-DKISSFFT_USE_ALLOCA=ON` (for CMake)
build kissfft with 'alloca' usage instead of 'malloc' / 'free'.
- `PREFIX=/full/path/to/installation/prefix/directory` (for Make) or
`-DCMAKE_INSTALL_PREFIX=/full/path/to/installation/prefix/directory` (for CMake)
specifies the prefix directory to install kissfft into.
For example, to build kissfft as a static library with 'int16_t' datatype and
OpenMP support using Make, run the command from kissfft source tree:
```
make KISSFFT_DATATYPE=int16_t KISSFFT_STATIC=1 KISSFFT_OPENMP=1 all
```
The same configuration for CMake is:
```
mkdir build && cd build
cmake -DKISSFFT_DATATYPE=int16_t -DKISSFFT_STATIC=ON -DKISSFFT_OPENMP=ON ..
make all
```
To specify '/tmp/1234' as installation prefix directory, run:
```
make PREFIX=/tmp/1234 KISSFFT_DATATYPE=int16_t KISSFFT_STATIC=1 KISSFFT_OPENMP=1 install
```
or
```
mkdir build && cd build
cmake -DCMAKE_INSTALL_PREFIX=/tmp/1234 -DKISSFFT_DATATYPE=int16_t -DKISSFFT_STATIC=ON -DKISSFFT_OPENMP=ON ..
make all
make install
```
## TESTING:
To validate the build configured as an example above, run the following command from
kissfft source tree:
```
make KISSFFT_DATATYPE=int16_t KISSFFT_STATIC=1 KISSFFT_OPENMP=1 testsingle
```
if using Make, or:
```
make test
```
if using CMake.
To test all possible build configurations, please run an extended testsuite from
kissfft source tree:
```
sh test/kissfft-testsuite.sh
```
Please note that the extended testsuite takes around 20-40 minutes depending on device
it runs on. This testsuite is useful for reporting bugs or testing the pull requests.
## BACKGROUND
I started coding this because I couldn't find a fixed point FFT that didn't
use assembly code. I started with floating point numbers so I could get the
theory straight before working on fixed point issues. In the end, I had a
little bit of code that could be recompiled easily to do ffts with short, float
or double (other types should be easy too).
Once I got my FFT working, I was curious about the speed compared to
a well respected and highly optimized fft library. I don't want to criticize
this great library, so let's call it FFT_BRANDX.
During this process, I learned:
> 1. FFT_BRANDX has more than 100K lines of code. The core of kiss_fft is about 500 lines (cpx 1-d).
> 2. It took me an embarrassingly long time to get FFT_BRANDX working.
> 3. A simple program using FFT_BRANDX is 522KB. A similar program using kiss_fft is 18KB (without optimizing for size).
> 4. FFT_BRANDX is roughly twice as fast as KISS FFT in default mode.
It is wonderful that free, highly optimized libraries like FFT_BRANDX exist.
But such libraries carry a huge burden of complexity necessary to extract every
last bit of performance.
**Sometimes simpler is better, even if it's not better.**
## FREQUENTLY ASKED QUESTIONS:
> Q: Can I use kissfft in a project with a ___ license?</br>
> A: Yes. See LICENSE below.
> Q: Why don't I get the output I expect?</br>
> A: The two most common causes of this are
> 1) scaling : is there a constant multiplier between what you got and what you want?
> 2) mixed build environment -- all code must be compiled with same preprocessor
> definitions for FIXED_POINT and kiss_fft_scalar
> Q: Will you write/debug my code for me?</br>
> A: Probably not unless you pay me. I am happy to answer pointed and topical questions, but
> I may refer you to a book, a forum, or some other resource.
## PERFORMANCE
(on Athlon XP 2100+, with gcc 2.96, float data type)
Kiss performed 10000 1024-pt cpx ffts in .63 s of cpu time.
For comparison, it took md5sum twice as long to process the same amount of data.
Transforming 5 minutes of CD quality audio takes less than a second (nfft=1024).
**DO NOT:**
- use Kiss if you need the Fastest Fourier Transform in the World
- ask me to add features that will bloat the code
## UNDER THE HOOD
Kiss FFT uses a time decimation, mixed-radix, out-of-place FFT. If you give it an input buffer
and output buffer that are the same, a temporary buffer will be created to hold the data.
No static data is used. The core routines of kiss_fft are thread-safe (but not all of the tools directory).[
No scaling is done for the floating point version (for speed).
Scaling is done both ways for the fixed-point version (for overflow prevention).
Optimized butterflies are used for factors 2,3,4, and 5.
The real (i.e. not complex) optimization code only works for even length ffts. It does two half-length
FFTs in parallel (packed into real&imag), and then combines them via twiddling. The result is
nfft/2+1 complex frequency bins from DC to Nyquist. If you don't know what this means, search the web.
The fast convolution filtering uses the overlap-scrap method, slightly
modified to put the scrap at the tail.
## LICENSE
Revised BSD License, see COPYING for verbiage.
Basically, "free to use&change, give credit where due, no guarantees"
Note this license is compatible with GPL at one end of the spectrum and closed, commercial software at
the other end. See http://www.fsf.org/licensing/licenses
## TODO
- Add real optimization for odd length FFTs
- Document/revisit the input/output fft scaling
- Make doc describing the overlap (tail) scrap fast convolution filtering in kiss_fastfir.c
- Test all the ./tools/ code with fixed point (kiss_fastfir.c doesn't work, maybe others)
## AUTHOR
Mark Borgerding
Mark@Borgerding.net

View File

@ -0,0 +1,78 @@
If you are reading this, it means you think you may be interested in using the SIMD extensions in kissfft
to do 4 *separate* FFTs at once.
Beware! Beyond here there be dragons!
This API is not easy to use, is not well documented, and breaks the KISS principle.
Still reading? Okay, you may get rewarded for your patience with a considerable speedup
(2-3x) on intel x86 machines with SSE if you are willing to jump through some hoops.
The basic idea is to use the packed 4 float __m128 data type as a scalar element.
This means that the format is pretty convoluted. It performs 4 FFTs per fft call on signals A,B,C,D.
For complex data, the data is interlaced as follows:
rA0,rB0,rC0,rD0, iA0,iB0,iC0,iD0, rA1,rB1,rC1,rD1, iA1,iB1,iC1,iD1 ...
where "rA0" is the real part of the zeroth sample for signal A
Real-only data is laid out:
rA0,rB0,rC0,rD0, rA1,rB1,rC1,rD1, ...
Compile with gcc flags something like
-O3 -mpreferred-stack-boundary=4 -DUSE_SIMD=1 -msse
Be aware of SIMD alignment. This is the most likely cause of segfaults.
The code within kissfft uses scratch variables on the stack.
With SIMD, these must have addresses on 16 byte boundaries.
Search on "SIMD alignment" for more info.
Robin at Divide Concept was kind enough to share his code for formatting to/from the SIMD kissfft.
I have not run it -- use it at your own risk. It appears to do 4xN and Nx4 transpositions
(out of place).
void SSETools::pack128(float* target, float* source, unsigned long size128)
{
__m128* pDest = (__m128*)target;
__m128* pDestEnd = pDest+size128;
float* source0=source;
float* source1=source0+size128;
float* source2=source1+size128;
float* source3=source2+size128;
while(pDest<pDestEnd)
{
*pDest=_mm_set_ps(*source3,*source2,*source1,*source0);
source0++;
source1++;
source2++;
source3++;
pDest++;
}
}
void SSETools::unpack128(float* target, float* source, unsigned long size128)
{
float* pSrc = source;
float* pSrcEnd = pSrc+size128*4;
float* target0=target;
float* target1=target0+size128;
float* target2=target1+size128;
float* target3=target2+size128;
while(pSrc<pSrcEnd)
{
*target0=pSrc[0];
*target1=pSrc[1];
*target2=pSrc[2];
*target3=pSrc[3];
target0++;
target1++;
target2++;
target3++;
pSrc+=4;
}
}

View File

@ -0,0 +1,39 @@
Speed:
* If you want to use multiple cores, then compile with -openmp or -fopenmp (see your compiler docs).
Realize that larger FFTs will reap more benefit than smaller FFTs. This generally uses more CPU time, but
less wall time.
* experiment with compiler flags
Special thanks to Oscar Lesta. He suggested some compiler flags
for gcc that make a big difference. They shave 10-15% off
execution time on some systems. Try some combination of:
-march=pentiumpro
-ffast-math
-fomit-frame-pointer
* If the input data has no imaginary component, use the kiss_fftr code under tools/.
Real ffts are roughly twice as fast as complex.
* If you can rearrange your code to do 4 FFTs in parallel and you are on a recent Intel or AMD machine,
then you might want to experiment with the USE_SIMD code. See README.simd
Reducing code size:
* remove some of the butterflies. There are currently butterflies optimized for radices
2,3,4,5. It is worth mentioning that you can still use FFT sizes that contain
other factors, they just won't be quite as fast. You can decide for yourself
whether to keep radix 2 or 4. If you do some work in this area, let me
know what you find.
* For platforms where ROM/code space is more plentiful than RAM,
consider creating a hardcoded kiss_fft_state. In other words, decide which
FFT size(s) you want and make a structure with the correct factors and twiddles.
* Frank van der Hulst offered numerous suggestions for smaller code size and correct operation
on embedded targets. "I'm happy to help anyone who is trying to implement KISSFFT on a micro"
Some of these were rolled into the mainline code base:
- using long casts to promote intermediate results of short*short multiplication
- delaying allocation of buffers that are sometimes unused.
In some cases, it may be desirable to limit capability in order to better suit the target:
- predefining the twiddle tables for the desired fft size.

View File

@ -0,0 +1,167 @@
/*
* Copyright (c) 2003-2010, Mark Borgerding. All rights reserved.
* This file is part of KISS FFT - https://github.com/mborgerding/kissfft
*
* SPDX-License-Identifier: BSD-3-Clause
* See COPYING file for more information.
*/
/* kiss_fft.h
defines kiss_fft_scalar as either short or a float type
and defines
typedef struct { kiss_fft_scalar r; kiss_fft_scalar i; }kiss_fft_cpx; */
#ifndef _kiss_fft_guts_h
#define _kiss_fft_guts_h
#include "kiss_fft.h"
#include "kiss_fft_log.h"
#include <limits.h>
#define MAXFACTORS 32
/* e.g. an fft of length 128 has 4 factors
as far as kissfft is concerned
4*4*4*2
*/
struct kiss_fft_state{
int nfft;
int inverse;
int factors[2*MAXFACTORS];
kiss_fft_cpx twiddles[1];
};
/*
Explanation of macros dealing with complex math:
C_MUL(m,a,b) : m = a*b
C_FIXDIV( c , div ) : if a fixed point impl., c /= div. noop otherwise
C_SUB( res, a,b) : res = a - b
C_SUBFROM( res , a) : res -= a
C_ADDTO( res , a) : res += a
* */
#ifdef FIXED_POINT
#include <stdint.h>
#if (FIXED_POINT==32)
# define FRACBITS 31
# define SAMPPROD int64_t
#define SAMP_MAX INT32_MAX
#define SAMP_MIN INT32_MIN
#else
# define FRACBITS 15
# define SAMPPROD int32_t
#define SAMP_MAX INT16_MAX
#define SAMP_MIN INT16_MIN
#endif
#if defined(CHECK_OVERFLOW)
# define CHECK_OVERFLOW_OP(a,op,b) \
if ( (SAMPPROD)(a) op (SAMPPROD)(b) > SAMP_MAX || (SAMPPROD)(a) op (SAMPPROD)(b) < SAMP_MIN ) { \
KISS_FFT_WARNING("overflow (%d " #op" %d) = %ld", (a),(b),(SAMPPROD)(a) op (SAMPPROD)(b)); }
#endif
# define smul(a,b) ( (SAMPPROD)(a)*(b) )
# define sround( x ) (kiss_fft_scalar)( ( (x) + (1<<(FRACBITS-1)) ) >> FRACBITS )
# define S_MUL(a,b) sround( smul(a,b) )
# define C_MUL(m,a,b) \
do{ (m).r = sround( smul((a).r,(b).r) - smul((a).i,(b).i) ); \
(m).i = sround( smul((a).r,(b).i) + smul((a).i,(b).r) ); }while(0)
# define DIVSCALAR(x,k) \
(x) = sround( smul( x, SAMP_MAX/k ) )
# define C_FIXDIV(c,div) \
do { DIVSCALAR( (c).r , div); \
DIVSCALAR( (c).i , div); }while (0)
# define C_MULBYSCALAR( c, s ) \
do{ (c).r = sround( smul( (c).r , s ) ) ;\
(c).i = sround( smul( (c).i , s ) ) ; }while(0)
#else /* not FIXED_POINT*/
# define S_MUL(a,b) ( (a)*(b) )
#define C_MUL(m,a,b) \
do{ (m).r = (a).r*(b).r - (a).i*(b).i;\
(m).i = (a).r*(b).i + (a).i*(b).r; }while(0)
# define C_FIXDIV(c,div) /* NOOP */
# define C_MULBYSCALAR( c, s ) \
do{ (c).r *= (s);\
(c).i *= (s); }while(0)
#endif
#ifndef CHECK_OVERFLOW_OP
# define CHECK_OVERFLOW_OP(a,op,b) /* noop */
#endif
#define C_ADD( res, a,b)\
do { \
CHECK_OVERFLOW_OP((a).r,+,(b).r)\
CHECK_OVERFLOW_OP((a).i,+,(b).i)\
(res).r=(a).r+(b).r; (res).i=(a).i+(b).i; \
}while(0)
#define C_SUB( res, a,b)\
do { \
CHECK_OVERFLOW_OP((a).r,-,(b).r)\
CHECK_OVERFLOW_OP((a).i,-,(b).i)\
(res).r=(a).r-(b).r; (res).i=(a).i-(b).i; \
}while(0)
#define C_ADDTO( res , a)\
do { \
CHECK_OVERFLOW_OP((res).r,+,(a).r)\
CHECK_OVERFLOW_OP((res).i,+,(a).i)\
(res).r += (a).r; (res).i += (a).i;\
}while(0)
#define C_SUBFROM( res , a)\
do {\
CHECK_OVERFLOW_OP((res).r,-,(a).r)\
CHECK_OVERFLOW_OP((res).i,-,(a).i)\
(res).r -= (a).r; (res).i -= (a).i; \
}while(0)
#ifdef FIXED_POINT
# define KISS_FFT_COS(phase) floor(.5+SAMP_MAX * cos (phase))
# define KISS_FFT_SIN(phase) floor(.5+SAMP_MAX * sin (phase))
# define HALF_OF(x) ((x)>>1)
#elif defined(USE_SIMD)
# define KISS_FFT_COS(phase) _mm_set1_ps( cos(phase) )
# define KISS_FFT_SIN(phase) _mm_set1_ps( sin(phase) )
# define HALF_OF(x) ((x)*_mm_set1_ps(.5))
#else
# define KISS_FFT_COS(phase) (kiss_fft_scalar) cos(phase)
# define KISS_FFT_SIN(phase) (kiss_fft_scalar) sin(phase)
# define HALF_OF(x) ((x)*((kiss_fft_scalar).5))
#endif
#define kf_cexp(x,phase) \
do{ \
(x)->r = KISS_FFT_COS(phase);\
(x)->i = KISS_FFT_SIN(phase);\
}while(0)
/* a debugging function */
#define pcpx(c)\
KISS_FFT_DEBUG("%g + %gi\n",(double)((c)->r),(double)((c)->i))
#ifdef KISS_FFT_USE_ALLOCA
// define this to allow use of alloca instead of malloc for temporary buffers
// Temporary buffers are used in two case:
// 1. FFT sizes that have "bad" factors. i.e. not 2,3 and 5
// 2. "in-place" FFTs. Notice the quotes, since kissfft does not really do an in-place transform.
#include <alloca.h>
#define KISS_FFT_TMP_ALLOC(nbytes) alloca(nbytes)
#define KISS_FFT_TMP_FREE(ptr)
#else
#define KISS_FFT_TMP_ALLOC(nbytes) KISS_FFT_MALLOC(nbytes)
#define KISS_FFT_TMP_FREE(ptr) KISS_FFT_FREE(ptr)
#endif
#endif /* _kiss_fft_guts_h */

View File

@ -0,0 +1,23 @@
# This module provides function for joining paths
# known from most languages
#
# SPDX-License-Identifier: (MIT OR CC0-1.0)
# Copyright 2020 Jan Tojnar
# https://github.com/jtojnar/cmake-snips
#
# Modelled after Pythons os.path.join
# https://docs.python.org/3.7/library/os.path.html#os.path.join
# Windows not supported
function(join_paths joined_path first_path_segment)
set(temp_path "${first_path_segment}")
foreach(current_segment IN LISTS ARGN)
if(NOT ("${current_segment}" STREQUAL ""))
if(IS_ABSOLUTE "${current_segment}")
set(temp_path "${current_segment}")
else()
set(temp_path "${temp_path}/${current_segment}")
endif()
endif()
endforeach()
set(${joined_path} "${temp_path}" PARENT_SCOPE)
endfunction()

109
src/cpp/viper/kissfft/kfc.c Normal file
View File

@ -0,0 +1,109 @@
/*
* Copyright (c) 2003-2004, Mark Borgerding. All rights reserved.
* This file is part of KISS FFT - https://github.com/mborgerding/kissfft
*
* SPDX-License-Identifier: BSD-3-Clause
* See COPYING file for more information.
*/
#include "kfc.h"
typedef struct cached_fft *kfc_cfg;
struct cached_fft
{
int nfft;
int inverse;
kiss_fft_cfg cfg;
kfc_cfg next;
};
static kfc_cfg cache_root=NULL;
static int ncached=0;
static kiss_fft_cfg find_cached_fft(int nfft,int inverse)
{
size_t len;
kfc_cfg cur=cache_root;
kfc_cfg prev=NULL;
while ( cur ) {
if ( cur->nfft == nfft && inverse == cur->inverse )
break;/*found the right node*/
prev = cur;
cur = prev->next;
}
if (cur== NULL) {
/* no cached node found, need to create a new one*/
kiss_fft_alloc(nfft,inverse,0,&len);
#ifdef USE_SIMD
int padding = (16-sizeof(struct cached_fft)) & 15;
// make sure the cfg aligns on a 16 byte boundary
len += padding;
#endif
cur = (kfc_cfg)KISS_FFT_MALLOC((sizeof(struct cached_fft) + len ));
if (cur == NULL)
return NULL;
cur->cfg = (kiss_fft_cfg)(cur+1);
#ifdef USE_SIMD
cur->cfg = (kiss_fft_cfg) ((char*)(cur+1)+padding);
#endif
kiss_fft_alloc(nfft,inverse,cur->cfg,&len);
cur->nfft=nfft;
cur->inverse=inverse;
cur->next = NULL;
if ( prev )
prev->next = cur;
else
cache_root = cur;
++ncached;
}
return cur->cfg;
}
void kfc_cleanup(void)
{
kfc_cfg cur=cache_root;
kfc_cfg next=NULL;
while (cur){
next = cur->next;
free(cur);
cur=next;
}
ncached=0;
cache_root = NULL;
}
void kfc_fft(int nfft, const kiss_fft_cpx * fin,kiss_fft_cpx * fout)
{
kiss_fft( find_cached_fft(nfft,0),fin,fout );
}
void kfc_ifft(int nfft, const kiss_fft_cpx * fin,kiss_fft_cpx * fout)
{
kiss_fft( find_cached_fft(nfft,1),fin,fout );
}
#ifdef KFC_TEST
static void check(int nc)
{
if (ncached != nc) {
fprintf(stderr,"ncached should be %d,but it is %d\n",nc,ncached);
exit(1);
}
}
int main(void)
{
kiss_fft_cpx buf1[1024],buf2[1024];
memset(buf1,0,sizeof(buf1));
check(0);
kfc_fft(512,buf1,buf2);
check(1);
kfc_fft(512,buf1,buf2);
check(1);
kfc_ifft(512,buf1,buf2);
check(2);
kfc_cleanup();
check(0);
return 0;
}
#endif

View File

@ -0,0 +1,54 @@
/*
* Copyright (c) 2003-2004, Mark Borgerding. All rights reserved.
* This file is part of KISS FFT - https://github.com/mborgerding/kissfft
*
* SPDX-License-Identifier: BSD-3-Clause
* See COPYING file for more information.
*/
#ifndef KFC_H
#define KFC_H
#include "kiss_fft.h"
#ifdef __cplusplus
extern "C" {
#endif
/*
KFC -- Kiss FFT Cache
Not needing to deal with kiss_fft_alloc and a config
object may be handy for a lot of programs.
KFC uses the underlying KISS FFT functions, but caches the config object.
The first time kfc_fft or kfc_ifft for a given FFT size, the cfg
object is created for it. All subsequent calls use the cached
configuration object.
NOTE:
You should probably not use this if your program will be using a lot
of various sizes of FFTs. There is a linear search through the
cached objects. If you are only using one or two FFT sizes, this
will be negligible. Otherwise, you may want to use another method
of managing the cfg objects.
There is no automated cleanup of the cached objects. This could lead
to large memory usage in a program that uses a lot of *DIFFERENT*
sized FFTs. If you want to force all cached cfg objects to be freed,
call kfc_cleanup.
*/
/*forward complex FFT */
void KISS_FFT_API kfc_fft(int nfft, const kiss_fft_cpx * fin,kiss_fft_cpx * fout);
/*reverse complex FFT */
void KISS_FFT_API kfc_ifft(int nfft, const kiss_fft_cpx * fin,kiss_fft_cpx * fout);
/*free all cached objects*/
void KISS_FFT_API kfc_cleanup(void);
#ifdef __cplusplus
}
#endif
#endif

View File

@ -0,0 +1,420 @@
/*
* Copyright (c) 2003-2010, Mark Borgerding. All rights reserved.
* This file is part of KISS FFT - https://github.com/mborgerding/kissfft
*
* SPDX-License-Identifier: BSD-3-Clause
* See COPYING file for more information.
*/
#include "_kiss_fft_guts.h"
/* The guts header contains all the multiplication and addition macros that are defined for
fixed or floating point complex numbers. It also delares the kf_ internal functions.
*/
static void kf_bfly2(
kiss_fft_cpx * Fout,
const size_t fstride,
const kiss_fft_cfg st,
int m
)
{
kiss_fft_cpx * Fout2;
kiss_fft_cpx * tw1 = st->twiddles;
kiss_fft_cpx t;
Fout2 = Fout + m;
do{
C_FIXDIV(*Fout,2); C_FIXDIV(*Fout2,2);
C_MUL (t, *Fout2 , *tw1);
tw1 += fstride;
C_SUB( *Fout2 , *Fout , t );
C_ADDTO( *Fout , t );
++Fout2;
++Fout;
}while (--m);
}
static void kf_bfly4(
kiss_fft_cpx * Fout,
const size_t fstride,
const kiss_fft_cfg st,
const size_t m
)
{
kiss_fft_cpx *tw1,*tw2,*tw3;
kiss_fft_cpx scratch[6];
size_t k=m;
const size_t m2=2*m;
const size_t m3=3*m;
tw3 = tw2 = tw1 = st->twiddles;
do {
C_FIXDIV(*Fout,4); C_FIXDIV(Fout[m],4); C_FIXDIV(Fout[m2],4); C_FIXDIV(Fout[m3],4);
C_MUL(scratch[0],Fout[m] , *tw1 );
C_MUL(scratch[1],Fout[m2] , *tw2 );
C_MUL(scratch[2],Fout[m3] , *tw3 );
C_SUB( scratch[5] , *Fout, scratch[1] );
C_ADDTO(*Fout, scratch[1]);
C_ADD( scratch[3] , scratch[0] , scratch[2] );
C_SUB( scratch[4] , scratch[0] , scratch[2] );
C_SUB( Fout[m2], *Fout, scratch[3] );
tw1 += fstride;
tw2 += fstride*2;
tw3 += fstride*3;
C_ADDTO( *Fout , scratch[3] );
if(st->inverse) {
Fout[m].r = scratch[5].r - scratch[4].i;
Fout[m].i = scratch[5].i + scratch[4].r;
Fout[m3].r = scratch[5].r + scratch[4].i;
Fout[m3].i = scratch[5].i - scratch[4].r;
}else{
Fout[m].r = scratch[5].r + scratch[4].i;
Fout[m].i = scratch[5].i - scratch[4].r;
Fout[m3].r = scratch[5].r - scratch[4].i;
Fout[m3].i = scratch[5].i + scratch[4].r;
}
++Fout;
}while(--k);
}
static void kf_bfly3(
kiss_fft_cpx * Fout,
const size_t fstride,
const kiss_fft_cfg st,
size_t m
)
{
size_t k=m;
const size_t m2 = 2*m;
kiss_fft_cpx *tw1,*tw2;
kiss_fft_cpx scratch[5];
kiss_fft_cpx epi3;
epi3 = st->twiddles[fstride*m];
tw1=tw2=st->twiddles;
do{
C_FIXDIV(*Fout,3); C_FIXDIV(Fout[m],3); C_FIXDIV(Fout[m2],3);
C_MUL(scratch[1],Fout[m] , *tw1);
C_MUL(scratch[2],Fout[m2] , *tw2);
C_ADD(scratch[3],scratch[1],scratch[2]);
C_SUB(scratch[0],scratch[1],scratch[2]);
tw1 += fstride;
tw2 += fstride*2;
Fout[m].r = Fout->r - HALF_OF(scratch[3].r);
Fout[m].i = Fout->i - HALF_OF(scratch[3].i);
C_MULBYSCALAR( scratch[0] , epi3.i );
C_ADDTO(*Fout,scratch[3]);
Fout[m2].r = Fout[m].r + scratch[0].i;
Fout[m2].i = Fout[m].i - scratch[0].r;
Fout[m].r -= scratch[0].i;
Fout[m].i += scratch[0].r;
++Fout;
}while(--k);
}
static void kf_bfly5(
kiss_fft_cpx * Fout,
const size_t fstride,
const kiss_fft_cfg st,
int m
)
{
kiss_fft_cpx *Fout0,*Fout1,*Fout2,*Fout3,*Fout4;
int u;
kiss_fft_cpx scratch[13];
kiss_fft_cpx * twiddles = st->twiddles;
kiss_fft_cpx *tw;
kiss_fft_cpx ya,yb;
ya = twiddles[fstride*m];
yb = twiddles[fstride*2*m];
Fout0=Fout;
Fout1=Fout0+m;
Fout2=Fout0+2*m;
Fout3=Fout0+3*m;
Fout4=Fout0+4*m;
tw=st->twiddles;
for ( u=0; u<m; ++u ) {
C_FIXDIV( *Fout0,5); C_FIXDIV( *Fout1,5); C_FIXDIV( *Fout2,5); C_FIXDIV( *Fout3,5); C_FIXDIV( *Fout4,5);
scratch[0] = *Fout0;
C_MUL(scratch[1] ,*Fout1, tw[u*fstride]);
C_MUL(scratch[2] ,*Fout2, tw[2*u*fstride]);
C_MUL(scratch[3] ,*Fout3, tw[3*u*fstride]);
C_MUL(scratch[4] ,*Fout4, tw[4*u*fstride]);
C_ADD( scratch[7],scratch[1],scratch[4]);
C_SUB( scratch[10],scratch[1],scratch[4]);
C_ADD( scratch[8],scratch[2],scratch[3]);
C_SUB( scratch[9],scratch[2],scratch[3]);
Fout0->r += scratch[7].r + scratch[8].r;
Fout0->i += scratch[7].i + scratch[8].i;
scratch[5].r = scratch[0].r + S_MUL(scratch[7].r,ya.r) + S_MUL(scratch[8].r,yb.r);
scratch[5].i = scratch[0].i + S_MUL(scratch[7].i,ya.r) + S_MUL(scratch[8].i,yb.r);
scratch[6].r = S_MUL(scratch[10].i,ya.i) + S_MUL(scratch[9].i,yb.i);
scratch[6].i = -S_MUL(scratch[10].r,ya.i) - S_MUL(scratch[9].r,yb.i);
C_SUB(*Fout1,scratch[5],scratch[6]);
C_ADD(*Fout4,scratch[5],scratch[6]);
scratch[11].r = scratch[0].r + S_MUL(scratch[7].r,yb.r) + S_MUL(scratch[8].r,ya.r);
scratch[11].i = scratch[0].i + S_MUL(scratch[7].i,yb.r) + S_MUL(scratch[8].i,ya.r);
scratch[12].r = - S_MUL(scratch[10].i,yb.i) + S_MUL(scratch[9].i,ya.i);
scratch[12].i = S_MUL(scratch[10].r,yb.i) - S_MUL(scratch[9].r,ya.i);
C_ADD(*Fout2,scratch[11],scratch[12]);
C_SUB(*Fout3,scratch[11],scratch[12]);
++Fout0;++Fout1;++Fout2;++Fout3;++Fout4;
}
}
/* perform the butterfly for one stage of a mixed radix FFT */
static void kf_bfly_generic(
kiss_fft_cpx * Fout,
const size_t fstride,
const kiss_fft_cfg st,
int m,
int p
)
{
int u,k,q1,q;
kiss_fft_cpx * twiddles = st->twiddles;
kiss_fft_cpx t;
int Norig = st->nfft;
kiss_fft_cpx * scratch = (kiss_fft_cpx*)KISS_FFT_TMP_ALLOC(sizeof(kiss_fft_cpx)*p);
if (scratch == NULL){
KISS_FFT_ERROR("Memory allocation failed.");
return;
}
for ( u=0; u<m; ++u ) {
k=u;
for ( q1=0 ; q1<p ; ++q1 ) {
scratch[q1] = Fout[ k ];
C_FIXDIV(scratch[q1],p);
k += m;
}
k=u;
for ( q1=0 ; q1<p ; ++q1 ) {
int twidx=0;
Fout[ k ] = scratch[0];
for (q=1;q<p;++q ) {
twidx += fstride * k;
if (twidx>=Norig) twidx-=Norig;
C_MUL(t,scratch[q] , twiddles[twidx] );
C_ADDTO( Fout[ k ] ,t);
}
k += m;
}
}
KISS_FFT_TMP_FREE(scratch);
}
static
void kf_work(
kiss_fft_cpx * Fout,
const kiss_fft_cpx * f,
const size_t fstride,
int in_stride,
int * factors,
const kiss_fft_cfg st
)
{
kiss_fft_cpx * Fout_beg=Fout;
const int p=*factors++; /* the radix */
const int m=*factors++; /* stage's fft length/p */
const kiss_fft_cpx * Fout_end = Fout + p*m;
#ifdef _OPENMP
// use openmp extensions at the
// top-level (not recursive)
if (fstride==1 && p<=5 && m!=1)
{
int k;
// execute the p different work units in different threads
# pragma omp parallel for
for (k=0;k<p;++k)
kf_work( Fout +k*m, f+ fstride*in_stride*k,fstride*p,in_stride,factors,st);
// all threads have joined by this point
switch (p) {
case 2: kf_bfly2(Fout,fstride,st,m); break;
case 3: kf_bfly3(Fout,fstride,st,m); break;
case 4: kf_bfly4(Fout,fstride,st,m); break;
case 5: kf_bfly5(Fout,fstride,st,m); break;
default: kf_bfly_generic(Fout,fstride,st,m,p); break;
}
return;
}
#endif
if (m==1) {
do{
*Fout = *f;
f += fstride*in_stride;
}while(++Fout != Fout_end );
}else{
do{
// recursive call:
// DFT of size m*p performed by doing
// p instances of smaller DFTs of size m,
// each one takes a decimated version of the input
kf_work( Fout , f, fstride*p, in_stride, factors,st);
f += fstride*in_stride;
}while( (Fout += m) != Fout_end );
}
Fout=Fout_beg;
// recombine the p smaller DFTs
switch (p) {
case 2: kf_bfly2(Fout,fstride,st,m); break;
case 3: kf_bfly3(Fout,fstride,st,m); break;
case 4: kf_bfly4(Fout,fstride,st,m); break;
case 5: kf_bfly5(Fout,fstride,st,m); break;
default: kf_bfly_generic(Fout,fstride,st,m,p); break;
}
}
/* facbuf is populated by p1,m1,p2,m2, ...
where
p[i] * m[i] = m[i-1]
m0 = n */
static
void kf_factor(int n,int * facbuf)
{
int p=4;
double floor_sqrt;
floor_sqrt = floor( sqrt((double)n) );
/*factor out powers of 4, powers of 2, then any remaining primes */
do {
while (n % p) {
switch (p) {
case 4: p = 2; break;
case 2: p = 3; break;
default: p += 2; break;
}
if (p > floor_sqrt)
p = n; /* no more factors, skip to end */
}
n /= p;
*facbuf++ = p;
*facbuf++ = n;
} while (n > 1);
}
/*
*
* User-callable function to allocate all necessary storage space for the fft.
*
* The return value is a contiguous block of memory, allocated with malloc. As such,
* It can be freed with free(), rather than a kiss_fft-specific function.
* */
kiss_fft_cfg kiss_fft_alloc(int nfft,int inverse_fft,void * mem,size_t * lenmem )
{
KISS_FFT_ALIGN_CHECK(mem)
kiss_fft_cfg st=NULL;
size_t memneeded = KISS_FFT_ALIGN_SIZE_UP(sizeof(struct kiss_fft_state)
+ sizeof(kiss_fft_cpx)*(nfft-1)); /* twiddle factors*/
if ( lenmem==NULL ) {
st = ( kiss_fft_cfg)KISS_FFT_MALLOC( memneeded );
}else{
if (mem != NULL && *lenmem >= memneeded)
st = (kiss_fft_cfg)mem;
*lenmem = memneeded;
}
if (st) {
int i;
st->nfft=nfft;
st->inverse = inverse_fft;
for (i=0;i<nfft;++i) {
const double pi=3.141592653589793238462643383279502884197169399375105820974944;
double phase = -2*pi*i / nfft;
if (st->inverse)
phase *= -1;
kf_cexp(st->twiddles+i, phase );
}
kf_factor(nfft,st->factors);
}
return st;
}
void kiss_fft_stride(kiss_fft_cfg st,const kiss_fft_cpx *fin,kiss_fft_cpx *fout,int in_stride)
{
if (fin == fout) {
//NOTE: this is not really an in-place FFT algorithm.
//It just performs an out-of-place FFT into a temp buffer
if (fout == NULL){
KISS_FFT_ERROR("fout buffer NULL.");
return;
}
kiss_fft_cpx * tmpbuf = (kiss_fft_cpx*)KISS_FFT_TMP_ALLOC( sizeof(kiss_fft_cpx)*st->nfft);
if (tmpbuf == NULL){
KISS_FFT_ERROR("Memory allocation error.");
return;
}
kf_work(tmpbuf,fin,1,in_stride, st->factors,st);
memcpy(fout,tmpbuf,sizeof(kiss_fft_cpx)*st->nfft);
KISS_FFT_TMP_FREE(tmpbuf);
}else{
kf_work( fout, fin, 1,in_stride, st->factors,st );
}
}
void kiss_fft(kiss_fft_cfg cfg,const kiss_fft_cpx *fin,kiss_fft_cpx *fout)
{
kiss_fft_stride(cfg,fin,fout,1);
}
void kiss_fft_cleanup(void)
{
// nothing needed any more
}
int kiss_fft_next_fast_size(int n)
{
while(1) {
int m=n;
while ( (m%2) == 0 ) m/=2;
while ( (m%3) == 0 ) m/=3;
while ( (m%5) == 0 ) m/=5;
if (m<=1)
break; /* n is completely factorable by twos, threes, and fives */
n++;
}
return n;
}

View File

@ -0,0 +1,160 @@
/*
* Copyright (c) 2003-2010, Mark Borgerding. All rights reserved.
* This file is part of KISS FFT - https://github.com/mborgerding/kissfft
*
* SPDX-License-Identifier: BSD-3-Clause
* See COPYING file for more information.
*/
#ifndef KISS_FFT_H
#define KISS_FFT_H
#include <stdlib.h>
#include <stdio.h>
#include <math.h>
#include <string.h>
// Define KISS_FFT_SHARED macro to properly export symbols
#ifdef KISS_FFT_SHARED
# ifdef _WIN32
# ifdef KISS_FFT_BUILD
# define KISS_FFT_API __declspec(dllexport)
# else
# define KISS_FFT_API __declspec(dllimport)
# endif
# else
# define KISS_FFT_API __attribute__ ((visibility ("default")))
# endif
#else
# define KISS_FFT_API
#endif
#ifdef __cplusplus
extern "C" {
#endif
/*
ATTENTION!
If you would like a :
-- a utility that will handle the caching of fft objects
-- real-only (no imaginary time component ) FFT
-- a multi-dimensional FFT
-- a command-line utility to perform ffts
-- a command-line utility to perform fast-convolution filtering
Then see kfc.h kiss_fftr.h kiss_fftnd.h fftutil.c kiss_fastfir.c
in the tools/ directory.
*/
/* User may override KISS_FFT_MALLOC and/or KISS_FFT_FREE. */
#ifdef USE_SIMD
# include <xmmintrin.h>
# define kiss_fft_scalar __m128
# ifndef KISS_FFT_MALLOC
# define KISS_FFT_MALLOC(nbytes) _mm_malloc(nbytes,16)
# define KISS_FFT_ALIGN_CHECK(ptr)
# define KISS_FFT_ALIGN_SIZE_UP(size) ((size + 15UL) & ~0xFUL)
# endif
# ifndef KISS_FFT_FREE
# define KISS_FFT_FREE _mm_free
# endif
#else
# define KISS_FFT_ALIGN_CHECK(ptr)
# define KISS_FFT_ALIGN_SIZE_UP(size) (size)
# ifndef KISS_FFT_MALLOC
# define KISS_FFT_MALLOC malloc
# endif
# ifndef KISS_FFT_FREE
# define KISS_FFT_FREE free
# endif
#endif
#ifdef FIXED_POINT
#include <stdint.h>
# if (FIXED_POINT == 32)
# define kiss_fft_scalar int32_t
# else
# define kiss_fft_scalar int16_t
# endif
#else
# ifndef kiss_fft_scalar
/* default is float */
# define kiss_fft_scalar float
# endif
#endif
typedef struct {
kiss_fft_scalar r;
kiss_fft_scalar i;
}kiss_fft_cpx;
typedef struct kiss_fft_state* kiss_fft_cfg;
/*
* kiss_fft_alloc
*
* Initialize a FFT (or IFFT) algorithm's cfg/state buffer.
*
* typical usage: kiss_fft_cfg mycfg=kiss_fft_alloc(1024,0,NULL,NULL);
*
* The return value from fft_alloc is a cfg buffer used internally
* by the fft routine or NULL.
*
* If lenmem is NULL, then kiss_fft_alloc will allocate a cfg buffer using malloc.
* The returned value should be free()d when done to avoid memory leaks.
*
* The state can be placed in a user supplied buffer 'mem':
* If lenmem is not NULL and mem is not NULL and *lenmem is large enough,
* then the function places the cfg in mem and the size used in *lenmem
* and returns mem.
*
* If lenmem is not NULL and ( mem is NULL or *lenmem is not large enough),
* then the function returns NULL and places the minimum cfg
* buffer size in *lenmem.
* */
kiss_fft_cfg KISS_FFT_API kiss_fft_alloc(int nfft,int inverse_fft,void * mem,size_t * lenmem);
/*
* kiss_fft(cfg,in_out_buf)
*
* Perform an FFT on a complex input buffer.
* for a forward FFT,
* fin should be f[0] , f[1] , ... ,f[nfft-1]
* fout will be F[0] , F[1] , ... ,F[nfft-1]
* Note that each element is complex and can be accessed like
f[k].r and f[k].i
* */
void KISS_FFT_API kiss_fft(kiss_fft_cfg cfg,const kiss_fft_cpx *fin,kiss_fft_cpx *fout);
/*
A more generic version of the above function. It reads its input from every Nth sample.
* */
void KISS_FFT_API kiss_fft_stride(kiss_fft_cfg cfg,const kiss_fft_cpx *fin,kiss_fft_cpx *fout,int fin_stride);
/* If kiss_fft_alloc allocated a buffer, it is one contiguous
buffer and can be simply free()d when no longer needed*/
#define kiss_fft_free KISS_FFT_FREE
/*
Cleans up some memory that gets managed internally. Not necessary to call, but it might clean up
your compiler output to call this before you exit.
*/
void KISS_FFT_API kiss_fft_cleanup(void);
/*
* Returns the smallest integer k, such that k>=n and k has only "fast" factors (2,3,5)
*/
int KISS_FFT_API kiss_fft_next_fast_size(int n);
/* for real ffts, we need an even size */
#define kiss_fftr_next_fast_size_real(n) \
(kiss_fft_next_fast_size( ((n)+1)>>1)<<1)
#ifdef __cplusplus
}
#endif
#endif

View File

@ -0,0 +1,36 @@
/*
* Copyright (c) 2003-2010, Mark Borgerding. All rights reserved.
* This file is part of KISS FFT - https://github.com/mborgerding/kissfft
*
* SPDX-License-Identifier: BSD-3-Clause
* See COPYING file for more information.
*/
#ifndef kiss_fft_log_h
#define kiss_fft_log_h
#define ERROR 1
#define WARNING 2
#define INFO 3
#define DEBUG 4
#define STRINGIFY(x) #x
#define TOSTRING(x) STRINGIFY(x)
#if defined(NDEBUG)
# define KISS_FFT_LOG_MSG(severity, ...) ((void)0)
#else
# define KISS_FFT_LOG_MSG(severity, ...) \
fprintf(stderr, "[" #severity "] " __FILE__ ":" TOSTRING(__LINE__) " "); \
fprintf(stderr, __VA_ARGS__); \
fprintf(stderr, "\n")
#endif
#define KISS_FFT_ERROR(...) KISS_FFT_LOG_MSG(ERROR, __VA_ARGS__)
#define KISS_FFT_WARNING(...) KISS_FFT_LOG_MSG(WARNING, __VA_ARGS__)
#define KISS_FFT_INFO(...) KISS_FFT_LOG_MSG(INFO, __VA_ARGS__)
#define KISS_FFT_DEBUG(...) KISS_FFT_LOG_MSG(DEBUG, __VA_ARGS__)
#endif /* kiss_fft_log_h */

View File

@ -0,0 +1,188 @@
/*
* Copyright (c) 2003-2004, Mark Borgerding. All rights reserved.
* This file is part of KISS FFT - https://github.com/mborgerding/kissfft
*
* SPDX-License-Identifier: BSD-3-Clause
* See COPYING file for more information.
*/
#include "kiss_fftnd.h"
#include "_kiss_fft_guts.h"
struct kiss_fftnd_state{
int dimprod; /* dimsum would be mighty tasty right now */
int ndims;
int *dims;
kiss_fft_cfg *states; /* cfg states for each dimension */
kiss_fft_cpx * tmpbuf; /*buffer capable of hold the entire input */
};
kiss_fftnd_cfg kiss_fftnd_alloc(const int *dims,int ndims,int inverse_fft,void*mem,size_t*lenmem)
{
KISS_FFT_ALIGN_CHECK(mem)
kiss_fftnd_cfg st = NULL;
int i;
int dimprod=1;
size_t memneeded = KISS_FFT_ALIGN_SIZE_UP(sizeof(struct kiss_fftnd_state));
char * ptr = NULL;
for (i=0;i<ndims;++i) {
size_t sublen=0;
kiss_fft_alloc (dims[i], inverse_fft, NULL, &sublen);
memneeded += sublen; /* st->states[i] */
dimprod *= dims[i];
}
memneeded += KISS_FFT_ALIGN_SIZE_UP(sizeof(int) * ndims);/* st->dims */
memneeded += KISS_FFT_ALIGN_SIZE_UP(sizeof(void*) * ndims);/* st->states */
memneeded += KISS_FFT_ALIGN_SIZE_UP(sizeof(kiss_fft_cpx) * dimprod); /* st->tmpbuf */
if (lenmem == NULL) {/* allocate for the caller*/
ptr = (char *) malloc (memneeded);
} else { /* initialize supplied buffer if big enough */
if (*lenmem >= memneeded)
ptr = (char *) mem;
*lenmem = memneeded; /*tell caller how big struct is (or would be) */
}
if (!ptr)
return NULL; /*malloc failed or buffer too small */
st = (kiss_fftnd_cfg) ptr;
st->dimprod = dimprod;
st->ndims = ndims;
ptr += KISS_FFT_ALIGN_SIZE_UP(sizeof(struct kiss_fftnd_state));
st->states = (kiss_fft_cfg *)ptr;
ptr += KISS_FFT_ALIGN_SIZE_UP(sizeof(void*) * ndims);
st->dims = (int*)ptr;
ptr += KISS_FFT_ALIGN_SIZE_UP(sizeof(int) * ndims);
st->tmpbuf = (kiss_fft_cpx*)ptr;
ptr += KISS_FFT_ALIGN_SIZE_UP(sizeof(kiss_fft_cpx) * dimprod);
for (i=0;i<ndims;++i) {
size_t len;
st->dims[i] = dims[i];
kiss_fft_alloc (st->dims[i], inverse_fft, NULL, &len);
st->states[i] = kiss_fft_alloc (st->dims[i], inverse_fft, ptr,&len);
ptr += len;
}
/*
Hi there!
If you're looking at this particular code, it probably means you've got a brain-dead bounds checker
that thinks the above code overwrites the end of the array.
It doesn't.
-- Mark
P.S.
The below code might give you some warm fuzzies and help convince you.
*/
if ( ptr - (char*)st != (int)memneeded ) {
fprintf(stderr,
"################################################################################\n"
"Internal error! Memory allocation miscalculation\n"
"################################################################################\n"
);
}
return st;
}
/*
This works by tackling one dimension at a time.
In viper,
Each stage starts out by reshaping the matrix into a DixSi 2d matrix.
A Di-sized fft is taken of each column, transposing the matrix as it goes.
Here's a 3-d example:
Take a 2x3x4 matrix, laid out in memory as a contiguous buffer
[ [ [ a b c d ] [ e f g h ] [ i j k l ] ]
[ [ m n o p ] [ q r s t ] [ u v w x ] ] ]
Stage 0 ( D=2): treat the buffer as a 2x12 matrix
[ [a b ... k l]
[m n ... w x] ]
FFT each column with size 2.
Transpose the matrix at the same time using kiss_fft_stride.
[ [ a+m a-m ]
[ b+n b-n]
...
[ k+w k-w ]
[ l+x l-x ] ]
Note fft([x y]) == [x+y x-y]
Stage 1 ( D=3) treats the buffer (the output of stage D=2) as an 3x8 matrix,
[ [ a+m a-m b+n b-n c+o c-o d+p d-p ]
[ e+q e-q f+r f-r g+s g-s h+t h-t ]
[ i+u i-u j+v j-v k+w k-w l+x l-x ] ]
And perform FFTs (size=3) on each of the columns as above, transposing
the matrix as it goes. The output of stage 1 is
(Legend: ap = [ a+m e+q i+u ]
am = [ a-m e-q i-u ] )
[ [ sum(ap) fft(ap)[0] fft(ap)[1] ]
[ sum(am) fft(am)[0] fft(am)[1] ]
[ sum(bp) fft(bp)[0] fft(bp)[1] ]
[ sum(bm) fft(bm)[0] fft(bm)[1] ]
[ sum(cp) fft(cp)[0] fft(cp)[1] ]
[ sum(cm) fft(cm)[0] fft(cm)[1] ]
[ sum(dp) fft(dp)[0] fft(dp)[1] ]
[ sum(dm) fft(dm)[0] fft(dm)[1] ] ]
Stage 2 ( D=4) treats this buffer as a 4*6 matrix,
[ [ sum(ap) fft(ap)[0] fft(ap)[1] sum(am) fft(am)[0] fft(am)[1] ]
[ sum(bp) fft(bp)[0] fft(bp)[1] sum(bm) fft(bm)[0] fft(bm)[1] ]
[ sum(cp) fft(cp)[0] fft(cp)[1] sum(cm) fft(cm)[0] fft(cm)[1] ]
[ sum(dp) fft(dp)[0] fft(dp)[1] sum(dm) fft(dm)[0] fft(dm)[1] ] ]
Then FFTs each column, transposing as it goes.
The resulting matrix is the 3d FFT of the 2x3x4 input matrix.
Note as a sanity check that the first element of the final
stage's output (DC term) is
sum( [ sum(ap) sum(bp) sum(cp) sum(dp) ] )
, i.e. the summation of all 24 input elements.
*/
void kiss_fftnd(kiss_fftnd_cfg st,const kiss_fft_cpx *fin,kiss_fft_cpx *fout)
{
int i,k;
const kiss_fft_cpx * bufin=fin;
kiss_fft_cpx * bufout;
/*arrange it so the last bufout == fout*/
if ( st->ndims & 1 ) {
bufout = fout;
if (fin==fout) {
memcpy( st->tmpbuf, fin, sizeof(kiss_fft_cpx) * st->dimprod );
bufin = st->tmpbuf;
}
}else
bufout = st->tmpbuf;
for ( k=0; k < st->ndims; ++k) {
int curdim = st->dims[k];
int stride = st->dimprod / curdim;
for ( i=0 ; i<stride ; ++i )
kiss_fft_stride( st->states[k], bufin+i , bufout+i*curdim, stride );
/*toggle back and forth between the two buffers*/
if (bufout == st->tmpbuf){
bufout = fout;
bufin = st->tmpbuf;
}else{
bufout = st->tmpbuf;
bufin = fout;
}
}
}

View File

@ -0,0 +1,26 @@
/*
* Copyright (c) 2003-2004, Mark Borgerding. All rights reserved.
* This file is part of KISS FFT - https://github.com/mborgerding/kissfft
*
* SPDX-License-Identifier: BSD-3-Clause
* See COPYING file for more information.
*/
#ifndef KISS_FFTND_H
#define KISS_FFTND_H
#include "kiss_fft.h"
#ifdef __cplusplus
extern "C" {
#endif
typedef struct kiss_fftnd_state * kiss_fftnd_cfg;
kiss_fftnd_cfg KISS_FFT_API kiss_fftnd_alloc(const int *dims,int ndims,int inverse_fft,void*mem,size_t*lenmem);
void KISS_FFT_API kiss_fftnd(kiss_fftnd_cfg cfg,const kiss_fft_cpx *fin,kiss_fft_cpx *fout);
#ifdef __cplusplus
}
#endif
#endif

View File

@ -0,0 +1,120 @@
/*
* Copyright (c) 2003-2004, Mark Borgerding. All rights reserved.
* This file is part of KISS FFT - https://github.com/mborgerding/kissfft
*
* SPDX-License-Identifier: BSD-3-Clause
* See COPYING file for more information.
*/
#include "kiss_fftndr.h"
#include "_kiss_fft_guts.h"
#define MAX(x,y) ( ( (x)<(y) )?(y):(x) )
struct kiss_fftndr_state
{
int dimReal;
int dimOther;
kiss_fftr_cfg cfg_r;
kiss_fftnd_cfg cfg_nd;
void * tmpbuf;
};
static int prod(const int *dims, int ndims)
{
int x=1;
while (ndims--)
x *= *dims++;
return x;
}
kiss_fftndr_cfg kiss_fftndr_alloc(const int *dims,int ndims,int inverse_fft,void*mem,size_t*lenmem)
{
KISS_FFT_ALIGN_CHECK(mem)
kiss_fftndr_cfg st = NULL;
size_t nr=0 , nd=0,ntmp=0;
int dimReal = dims[ndims-1];
int dimOther = prod(dims,ndims-1);
size_t memneeded;
char * ptr = NULL;
(void)kiss_fftr_alloc(dimReal,inverse_fft,NULL,&nr);
(void)kiss_fftnd_alloc(dims,ndims-1,inverse_fft,NULL,&nd);
ntmp =
MAX( 2*dimOther , dimReal+2) * sizeof(kiss_fft_scalar) // freq buffer for one pass
+ dimOther*(dimReal+2) * sizeof(kiss_fft_scalar); // large enough to hold entire input in case of in-place
memneeded = KISS_FFT_ALIGN_SIZE_UP(sizeof( struct kiss_fftndr_state )) + KISS_FFT_ALIGN_SIZE_UP(nr) + KISS_FFT_ALIGN_SIZE_UP(nd) + KISS_FFT_ALIGN_SIZE_UP(ntmp);
if (lenmem==NULL) {
ptr = (char*) malloc(memneeded);
}else{
if (*lenmem >= memneeded)
ptr = (char *)mem;
*lenmem = memneeded;
}
if (ptr==NULL)
return NULL;
st = (kiss_fftndr_cfg) ptr;
memset( st , 0 , memneeded);
ptr += KISS_FFT_ALIGN_SIZE_UP(sizeof(struct kiss_fftndr_state));
st->dimReal = dimReal;
st->dimOther = dimOther;
st->cfg_r = kiss_fftr_alloc( dimReal,inverse_fft,ptr,&nr);
ptr += KISS_FFT_ALIGN_SIZE_UP(nr);
st->cfg_nd = kiss_fftnd_alloc(dims,ndims-1,inverse_fft, ptr,&nd);
ptr += KISS_FFT_ALIGN_SIZE_UP(nd);
st->tmpbuf = ptr;
return st;
}
void kiss_fftndr(kiss_fftndr_cfg st,const kiss_fft_scalar *timedata,kiss_fft_cpx *freqdata)
{
int k1,k2;
int dimReal = st->dimReal;
int dimOther = st->dimOther;
int nrbins = dimReal/2+1;
kiss_fft_cpx * tmp1 = (kiss_fft_cpx*)st->tmpbuf;
kiss_fft_cpx * tmp2 = tmp1 + MAX(nrbins,dimOther);
// timedata is N0 x N1 x ... x Nk real
// take a real chunk of data, fft it and place the output at correct intervals
for (k1=0;k1<dimOther;++k1) {
kiss_fftr( st->cfg_r, timedata + k1*dimReal , tmp1 ); // tmp1 now holds nrbins complex points
for (k2=0;k2<nrbins;++k2)
tmp2[ k2*dimOther+k1 ] = tmp1[k2];
}
for (k2=0;k2<nrbins;++k2) {
kiss_fftnd(st->cfg_nd, tmp2+k2*dimOther, tmp1); // tmp1 now holds dimOther complex points
for (k1=0;k1<dimOther;++k1)
freqdata[ k1*(nrbins) + k2] = tmp1[k1];
}
}
void kiss_fftndri(kiss_fftndr_cfg st,const kiss_fft_cpx *freqdata,kiss_fft_scalar *timedata)
{
int k1,k2;
int dimReal = st->dimReal;
int dimOther = st->dimOther;
int nrbins = dimReal/2+1;
kiss_fft_cpx * tmp1 = (kiss_fft_cpx*)st->tmpbuf;
kiss_fft_cpx * tmp2 = tmp1 + MAX(nrbins,dimOther);
for (k2=0;k2<nrbins;++k2) {
for (k1=0;k1<dimOther;++k1)
tmp1[k1] = freqdata[ k1*(nrbins) + k2 ];
kiss_fftnd(st->cfg_nd, tmp1, tmp2+k2*dimOther);
}
for (k1=0;k1<dimOther;++k1) {
for (k2=0;k2<nrbins;++k2)
tmp1[k2] = tmp2[ k2*dimOther+k1 ];
kiss_fftri( st->cfg_r,tmp1,timedata + k1*dimReal);
}
}

View File

@ -0,0 +1,55 @@
/*
* Copyright (c) 2003-2004, Mark Borgerding. All rights reserved.
* This file is part of KISS FFT - https://github.com/mborgerding/kissfft
*
* SPDX-License-Identifier: BSD-3-Clause
* See COPYING file for more information.
*/
#ifndef KISS_NDR_H
#define KISS_NDR_H
#include "kiss_fft.h"
#include "kiss_fftr.h"
#include "kiss_fftnd.h"
#ifdef __cplusplus
extern "C" {
#endif
typedef struct kiss_fftndr_state *kiss_fftndr_cfg;
kiss_fftndr_cfg KISS_FFT_API kiss_fftndr_alloc(const int *dims,int ndims,int inverse_fft,void*mem,size_t*lenmem);
/*
dims[0] must be even
If you don't care to allocate space, use mem = lenmem = NULL
*/
void KISS_FFT_API kiss_fftndr(
kiss_fftndr_cfg cfg,
const kiss_fft_scalar *timedata,
kiss_fft_cpx *freqdata);
/*
input timedata has dims[0] X dims[1] X ... X dims[ndims-1] scalar points
output freqdata has dims[0] X dims[1] X ... X dims[ndims-1]/2+1 complex points
*/
void KISS_FFT_API kiss_fftndri(
kiss_fftndr_cfg cfg,
const kiss_fft_cpx *freqdata,
kiss_fft_scalar *timedata);
/*
input and output dimensions are the exact opposite of kiss_fftndr
*/
#define kiss_fftndr_free free
#ifdef __cplusplus
}
#endif
#endif

View File

@ -0,0 +1,155 @@
/*
* Copyright (c) 2003-2004, Mark Borgerding. All rights reserved.
* This file is part of KISS FFT - https://github.com/mborgerding/kissfft
*
* SPDX-License-Identifier: BSD-3-Clause
* See COPYING file for more information.
*/
#include "kiss_fftr.h"
#include "_kiss_fft_guts.h"
struct kiss_fftr_state{
kiss_fft_cfg substate;
kiss_fft_cpx * tmpbuf;
kiss_fft_cpx * super_twiddles;
#ifdef USE_SIMD
void * pad;
#endif
};
kiss_fftr_cfg kiss_fftr_alloc(int nfft,int inverse_fft,void * mem,size_t * lenmem)
{
KISS_FFT_ALIGN_CHECK(mem)
int i;
kiss_fftr_cfg st = NULL;
size_t subsize = 0, memneeded;
if (nfft & 1) {
KISS_FFT_ERROR("Real FFT optimization must be even.");
return NULL;
}
nfft >>= 1;
kiss_fft_alloc (nfft, inverse_fft, NULL, &subsize);
memneeded = sizeof(struct kiss_fftr_state) + subsize + sizeof(kiss_fft_cpx) * ( nfft * 3 / 2);
if (lenmem == NULL) {
st = (kiss_fftr_cfg) KISS_FFT_MALLOC (memneeded);
} else {
if (*lenmem >= memneeded)
st = (kiss_fftr_cfg) mem;
*lenmem = memneeded;
}
if (!st)
return NULL;
st->substate = (kiss_fft_cfg) (st + 1); /*just beyond kiss_fftr_state struct */
st->tmpbuf = (kiss_fft_cpx *) (((char *) st->substate) + subsize);
st->super_twiddles = st->tmpbuf + nfft;
kiss_fft_alloc(nfft, inverse_fft, st->substate, &subsize);
for (i = 0; i < nfft/2; ++i) {
double phase =
-3.14159265358979323846264338327 * ((double) (i+1) / nfft + .5);
if (inverse_fft)
phase *= -1;
kf_cexp (st->super_twiddles+i,phase);
}
return st;
}
void kiss_fftr(kiss_fftr_cfg st,const kiss_fft_scalar *timedata,kiss_fft_cpx *freqdata)
{
/* input buffer timedata is stored row-wise */
int k,ncfft;
kiss_fft_cpx fpnk,fpk,f1k,f2k,tw,tdc;
if ( st->substate->inverse) {
KISS_FFT_ERROR("kiss fft usage error: improper alloc");
return;/* The caller did not call the correct function */
}
ncfft = st->substate->nfft;
/*perform the parallel fft of two real signals packed in real,imag*/
kiss_fft( st->substate , (const kiss_fft_cpx*)timedata, st->tmpbuf );
/* The real part of the DC element of the frequency spectrum in st->tmpbuf
* contains the sum of the even-numbered elements of the input time sequence
* The imag part is the sum of the odd-numbered elements
*
* The sum of tdc.r and tdc.i is the sum of the input time sequence.
* yielding DC of input time sequence
* The difference of tdc.r - tdc.i is the sum of the input (dot product) [1,-1,1,-1...
* yielding Nyquist bin of input time sequence
*/
tdc.r = st->tmpbuf[0].r;
tdc.i = st->tmpbuf[0].i;
C_FIXDIV(tdc,2);
CHECK_OVERFLOW_OP(tdc.r ,+, tdc.i);
CHECK_OVERFLOW_OP(tdc.r ,-, tdc.i);
freqdata[0].r = tdc.r + tdc.i;
freqdata[ncfft].r = tdc.r - tdc.i;
#ifdef USE_SIMD
freqdata[ncfft].i = freqdata[0].i = _mm_set1_ps(0);
#else
freqdata[ncfft].i = freqdata[0].i = 0;
#endif
for ( k=1;k <= ncfft/2 ; ++k ) {
fpk = st->tmpbuf[k];
fpnk.r = st->tmpbuf[ncfft-k].r;
fpnk.i = - st->tmpbuf[ncfft-k].i;
C_FIXDIV(fpk,2);
C_FIXDIV(fpnk,2);
C_ADD( f1k, fpk , fpnk );
C_SUB( f2k, fpk , fpnk );
C_MUL( tw , f2k , st->super_twiddles[k-1]);
freqdata[k].r = HALF_OF(f1k.r + tw.r);
freqdata[k].i = HALF_OF(f1k.i + tw.i);
freqdata[ncfft-k].r = HALF_OF(f1k.r - tw.r);
freqdata[ncfft-k].i = HALF_OF(tw.i - f1k.i);
}
}
void kiss_fftri(kiss_fftr_cfg st,const kiss_fft_cpx *freqdata,kiss_fft_scalar *timedata)
{
/* input buffer timedata is stored row-wise */
int k, ncfft;
if (st->substate->inverse == 0) {
KISS_FFT_ERROR("kiss fft usage error: improper alloc");
return;/* The caller did not call the correct function */
}
ncfft = st->substate->nfft;
st->tmpbuf[0].r = freqdata[0].r + freqdata[ncfft].r;
st->tmpbuf[0].i = freqdata[0].r - freqdata[ncfft].r;
C_FIXDIV(st->tmpbuf[0],2);
for (k = 1; k <= ncfft / 2; ++k) {
kiss_fft_cpx fk, fnkc, fek, fok, tmp;
fk = freqdata[k];
fnkc.r = freqdata[ncfft - k].r;
fnkc.i = -freqdata[ncfft - k].i;
C_FIXDIV( fk , 2 );
C_FIXDIV( fnkc , 2 );
C_ADD (fek, fk, fnkc);
C_SUB (tmp, fk, fnkc);
C_MUL (fok, tmp, st->super_twiddles[k-1]);
C_ADD (st->tmpbuf[k], fek, fok);
C_SUB (st->tmpbuf[ncfft - k], fek, fok);
#ifdef USE_SIMD
st->tmpbuf[ncfft - k].i *= _mm_set1_ps(-1.0);
#else
st->tmpbuf[ncfft - k].i *= -1;
#endif
}
kiss_fft (st->substate, st->tmpbuf, (kiss_fft_cpx *) timedata);
}

View File

@ -0,0 +1,54 @@
/*
* Copyright (c) 2003-2004, Mark Borgerding. All rights reserved.
* This file is part of KISS FFT - https://github.com/mborgerding/kissfft
*
* SPDX-License-Identifier: BSD-3-Clause
* See COPYING file for more information.
*/
#ifndef KISS_FTR_H
#define KISS_FTR_H
#include "kiss_fft.h"
#ifdef __cplusplus
extern "C" {
#endif
/*
Real optimized version can save about 45% cpu time vs. complex fft of a real seq.
*/
typedef struct kiss_fftr_state *kiss_fftr_cfg;
kiss_fftr_cfg KISS_FFT_API kiss_fftr_alloc(int nfft,int inverse_fft,void * mem, size_t * lenmem);
/*
nfft must be even
If you don't care to allocate space, use mem = lenmem = NULL
*/
void KISS_FFT_API kiss_fftr(kiss_fftr_cfg cfg,const kiss_fft_scalar *timedata,kiss_fft_cpx *freqdata);
/*
input timedata has nfft scalar points
output freqdata has nfft/2+1 complex points
*/
void KISS_FFT_API kiss_fftri(kiss_fftr_cfg cfg,const kiss_fft_cpx *freqdata,kiss_fft_scalar *timedata);
/*
input freqdata has nfft/2+1 complex points
output timedata has nfft scalar points
*/
#define kiss_fftr_free KISS_FFT_FREE
#ifdef __cplusplus
}
#endif
#endif

View File

@ -0,0 +1,82 @@
# kissfft-config.ccmake accept the following components:
#
# SHARED/STATIC:
# This components allows one to choose a shared/static kissfft library.
# The default is selected by BUILD_SHARED_LIBS.
# They are to be used exclusively. Using them together is an error.
#
# example:
# find_package(kissfft CONFIG REQUIRED COMPONENTS STATIC)
#
# simd/int16/int32/float/double:
# This components allows one to choose the datatype.
# When using this component, the target kissfft::kissfft becomes available.
# When not using this component, you will have to choose the correct kissfft target.
#
# example:
# find_package(kissfft CONFIG REQUIRED)
# # - kissfft::kissfft-float, kissfft::kissfft-int32_t/ ... are available (if they are installed)
# # - kissfft::kissfft is not available,
#
# find_package(kissfft CONFIG REQUIRED COMPONENTS int32_t)
# # - kissfft::kissfft-float, kissfft::kissfft-int32_t/ ... are available (if they are installed)
# # - kissfft::kissfft is available (as an alias for kissfft::kissfft-int32_t),
@PACKAGE_INIT@
cmake_minimum_required(VERSION 3.3)
# Set include glob of config files using SHARED/static component, BUILD_SHARED_LIBS by default
set(_kissfft_shared_detected OFF)
set(_kissfft_shared ${BUILD_SHARED_LIBS})
if("SHARED" IN_LIST kissfft_FIND_COMPONENTS)
set(_kissfft_shared_detected ON)
set(_kissfft_shared ON)
endif()
if("STATIC" IN_LIST kissfft_FIND_COMPONENTS)
if(_kissfft_shared_detected)
message(FATAL_ERROR "SHARED and STATIC components cannot be used together")
endif()
set(_kissfft_shared_detected ON)
set(_kissfft_shared OFF)
endif()
if(_kissfft_shared)
set(_kissfft_config_glob "kissfft-*-shared-targets.cmake")
else()
set(_kissfft_config_glob "kissfft-*-static-targets.cmake")
endif()
# Load information for all configured kissfft
get_filename_component(_DIR "${CMAKE_CURRENT_LIST_FILE}" PATH)
file(GLOB CONFIG_FILES "${_DIR}/${_kissfft_config_glob}")
foreach(f ${CONFIG_FILES})
include(${f})
endforeach()
# If a datatype component is passed, create kissfft::kissfft
set(_kissfft_datatype_detected)
foreach(_kissfft_datatype simd int16 int32 float double)
if(_kissfft_datatype IN_LIST kissfft_FIND_COMPONENTS)
if(_kissfft_datatype_detected)
message(FATAL_ERROR "Cannot define datatype COMPONENT twice: ${_kissfft_datatype_detected} and ${_kissfft_datatype}")
endif()
set(_kissfft_datatype_detected ${_kissfft_datatype})
endif()
endforeach()
if(_kissfft_datatype_detected)
if(NOT TARGET kissfft::kissfft-${_kissfft_datatype_detected})
message(FATAL_ERROR "kissfft with datatype=${_kissfft_datatype_detected} is not installed")
endif()
if(TARGET kissfft::kissfft)
message(SEND_ERROR "kissfft::kissfft already exists. You cannot use 2 find_package's with datatype that are visible to eachother.")
else()
add_library(kissfft::kissfft INTERFACE IMPORTED)
set_property(TARGET kissfft::kissfft PROPERTY INTERFACE_LINK_LIBRARIES kissfft::kissfft-${_kissfft_datatype_detected})
endif()
endif()
set(kissfft_FOUND ON)
set(KISSFFT_VERSION @kissfft_VERSION@)

View File

@ -0,0 +1,361 @@
/*
* Copyright (c) 2003-2010, Mark Borgerding. All rights reserved.
* This file is part of KISS FFT - https://github.com/mborgerding/kissfft
*
* SPDX-License-Identifier: BSD-3-Clause
* See COPYING file for more information.
*/
#ifndef KISSFFT_CLASS_HH
#define KISSFFT_CLASS_HH
#include <complex>
#include <utility>
#include <vector>
template <typename scalar_t>
class kissfft
{
public:
typedef std::complex<scalar_t> cpx_t;
kissfft( const std::size_t nfft,
const bool inverse )
:_nfft(nfft)
,_inverse(inverse)
{
// fill twiddle factors
_twiddles.resize(_nfft);
const scalar_t phinc = (_inverse?2:-2)* std::acos( (scalar_t) -1) / _nfft;
for (std::size_t i=0;i<_nfft;++i)
_twiddles[i] = std::exp( cpx_t(0,i*phinc) );
//factorize
//start factoring out 4's, then 2's, then 3,5,7,9,...
std::size_t n= _nfft;
std::size_t p=4;
do {
while (n % p) {
switch (p) {
case 4: p = 2; break;
case 2: p = 3; break;
default: p += 2; break;
}
if (p*p>n)
p = n;// no more factors
}
n /= p;
_stageRadix.push_back(p);
_stageRemainder.push_back(n);
}while(n>1);
}
/// Changes the FFT-length and/or the transform direction.
///
/// @post The @c kissfft object will be in the same state as if it
/// had been newly constructed with the passed arguments.
/// However, the implementation may be faster than constructing a
/// new fft object.
void assign( const std::size_t nfft,
const bool inverse )
{
if ( nfft != _nfft )
{
kissfft tmp( nfft, inverse ); // O(n) time.
std::swap( tmp, *this ); // this is O(1) in C++11, O(n) otherwise.
}
else if ( inverse != _inverse )
{
// conjugate the twiddle factors.
for ( typename std::vector<cpx_t>::iterator it = _twiddles.begin();
it != _twiddles.end(); ++it )
it->imag( -it->imag() );
}
}
/// Calculates the complex Discrete Fourier Transform.
///
/// The size of the passed arrays must be passed in the constructor.
/// The sum of the squares of the absolute values in the @c dst
/// array will be @c N times the sum of the squares of the absolute
/// values in the @c src array, where @c N is the size of the array.
/// In other words, the l_2 norm of the resulting array will be
/// @c sqrt(N) times as big as the l_2 norm of the input array.
/// This is also the case when the inverse flag is set in the
/// constructor. Hence when applying the same transform twice, but with
/// the inverse flag changed the second time, then the result will
/// be equal to the original input times @c N.
void transform(const cpx_t * fft_in, cpx_t * fft_out, const std::size_t stage = 0, const std::size_t fstride = 1, const std::size_t in_stride = 1) const
{
const std::size_t p = _stageRadix[stage];
const std::size_t m = _stageRemainder[stage];
cpx_t * const Fout_beg = fft_out;
cpx_t * const Fout_end = fft_out + p*m;
if (m==1) {
do{
*fft_out = *fft_in;
fft_in += fstride*in_stride;
}while(++fft_out != Fout_end );
}else{
do{
// recursive call:
// DFT of size m*p performed by doing
// p instances of smaller DFTs of size m,
// each one takes a decimated version of the input
transform(fft_in, fft_out, stage+1, fstride*p,in_stride);
fft_in += fstride*in_stride;
}while( (fft_out += m) != Fout_end );
}
fft_out=Fout_beg;
// recombine the p smaller DFTs
switch (p) {
case 2: kf_bfly2(fft_out,fstride,m); break;
case 3: kf_bfly3(fft_out,fstride,m); break;
case 4: kf_bfly4(fft_out,fstride,m); break;
case 5: kf_bfly5(fft_out,fstride,m); break;
default: kf_bfly_generic(fft_out,fstride,m,p); break;
}
}
/// Calculates the Discrete Fourier Transform (DFT) of a real input
/// of size @c 2*N.
///
/// The 0-th and N-th value of the DFT are real numbers. These are
/// stored in @c dst[0].real() and @c dst[0].imag() respectively.
/// The remaining DFT values up to the index N-1 are stored in
/// @c dst[1] to @c dst[N-1].
/// The other half of the DFT values can be calculated from the
/// symmetry relation
/// @code
/// DFT(src)[2*N-k] == conj( DFT(src)[k] );
/// @endcode
/// The same scaling factors as in @c transform() apply.
///
/// @note For this to work, the types @c scalar_t and @c cpx_t
/// must fulfill the following requirements:
///
/// For any object @c z of type @c cpx_t,
/// @c reinterpret_cast<scalar_t(&)[2]>(z)[0] is the real part of @c z and
/// @c reinterpret_cast<scalar_t(&)[2]>(z)[1] is the imaginary part of @c z.
/// For any pointer to an element of an array of @c cpx_t named @c p
/// and any valid array index @c i, @c reinterpret_cast<T*>(p)[2*i]
/// is the real part of the complex number @c p[i], and
/// @c reinterpret_cast<T*>(p)[2*i+1] is the imaginary part of the
/// complex number @c p[i].
///
/// Since C++11, these requirements are guaranteed to be satisfied for
/// @c scalar_ts being @c float, @c double or @c long @c double
/// together with @c cpx_t being @c std::complex<scalar_t>.
void transform_real( const scalar_t * const src,
cpx_t * const dst ) const
{
const std::size_t N = _nfft;
if ( N == 0 )
return;
// perform complex FFT
transform( reinterpret_cast<const cpx_t*>(src), dst );
// post processing for k = 0 and k = N
dst[0] = cpx_t( dst[0].real() + dst[0].imag(),
dst[0].real() - dst[0].imag() );
// post processing for all the other k = 1, 2, ..., N-1
const scalar_t pi = std::acos( (scalar_t) -1);
const scalar_t half_phi_inc = ( _inverse ? pi : -pi ) / N;
const cpx_t twiddle_mul = std::exp( cpx_t(0, half_phi_inc) );
for ( std::size_t k = 1; 2*k < N; ++k )
{
const cpx_t w = (scalar_t)0.5 * cpx_t(
dst[k].real() + dst[N-k].real(),
dst[k].imag() - dst[N-k].imag() );
const cpx_t z = (scalar_t)0.5 * cpx_t(
dst[k].imag() + dst[N-k].imag(),
-dst[k].real() + dst[N-k].real() );
const cpx_t twiddle =
k % 2 == 0 ?
_twiddles[k/2] :
_twiddles[k/2] * twiddle_mul;
dst[ k] = w + twiddle * z;
dst[N-k] = std::conj( w - twiddle * z );
}
if ( N % 2 == 0 )
dst[N/2] = std::conj( dst[N/2] );
}
private:
void kf_bfly2( cpx_t * Fout, const size_t fstride, const std::size_t m) const
{
for (std::size_t k=0;k<m;++k) {
const cpx_t t = Fout[m+k] * _twiddles[k*fstride];
Fout[m+k] = Fout[k] - t;
Fout[k] += t;
}
}
void kf_bfly3( cpx_t * Fout, const std::size_t fstride, const std::size_t m) const
{
std::size_t k=m;
const std::size_t m2 = 2*m;
const cpx_t *tw1,*tw2;
cpx_t scratch[5];
const cpx_t epi3 = _twiddles[fstride*m];
tw1=tw2=&_twiddles[0];
do{
scratch[1] = Fout[m] * *tw1;
scratch[2] = Fout[m2] * *tw2;
scratch[3] = scratch[1] + scratch[2];
scratch[0] = scratch[1] - scratch[2];
tw1 += fstride;
tw2 += fstride*2;
Fout[m] = Fout[0] - scratch[3]*scalar_t(0.5);
scratch[0] *= epi3.imag();
Fout[0] += scratch[3];
Fout[m2] = cpx_t( Fout[m].real() + scratch[0].imag() , Fout[m].imag() - scratch[0].real() );
Fout[m] += cpx_t( -scratch[0].imag(),scratch[0].real() );
++Fout;
}while(--k);
}
void kf_bfly4( cpx_t * const Fout, const std::size_t fstride, const std::size_t m) const
{
cpx_t scratch[7];
const scalar_t negative_if_inverse = _inverse ? -1 : +1;
for (std::size_t k=0;k<m;++k) {
scratch[0] = Fout[k+ m] * _twiddles[k*fstride ];
scratch[1] = Fout[k+2*m] * _twiddles[k*fstride*2];
scratch[2] = Fout[k+3*m] * _twiddles[k*fstride*3];
scratch[5] = Fout[k] - scratch[1];
Fout[k] += scratch[1];
scratch[3] = scratch[0] + scratch[2];
scratch[4] = scratch[0] - scratch[2];
scratch[4] = cpx_t( scratch[4].imag()*negative_if_inverse ,
-scratch[4].real()*negative_if_inverse );
Fout[k+2*m] = Fout[k] - scratch[3];
Fout[k ]+= scratch[3];
Fout[k+ m] = scratch[5] + scratch[4];
Fout[k+3*m] = scratch[5] - scratch[4];
}
}
void kf_bfly5( cpx_t * const Fout, const std::size_t fstride, const std::size_t m) const
{
cpx_t *Fout0,*Fout1,*Fout2,*Fout3,*Fout4;
cpx_t scratch[13];
const cpx_t ya = _twiddles[fstride*m];
const cpx_t yb = _twiddles[fstride*2*m];
Fout0=Fout;
Fout1=Fout0+m;
Fout2=Fout0+2*m;
Fout3=Fout0+3*m;
Fout4=Fout0+4*m;
for ( std::size_t u=0; u<m; ++u ) {
scratch[0] = *Fout0;
scratch[1] = *Fout1 * _twiddles[ u*fstride];
scratch[2] = *Fout2 * _twiddles[2*u*fstride];
scratch[3] = *Fout3 * _twiddles[3*u*fstride];
scratch[4] = *Fout4 * _twiddles[4*u*fstride];
scratch[7] = scratch[1] + scratch[4];
scratch[10]= scratch[1] - scratch[4];
scratch[8] = scratch[2] + scratch[3];
scratch[9] = scratch[2] - scratch[3];
*Fout0 += scratch[7];
*Fout0 += scratch[8];
scratch[5] = scratch[0] + cpx_t(
scratch[7].real()*ya.real() + scratch[8].real()*yb.real(),
scratch[7].imag()*ya.real() + scratch[8].imag()*yb.real()
);
scratch[6] = cpx_t(
scratch[10].imag()*ya.imag() + scratch[9].imag()*yb.imag(),
-scratch[10].real()*ya.imag() - scratch[9].real()*yb.imag()
);
*Fout1 = scratch[5] - scratch[6];
*Fout4 = scratch[5] + scratch[6];
scratch[11] = scratch[0] +
cpx_t(
scratch[7].real()*yb.real() + scratch[8].real()*ya.real(),
scratch[7].imag()*yb.real() + scratch[8].imag()*ya.real()
);
scratch[12] = cpx_t(
-scratch[10].imag()*yb.imag() + scratch[9].imag()*ya.imag(),
scratch[10].real()*yb.imag() - scratch[9].real()*ya.imag()
);
*Fout2 = scratch[11] + scratch[12];
*Fout3 = scratch[11] - scratch[12];
++Fout0;
++Fout1;
++Fout2;
++Fout3;
++Fout4;
}
}
/* perform the butterfly for one stage of a mixed radix FFT */
void kf_bfly_generic(
cpx_t * const Fout,
const size_t fstride,
const std::size_t m,
const std::size_t p
) const
{
const cpx_t * twiddles = &_twiddles[0];
if(p > _scratchbuf.size()) _scratchbuf.resize(p);
for ( std::size_t u=0; u<m; ++u ) {
std::size_t k = u;
for ( std::size_t q1=0 ; q1<p ; ++q1 ) {
_scratchbuf[q1] = Fout[ k ];
k += m;
}
k=u;
for ( std::size_t q1=0 ; q1<p ; ++q1 ) {
std::size_t twidx=0;
Fout[ k ] = _scratchbuf[0];
for ( std::size_t q=1;q<p;++q ) {
twidx += fstride * k;
if (twidx>=_nfft)
twidx-=_nfft;
Fout[ k ] += _scratchbuf[q] * twiddles[twidx];
}
k += m;
}
}
}
std::size_t _nfft;
bool _inverse;
std::vector<cpx_t> _twiddles;
std::vector<std::size_t> _stageRadix;
std::vector<std::size_t> _stageRemainder;
mutable std::vector<cpx_t> _scratchbuf;
};
#endif

View File

@ -0,0 +1,10 @@
prefix=@PKGCONFIG_KISSFFT_PREFIX@
libdir=@PKGCONFIG_KISSFFT_LIBDIR@
includedir=@PKGCONFIG_KISSFFT_INCLUDEDIR@
Name: kissfft
Description: A Fast Fourier Transform (FFT) library that tries to Keep it Simple, Stupid
Version: @PKGCONFIG_KISSFFT_VERSION@
Libs: @PKG_OPENMP@ -L${libdir} -l@KISSFFT_OUTPUT_NAME@
Cflags: -I@PKGCONFIG_KISSFFT_PKGINCLUDEDIR@ @PKG_KISSFFT_DEFS@

View File

@ -0,0 +1,304 @@
#ifndef KISSFFT_I32_CLASS_HH
#define KISSFFT_I32_CLASS_HH
#include <complex>
#include <utility>
#include <vector>
// TODO1: substitute complex<type> (behaviour not defined for nonfloats), should be faster
// TODO2: use std:: namespace
// TODO3: make unittests for all ffts (c, cpp, i32)
template <typename DType>
struct complex_s
{
DType real;
DType imag;
};
class kissfft_i32
{
private:
using scalar_type = int32_t;
using cpx_type = complex<int32_t>;
scalar_type _scale_factor;
std::size_t _nfft;
bool _inverse;
std::vector<cpx_type> _twiddles;
std::vector<std::size_t> _stageRadix;
std::vector<std::size_t> _stageRemainder;
public:
// scale_factor: upscale twiddle-factors otherwise they lie between 0..1 (out of range for integer) --> fixed point math
kissfft_i32(const std::size_t nfft, const bool inverse, const double scale_factor = 1024.0)
: _scale_factor(scalar_type(scale_factor)), _nfft(nfft), _inverse(inverse)
{
// fill twiddle factors
_twiddles.resize(_nfft);
const double phinc = (_inverse ? 2 : -2) * acos(-1.0) / _nfft;
for (std::size_t i = 0; i < _nfft; ++i)
{
_twiddles[i] = scale_factor * exp(complex<double>(0, i * phinc));
}
//factorize
//start factoring out 4's, then 2's, then 3,5,7,9,...
std::size_t n = _nfft;
std::size_t p = 4;
do
{
while (n % p)
{
switch (p)
{
case 4:
p = 2;
break;
case 2:
p = 3;
break;
default:
p += 2;
break;
}
if (p * p > n) p = n;// no more factors
}
n /= p;
_stageRadix.push_back(p);
_stageRemainder.push_back(n);
} while (n > 1);
}
/// Calculates the complex Discrete Fourier Transform.
///
/// The size of the passed arrays must be passed in the constructor.
/// The sum of the squares of the absolute values in the @c dst
/// array will be @c N times the sum of the squares of the absolute
/// values in the @c src array, where @c N is the size of the array.
/// In other words, the l_2 norm of the resulting array will be
/// @c sqrt(N) times as big as the l_2 norm of the input array.
/// This is also the case when the inverse flag is set in the
/// constructor. Hence when applying the same transform twice, but with
/// the inverse flag changed the second time, then the result will
/// be equal to the original input times @c N.
void transform(const cpx_type * FSrc,
cpx_type * FDst,
const std::size_t stage = 0,
const std::size_t fstride = 1,
const std::size_t in_stride = 1) const
{
const std::size_t p = _stageRadix[stage];
const std::size_t m = _stageRemainder[stage];
cpx_type *const Fout_beg = FDst;
cpx_type *const Fout_end = FDst + p * m;
if (m == 1)
{
do
{
*FDst = *FSrc;
FSrc += fstride * in_stride;
} while (++FDst != Fout_end);
}
else
{
do
{
// recursive call:
// DFT of size m*p performed by doing
// p instances of smaller DFTs of size m,
// each one takes a decimated version of the input
transform(FSrc, FDst, stage + 1, fstride * p, in_stride);
FSrc += fstride * in_stride;
} while ((FDst += m) != Fout_end);
}
FDst = Fout_beg;
// recombine the p smaller DFTs
switch (p)
{
case 2:
kf_bfly2(FDst, fstride, m);
break;
case 3:
kf_bfly3(FDst, fstride, m);
break;
case 4:
kf_bfly4(FDst, fstride, m);
break;
case 5:
kf_bfly5(FDst, fstride, m);
break;
default:
kf_bfly_generic(FDst, fstride, m, p);
break;
}
}
private:
void kf_bfly2(cpx_type *const Fout, const size_t fstride, const std::size_t m) const
{
for (std::size_t k = 0; k < m; ++k)
{
const cpx_type t = (Fout[m + k] * _twiddles[k * fstride]) / _scale_factor;
Fout[m + k] = Fout[k] - t;
Fout[k] += t;
}
}
void kf_bfly3(cpx_type *Fout, const std::size_t fstride, const std::size_t m) const
{
std::size_t k = m;
const std::size_t m2 = 2 * m;
const cpx_type *tw1, *tw2;
cpx_type scratch[5];
const cpx_type epi3 = _twiddles[fstride * m];
tw1 = tw2 = &_twiddles[0];
do
{
scratch[1] = (Fout[m] * *tw1) / _scale_factor;
scratch[2] = (Fout[m2] * *tw2) / _scale_factor;
scratch[3] = scratch[1] + scratch[2];
scratch[0] = scratch[1] - scratch[2];
tw1 += fstride;
tw2 += fstride * 2;
Fout[m] = Fout[0] - (scratch[3] / 2);
scratch[0] *= epi3.imag();
scratch[0] /= _scale_factor;
Fout[0] += scratch[3];
Fout[m2] = cpx_type(Fout[m].real() + scratch[0].imag(), Fout[m].imag() - scratch[0].real());
Fout[m] += cpx_type(-scratch[0].imag(), scratch[0].real());
++Fout;
} while (--k);
}
void kf_bfly4(cpx_type *const Fout, const std::size_t fstride, const std::size_t m) const
{
cpx_type scratch[7];
const scalar_type negative_if_inverse = _inverse ? -1 : +1;
for (std::size_t k = 0; k < m; ++k)
{
scratch[0] = (Fout[k + m] * _twiddles[k * fstride]) / _scale_factor;
scratch[1] = (Fout[k + 2 * m] * _twiddles[k * fstride * 2]) / _scale_factor;
scratch[2] = (Fout[k + 3 * m] * _twiddles[k * fstride * 3]) / _scale_factor;
scratch[5] = Fout[k] - scratch[1];
Fout[k] += scratch[1];
scratch[3] = scratch[0] + scratch[2];
scratch[4] = scratch[0] - scratch[2];
scratch[4] = cpx_type(scratch[4].imag() * negative_if_inverse,
-scratch[4].real() * negative_if_inverse);
Fout[k + 2 * m] = Fout[k] - scratch[3];
Fout[k] += scratch[3];
Fout[k + m] = scratch[5] + scratch[4];
Fout[k + 3 * m] = scratch[5] - scratch[4];
}
}
void kf_bfly5(cpx_type *const Fout, const std::size_t fstride, const std::size_t m) const
{
cpx_type *Fout0, *Fout1, *Fout2, *Fout3, *Fout4;
cpx_type scratch[13];
const cpx_type ya = _twiddles[fstride * m];
const cpx_type yb = _twiddles[fstride * 2 * m];
Fout0 = Fout;
Fout1 = Fout0 + m;
Fout2 = Fout0 + 2 * m;
Fout3 = Fout0 + 3 * m;
Fout4 = Fout0 + 4 * m;
for (std::size_t u = 0; u < m; ++u)
{
scratch[0] = *Fout0;
scratch[1] = (*Fout1 * _twiddles[u * fstride]) / _scale_factor;
scratch[2] = (*Fout2 * _twiddles[2 * u * fstride]) / _scale_factor;
scratch[3] = (*Fout3 * _twiddles[3 * u * fstride]) / _scale_factor;
scratch[4] = (*Fout4 * _twiddles[4 * u * fstride]) / _scale_factor;
scratch[7] = scratch[1] + scratch[4];
scratch[10] = scratch[1] - scratch[4];
scratch[8] = scratch[2] + scratch[3];
scratch[9] = scratch[2] - scratch[3];
*Fout0 += scratch[7];
*Fout0 += scratch[8];
scratch[5] = scratch[0] + (cpx_type(
scratch[7].real() * ya.real() + scratch[8].real() * yb.real(),
scratch[7].imag() * ya.real() + scratch[8].imag() * yb.real() ) / _scale_factor);
scratch[6] = cpx_type(
scratch[10].imag() * ya.imag() + scratch[9].imag() * yb.imag(),
-scratch[10].real() * ya.imag() - scratch[9].real() * yb.imag() ) / _scale_factor;
*Fout1 = scratch[5] - scratch[6];
*Fout4 = scratch[5] + scratch[6];
scratch[11] = scratch[0] + (cpx_type(
scratch[7].real() * yb.real() + scratch[8].real() * ya.real(),
scratch[7].imag() * yb.real() + scratch[8].imag() * ya.real() ) / _scale_factor);
scratch[12] = cpx_type(
-scratch[10].imag() * yb.imag() + scratch[9].imag() * ya.imag(),
scratch[10].real() * yb.imag() - scratch[9].real() * ya.imag() ) / _scale_factor;
*Fout2 = scratch[11] + scratch[12];
*Fout3 = scratch[11] - scratch[12];
++Fout0;
++Fout1;
++Fout2;
++Fout3;
++Fout4;
}
}
/* perform the butterfly for one stage of a mixed radix FFT */
void kf_bfly_generic(cpx_type * const Fout, const size_t fstride, const std::size_t m, const std::size_t p) const
{
const cpx_type *twiddles = &_twiddles[0];
cpx_type scratchbuf[p];
for (std::size_t u = 0; u < m; ++u)
{
std::size_t k = u;
for (std::size_t q1 = 0; q1 < p; ++q1)
{
scratchbuf[q1] = Fout[k];
k += m;
}
k = u;
for (std::size_t q1 = 0; q1 < p; ++q1)
{
std::size_t twidx = 0;
Fout[k] = scratchbuf[0];
for (std::size_t q = 1; q < p; ++q)
{
twidx += fstride * k;
if (twidx >= _nfft)
twidx -= _nfft;
Fout[k] += (scratchbuf[q] * twiddles[twidx]) / _scale_factor;
}
k += m;
}
}
}
};
#endif

View File

@ -0,0 +1,63 @@
function(add_kissfft_test_executable NAME)
add_kissfft_executable(${NAME} ${ARGN})
target_include_directories(${NAME} PRIVATE ..)
add_test(NAME ${NAME} COMMAND ${NAME})
set_tests_properties(${NAME} PROPERTIES TIMEOUT 3600)
endfunction()
set(KISSFFT_TEST_NUMFFTS 10000)
#
# Add tools-independent fastfilt_* (../tools/fft_*) executable without adding a test
#
add_kissfft_executable(fastfilt ../tools/fftutil.c)
target_include_directories(fastfilt PRIVATE ..)
#
# Add test executables and define tests
#
add_kissfft_test_executable(bm_kiss benchkiss.c pstats.c)
# add_test(NAME benchmar COMMAND ${NAME})
# set_tests_properties(${NAME} PROPERTIES TIMEOUT 3600)
include(FindPkgConfig)
if(KISSFFT_FLOAT)
set(fftw3_pkg fftw3f)
else()
set(fftw3_pkg fftw3)
endif()
pkg_check_modules(fftw3 REQUIRED IMPORTED_TARGET ${fftw3_pkg})
add_kissfft_test_executable(bm_fftw benchfftw.c pstats.c)
target_link_libraries(bm_fftw PRIVATE PkgConfig::fftw3)
add_kissfft_test_executable(st twotonetest.c)
add_kissfft_test_executable(tkfc twotonetest.c)
target_compile_definitions(tkfc PRIVATE KFC_TEST)
add_kissfft_test_executable(ffr twotonetest.c)
add_kissfft_test_executable(tr test_real.c)
add_kissfft_test_executable(testcpp testcpp.cc)
if(KISSFFT_DATATYPE MATCHES "^simd$")
add_kissfft_test_executable(tsimd test_simd.c)
target_compile_definitions(tsimd PRIVATE USE_SIMD)
if (NOT MSVC)
target_compile_options(kissfft PRIVATE -msse)
else()
target_compile_options(kissfft PRIVATE "/arch:SSE")
endif()
endif()
find_package(PythonInterp REQUIRED)
add_test(NAME testkiss.py COMMAND "${PYTHON_EXECUTABLE}" "testkiss.py")
list(APPEND TESTKISS_PY_ENV "KISSFFT_DATATYPE=${KISSFFT_DATATYPE}")
list(APPEND TESTKISS_PY_ENV "KISSFFT_OPENMP=${KISSFFT_OPENMP}")
set_tests_properties(testkiss.py PROPERTIES
TIMEOUT 3600
ENVIRONMENT "${TESTKISS_PY_ENV}"
WORKING_DIRECTORY "${CMAKE_CURRENT_BINARY_DIR}")

View File

@ -0,0 +1,101 @@
/*
* Copyright (c) 2003-2010, Mark Borgerding. All rights reserved.
* This file is part of KISS FFT - https://github.com/mborgerding/kissfft
*
* SPDX-License-Identifier: BSD-3-Clause
* See COPYING file for more information.
*/
#include <stdio.h>
#include <stdlib.h>
#include <fftw3.h>
#include <unistd.h>
#include "pstats.h"
#ifdef DATATYPEdouble
#define CPXTYPE fftw_complex
#define PLAN fftw_plan
#define FFTMALLOC fftw_malloc
#define MAKEPLAN fftw_plan_dft_1d
#define DOFFT fftw_execute
#define DESTROYPLAN fftw_destroy_plan
#define FFTFREE fftw_free
#elif defined(DATATYPEfloat)
#define CPXTYPE fftwf_complex
#define PLAN fftwf_plan
#define FFTMALLOC fftwf_malloc
#define MAKEPLAN fftwf_plan_dft_1d
#define DOFFT fftwf_execute
#define DESTROYPLAN fftwf_destroy_plan
#define FFTFREE fftwf_free
#endif
#ifndef CPXTYPE
int main(void)
{
fprintf(stderr,"Datatype not available in FFTW\n" );
return 0;
}
#else
int main(int argc,char ** argv)
{
int nfft=1024;
int isinverse=0;
int numffts=1000,i;
CPXTYPE * in=NULL;
CPXTYPE * out=NULL;
PLAN p;
pstats_init();
while (1) {
int c = getopt (argc, argv, "n:ix:h");
if (c == -1)
break;
switch (c) {
case 'n':
nfft = atoi (optarg);
break;
case 'x':
numffts = atoi (optarg);
break;
case 'i':
isinverse = 1;
break;
case 'h':
case '?':
default:
fprintf(stderr,"options:\n-n N: complex fft length\n-i: inverse\n-x N: number of ffts to compute\n"
"");
}
}
in=FFTMALLOC(sizeof(CPXTYPE) * nfft);
out=FFTMALLOC(sizeof(CPXTYPE) * nfft);
for (i=0;i<nfft;++i ) {
in[i][0] = rand() - RAND_MAX/2;
in[i][1] = rand() - RAND_MAX/2;
}
if ( isinverse )
p = MAKEPLAN(nfft, in, out, FFTW_BACKWARD, FFTW_ESTIMATE);
else
p = MAKEPLAN(nfft, in, out, FFTW_FORWARD, FFTW_ESTIMATE);
for (i=0;i<numffts;++i)
DOFFT(p);
DESTROYPLAN(p);
FFTFREE(in); FFTFREE(out);
fprintf(stderr,"fftw\tnfft=%d\tnumffts=%d\n", nfft,numffts);
pstats_report();
return 0;
}
#endif

View File

@ -0,0 +1,129 @@
/*
* Copyright (c) 2003-2010, Mark Borgerding. All rights reserved.
* This file is part of KISS FFT - https://github.com/mborgerding/kissfft
*
* SPDX-License-Identifier: BSD-3-Clause
* See COPYING file for more information.
*/
#include <stdio.h>
#include <stdlib.h>
#include <sys/times.h>
#include <unistd.h>
#include "kiss_fft.h"
#include "kiss_fftr.h"
#include "kiss_fftnd.h"
#include "kiss_fftndr.h"
#include "pstats.h"
static
int getdims(int * dims, char * arg)
{
char *s;
int ndims=0;
while ( (s=strtok( arg , ",") ) ) {
dims[ndims++] = atoi(s);
//printf("%s=%d\n",s,dims[ndims-1]);
arg=NULL;
}
return ndims;
}
int main(int argc,char ** argv)
{
int k;
int nfft[32];
int ndims = 1;
int isinverse=0;
int numffts=1000,i;
kiss_fft_cpx * buf;
kiss_fft_cpx * bufout;
int real = 0;
nfft[0] = 1024;// default
while (1) {
int c = getopt (argc, argv, "n:ix:r");
if (c == -1)
break;
switch (c) {
case 'r':
real = 1;
break;
case 'n':
ndims = getdims(nfft, optarg );
if (nfft[0] != kiss_fft_next_fast_size(nfft[0]) ) {
int ng = kiss_fft_next_fast_size(nfft[0]);
fprintf(stderr,"warning: %d might be a better choice for speed than %d\n",ng,nfft[0]);
}
break;
case 'x':
numffts = atoi (optarg);
break;
case 'i':
isinverse = 1;
break;
}
}
int nbytes = sizeof(kiss_fft_cpx);
for (k=0;k<ndims;++k)
nbytes *= nfft[k];
#ifdef USE_SIMD
numffts /= 4;
fprintf(stderr,"since SIMD implementation does 4 ffts at a time, numffts is being reduced to %d\n",numffts);
#endif
buf=(kiss_fft_cpx*)KISS_FFT_MALLOC(nbytes);
bufout=(kiss_fft_cpx*)KISS_FFT_MALLOC(nbytes);
memset(buf,0,nbytes);
pstats_init();
if (ndims==1) {
if (real) {
kiss_fftr_cfg st = kiss_fftr_alloc( nfft[0] ,isinverse ,0,0);
if (isinverse)
for (i=0;i<numffts;++i)
kiss_fftri( st ,(kiss_fft_cpx*)buf,(kiss_fft_scalar*)bufout );
else
for (i=0;i<numffts;++i)
kiss_fftr( st ,(kiss_fft_scalar*)buf,(kiss_fft_cpx*)bufout );
free(st);
}else{
kiss_fft_cfg st = kiss_fft_alloc( nfft[0] ,isinverse ,0,0);
for (i=0;i<numffts;++i)
kiss_fft( st ,buf,bufout );
free(st);
}
}else{
if (real) {
kiss_fftndr_cfg st = kiss_fftndr_alloc( nfft,ndims ,isinverse ,0,0);
if (isinverse)
for (i=0;i<numffts;++i)
kiss_fftndri( st ,(kiss_fft_cpx*)buf,(kiss_fft_scalar*)bufout );
else
for (i=0;i<numffts;++i)
kiss_fftndr( st ,(kiss_fft_scalar*)buf,(kiss_fft_cpx*)bufout );
free(st);
}else{
kiss_fftnd_cfg st= kiss_fftnd_alloc(nfft,ndims,isinverse ,0,0);
for (i=0;i<numffts;++i)
kiss_fftnd( st ,buf,bufout );
free(st);
}
}
free(buf); free(bufout);
fprintf(stderr,"KISS\tnfft=");
for (k=0;k<ndims;++k)
fprintf(stderr, "%d,",nfft[k]);
fprintf(stderr,"\tnumffts=%d\n" ,numffts);
pstats_report();
kiss_fft_cleanup();
return 0;
}

View File

@ -0,0 +1,134 @@
/*
* This program is in the public domain
* A program that helps the authors of the fine fftw library benchmark kiss
* This file is part of KISS FFT - https://github.com/mborgerding/kissfft
*
* SPDX-License-Identifier: Unlicense
* See LICENSES/Unlicense for more information.
*/
#include "bench-user.h"
#include <math.h>
#include "kiss_fft.h"
#include "kiss_fftnd.h"
#include "kiss_fftr.h"
BEGIN_BENCH_DOC
BENCH_DOC("name", "kissfft")
BENCH_DOC("version", "1.0.1")
BENCH_DOC("year", "2004")
BENCH_DOC("author", "Mark Borgerding")
BENCH_DOC("language", "C")
BENCH_DOC("url", "http://sourceforge.net/projects/kissfft/")
BENCH_DOC("copyright",
"Copyright (c) 2003,4 Mark Borgerding\n"
"\n"
"All rights reserved.\n"
"\n"
"Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:\n"
"\n"
" * Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.\n"
" * Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.\n"
" * Neither the author nor the names of any contributors may be used to endorse or promote products derived from this software without specific prior written permission.\n"
"\n"
"THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.\n")
END_BENCH_DOC
int can_do(struct problem *p)
{
if (p->rank == 1) {
if (p->kind == PROBLEM_REAL) {
return (p->n[0] & 1) == 0; /* only even real is okay */
} else {
return 1;
}
} else {
return p->kind == PROBLEM_COMPLEX;
}
}
static kiss_fft_cfg cfg=NULL;
static kiss_fftr_cfg cfgr=NULL;
static kiss_fftnd_cfg cfgnd=NULL;
#define FAILIF( c ) \
if ( c ) do {\
fprintf(stderr,\
"kissfft: " #c " (file=%s:%d errno=%d %s)\n",\
__FILE__,__LINE__ , errno,strerror( errno ) ) ;\
exit(1);\
}while(0)
void setup(struct problem *p)
{
size_t i;
/*
fprintf(stderr,"%s %s %d-d ",
(p->sign == 1)?"Inverse":"Forward",
p->kind == PROBLEM_COMPLEX?"Complex":"Real",
p->rank);
*/
if (p->rank == 1) {
if (p->kind == PROBLEM_COMPLEX) {
cfg = kiss_fft_alloc (p->n[0], (p->sign == 1), 0, 0);
FAILIF(cfg==NULL);
}else{
cfgr = kiss_fftr_alloc (p->n[0], (p->sign == 1), 0, 0);
FAILIF(cfgr==NULL);
}
}else{
int dims[5];
for (i=0;i<p->rank;++i){
dims[i] = p->n[i];
}
/* multi-dimensional */
if (p->kind == PROBLEM_COMPLEX) {
cfgnd = kiss_fftnd_alloc( dims , p->rank, (p->sign == 1), 0, 0 );
FAILIF(cfgnd==NULL);
}
}
}
void doit(int iter, struct problem *p)
{
int i;
void *in = p->in;
void *out = p->out;
if (p->in_place)
out = p->in;
if (p->rank == 1) {
if (p->kind == PROBLEM_COMPLEX){
for (i = 0; i < iter; ++i)
kiss_fft (cfg, in, out);
} else {
/* PROBLEM_REAL */
if (p->sign == -1) /* FORWARD */
for (i = 0; i < iter; ++i)
kiss_fftr (cfgr, in, out);
else
for (i = 0; i < iter; ++i)
kiss_fftri (cfgr, in, out);
}
}else{
/* multi-dimensional */
for (i = 0; i < iter; ++i)
kiss_fftnd(cfgnd,in,out);
}
}
void done(struct problem *p)
{
free(cfg);
cfg=NULL;
free(cfgr);
cfgr=NULL;
free(cfgnd);
cfgnd=NULL;
UNUSED(p);
}

View File

@ -0,0 +1,275 @@
#!/bin/sh
#
# Test suite for kissfft
#
# Copyright (c) 2021, Vasyl Gello.
# This file is part of KISS FFT - https://github.com/mborgerding/kissfft
#
# SPDX-License-Identifier: BSD-3-Clause
# See COPYING file for more information.
#
if [ ! -f CHANGELOG ] && [ ! -f kiss_fft.h ]; then
echo "ERROR: Please run this testsuite from top level of kissfft source tree!" >&2
return 1
fi
TESTSUITEOUTDIR="$2"
if [ -z "$TESTSUITEOUTDIR" ]; then
TESTSUITEOUTDIR="/tmp/kissfft-testsuite"
fi
if ! mkdir -p "$TESTSUITEOUTDIR"; then
echo "ERROR: Can not create directory '$TESTSUITEOUTDIR'!" >&2
return 1
fi
#
# Test runner function
#
# Parameters:
#
# $1 - Action: "test" or "install"
# $2 - Build type: "make" or "cmake"
# $3 - Data type: "float" "double" "int16_t" "int32_t" "simd"
# $4 - library type: "shared" or "static"
# $5 - Include tools: "yes" or "no"
# $6 - Install root dir: "existing writable directory"
#
test_runner() {
_ACTION="$1"
_BUILD_TYPE="$2"
_DATA_TYPE="$3"
_LIB_TYPE="$4"
_OPENMP="$5"
_INCLUDE_TOOLS="$6"
_INSTALL_ROOT_DIR="$7"
_CMAKE_OPTS=""
_MAKE_OPTS=""
# Prepare install directory name without "$_OPENMP" and "$_INCLUDE_TOOLS"
_INSTALL_DIR="$_INSTALL_ROOT_DIR/$_BUILD_TYPE/$_DATA_TYPE/$_LIB_TYPE"
# Prepare log file without "$_OPENMP" and "$_INCLUDE_TOOLS"
_LOG_FILE="$_INSTALL_ROOT_DIR/$_ACTION-$_BUILD_TYPE-$_DATA_TYPE-$_LIB_TYPE"
# Validate parameters
# Create install root directory
if [ -z "$_INSTALL_ROOT_DIR" ]; then
echo "" >&2
echo "ERROR: Empty path to writeable directory" >&2
echo "" >&2
return 1
fi
if [ ! -d "$_INSTALL_ROOT_DIR" ]; then
if ! mkdir -p "$_INSTALL_ROOT_DIR"; then
echo "" >&2
echo "ERROR: Can not create directory '$_INSTALL_ROOT_DIR'" >&2
echo "" >&2
return 1
fi
fi
if [ "$_BUILD_TYPE" != "make" ] && [ "$_BUILD_TYPE" != "cmake" ]; then
echo "ERROR: Build type must be one of: cmake make" >&2
echo "" >&2
return 1
fi
if [ "$_DATA_TYPE" != "double" ] &&
[ "$_DATA_TYPE" != "float" ] &&
[ "$_DATA_TYPE" != "int16_t" ] &&
[ "$_DATA_TYPE" != "int32_t" ] &&
[ "$_DATA_TYPE" != "simd" ];
then
echo "ERROR: Data type must be one of: double float int16_t int32_t simd" >&2
echo "" >&2
return 1
else
_MAKE_OPTS="$_MAKE_OPTS KISSFFT_DATATYPE=$_DATA_TYPE"
_CMAKE_OPTS="$_CMAKE_OPTS -DKISSFFT_DATATYPE=$_DATA_TYPE"
fi
if [ "$_LIB_TYPE" != "shared" ] && [ "$_LIB_TYPE" != "static" ]; then
echo "ERROR: Library type must be one of: shared static" >&2
echo "" >&2
return 1
fi
case "$_LIB_TYPE" in
"shared")
;;
"static")
_MAKE_OPTS="$_MAKE_OPTS KISSFFT_STATIC=1"
_CMAKE_OPTS="$_CMAKE_OPTS -DKISSFFT_STATIC=ON"
;;
"*")
echo "ERROR: OpenMP inclusion must be one of: no yes" >&2
echo "" >&2
return 1
;;
esac
case "$_OPENMP" in
"yes")
_INSTALL_DIR="$_INSTALL_DIR/openmp"
_LOG_FILE="$_LOG_FILE-openmp"
_MAKE_OPTS="$_MAKE_OPTS KISSFFT_OPENMP=1"
_CMAKE_OPTS="$_CMAKE_OPTS -DKISSFFT_OPENMP=ON"
;;
"no")
_INSTALL_DIR="$_INSTALL_DIR/noopenmp"
_LOG_FILE="$_LOG_FILE-noopenmp"
;;
"*")
echo "ERROR: OpenMP inclusion must be one of: no yes" >&2
echo "" >&2
return 1
;;
esac
case "$_INCLUDE_TOOLS" in
"yes")
_INSTALL_DIR="$_INSTALL_DIR/tools"
_LOG_FILE="$_LOG_FILE-tools"
;;
"no")
_INSTALL_DIR="$_INSTALL_DIR/notools"
_LOG_FILE="$_LOG_FILE-notools"
_MAKE_OPTS="$_MAKE_OPTS KISSFFT_TOOLS=0"
_CMAKE_OPTS="$_CMAKE_OPTS -DKISSFFT_TOOLS=OFF"
;;
"*")
echo "ERROR: Tools inclusion must be one of: no yes" >&2
echo "" >&2
return 1
;;
esac
# Clean kissfft
rm -rf build 1>/dev/null 2>/dev/null
make clean 1>/dev/null 2>&1
# Prepare status line
_STATUS_LINE="Running: $(printf "% 10s" "$_ACTION") |"
_STATUS_LINE="$_STATUS_LINE Build Type: $(printf "% 7s" "$_BUILD_TYPE") |"
_STATUS_LINE="$_STATUS_LINE Data Type: $(printf "% 7s" "$_DATA_TYPE") |"
_STATUS_LINE="$_STATUS_LINE Lib Type: $(printf "% 7s" "$_LIB_TYPE") |"
_STATUS_LINE="$_STATUS_LINE OpenMP: $(printf "% 3s" "$_OPENMP") |"
_STATUS_LINE="$_STATUS_LINE Tools: $(printf "% 3s" "$_INCLUDE_TOOLS") |"
# Skip tests with tools not installed as they are same as with tools
if [ "$_ACTION" = "test" ] && [ "$_INCLUDE_TOOLS" = "no" ]; then
return 2
fi
# Run selected action
echo "$_STATUS_LINE"
case "$_ACTION" in
"test")
_MAKE_OPTS="$_MAKE_OPTS PREFIX=$_INSTALL_DIR"
_CMAKE_OPTS="$_CMAKE_OPTS -DCMAKE_INSTALL_PREFIX=$_INSTALL_DIR"
case "$_BUILD_TYPE" in
"make")
make $_MAKE_OPTS all 1>>"$_LOG_FILE" 2>&1 &&
make $_MAKE_OPTS testsingle 1>>"$_LOG_FILE" 2>&1 &&
_RET=$?
;;
"cmake")
mkdir build 1>/dev/null 2>&1 &&
cd build &&
cmake $_CMAKE_OPTS .. 1>"$_LOG_FILE" 2>&1 &&
make all 1>>"$_LOG_FILE" 2>&1 &&
make test 1>>"$_LOG_FILE" 2>&1
_RET=$?
cd ..
;;
esac
;;
"install")
_MAKE_OPTS="$_MAKE_OPTS PREFIX=$_INSTALL_DIR"
_CMAKE_OPTS="$_CMAKE_OPTS -DCMAKE_INSTALL_PREFIX=$_INSTALL_DIR"
case "$_BUILD_TYPE" in
"make")
make $_MAKE_OPTS install 1>>"$_LOG_FILE" 2>&1
_RET=$?
;;
"cmake")
mkdir build 1>/dev/null 2>&1 &&
cd build &&
cmake $_CMAKE_OPTS .. 1>"$_LOG_FILE" 2>&1 &&
make all 1>>"$_LOG_FILE" 2>&1 &&
make install 1>>"$_LOG_FILE" 2>&1
_RET=$?
cd ..
;;
esac
;;
*)
echo "ERROR: Action must be one of: test install" >&2
echo "" >&2
return 1
;;
esac
# Clean kissfft
rm -rf build 1>/dev/null 2>/dev/null
make clean 1>/dev/null 2>&1
# Return result
[ $_RET -eq 0 ] && return 0 || return 1
}
# Main script
for ACTION in test install; do
for BUILD_TYPE in make cmake; do
for DATA_TYPE in float double int16_t int32_t simd; do
for LIB_TYPE in shared static; do
for OPENMP in no yes; do
for INCLUDE_TOOLS in no yes; do
test_runner \
"$ACTION" \
"$BUILD_TYPE" \
"$DATA_TYPE" \
"$LIB_TYPE" \
"$OPENMP" \
"$INCLUDE_TOOLS" \
"$TESTSUITEOUTDIR"
case $? in
0)
echo "Result: OK"
;;
1)
echo "Result: FAIL"
;;
2)
# Ignore it
echo "Result: IGNORE" 1>/dev/null
;;
esac
done
done
done
done
done
done 2>&1 | tee "$TESTSUITEOUTDIR/all-tests.log"

View File

@ -0,0 +1,56 @@
/*
* Copyright (c) 2003-2010, Mark Borgerding. All rights reserved.
* This file is part of KISS FFT - https://github.com/mborgerding/kissfft
*
* SPDX-License-Identifier: BSD-3-Clause
* See COPYING file for more information.
*/
#include <stdio.h>
#include <stdlib.h>
#include <sys/times.h>
#include <sys/types.h>
#include <unistd.h>
#include "pstats.h"
static struct tms tms_beg;
static struct tms tms_end;
static int has_times = 0;
void pstats_init(void)
{
has_times = times(&tms_beg) != -1;
}
static void tms_report(void)
{
double cputime;
if (! has_times )
return;
times(&tms_end);
cputime = ( ((float)tms_end.tms_utime + tms_end.tms_stime + tms_end.tms_cutime + tms_end.tms_cstime ) -
((float)tms_beg.tms_utime + tms_beg.tms_stime + tms_beg.tms_cutime + tms_beg.tms_cstime ) )
/ sysconf(_SC_CLK_TCK);
fprintf(stderr,"\tcputime=%.3f\n" , cputime);
}
static void ps_report(void)
{
char buf[1024];
#ifdef __APPLE__ /* MAC OS X */
sprintf(buf,"ps -o command,majflt,minflt,rss,pagein,vsz -p %d 1>&2",getpid() );
#else /* GNU/Linux */
sprintf(buf,"ps -o comm,majflt,minflt,rss,drs,pagein,sz,trs,vsz %d 1>&2",getpid() );
#endif
if (system( buf )==-1) {
perror("system call to ps failed");
}
}
void pstats_report()
{
ps_report();
tms_report();
}

View File

@ -0,0 +1,14 @@
/*
* Copyright (c) 2003-2010, Mark Borgerding. All rights reserved.
* This file is part of KISS FFT - https://github.com/mborgerding/kissfft
*
* SPDX-License-Identifier: BSD-3-Clause
* See COPYING file for more information.
*/
#ifndef PSTATS_H
#define PSTATS_H
void pstats_init(void);
void pstats_report(void);
#endif

View File

@ -0,0 +1,179 @@
/*
* Copyright (c) 2003-2010, Mark Borgerding. All rights reserved.
* This file is part of KISS FFT - https://github.com/mborgerding/kissfft
*
* SPDX-License-Identifier: BSD-3-Clause
* See COPYING file for more information.
*/
#include "kiss_fftr.h"
#include "_kiss_fft_guts.h"
#include <sys/times.h>
#include <time.h>
#include <unistd.h>
static double cputime(void)
{
struct tms t;
times(&t);
return (double)(t.tms_utime + t.tms_stime)/ sysconf(_SC_CLK_TCK) ;
}
static
kiss_fft_scalar rand_scalar(void)
{
#ifdef USE_SIMD
return _mm_set1_ps(rand()-RAND_MAX/2);
#else
kiss_fft_scalar s = (kiss_fft_scalar)(rand() -RAND_MAX/2);
return s/2;
#endif
}
static
double snr_compare( kiss_fft_cpx * vec1,kiss_fft_cpx * vec2, int n)
{
int k;
double sigpow=1e-10,noisepow=1e-10,err,snr,scale=0;
#ifdef USE_SIMD
float *fv1 = (float*)vec1;
float *fv2 = (float*)vec2;
for (k=0;k<8*n;++k) {
sigpow += *fv1 * *fv1;
err = *fv1 - *fv2;
noisepow += err*err;
++fv1;
++fv2;
}
#else
for (k=0;k<n;++k) {
sigpow += (double)vec1[k].r * (double)vec1[k].r +
(double)vec1[k].i * (double)vec1[k].i;
err = (double)vec1[k].r - (double)vec2[k].r;
noisepow += err * err;
err = (double)vec1[k].i - (double)vec2[k].i;
noisepow += err * err;
if (vec1[k].r)
scale +=(double) vec2[k].r / (double)vec1[k].r;
}
#endif
snr = 10*log10( sigpow / noisepow );
scale /= n;
if (snr<10) {
printf( "\npoor snr, try a scaling factor %f\n" , scale );
exit(1);
}
return snr;
}
#ifndef NUMFFTS
#define NUMFFTS 10000
#endif
int main(int argc,char ** argv)
{
int nfft = 8*3*5;
double ts,tfft,trfft;
int i;
if (argc>1)
nfft = atoi(argv[1]);
kiss_fft_cpx cin[nfft];
kiss_fft_cpx cout[nfft];
kiss_fft_cpx sout[nfft];
kiss_fft_cfg kiss_fft_state;
kiss_fftr_cfg kiss_fftr_state;
kiss_fft_scalar rin[nfft+2];
kiss_fft_scalar rout[nfft+2];
kiss_fft_scalar zero;
memset(&zero,0,sizeof(zero) ); // ugly way of setting short,int,float,double, or __m128 to zero
srand(time(0));
for (i=0;i<nfft;++i) {
rin[i] = rand_scalar();
cin[i].r = rin[i];
cin[i].i = zero;
}
kiss_fft_state = kiss_fft_alloc(nfft,0,0,0);
kiss_fftr_state = kiss_fftr_alloc(nfft,0,0,0);
kiss_fft(kiss_fft_state,cin,cout);
kiss_fftr(kiss_fftr_state,rin,sout);
/*
printf(" results from kiss_fft : (%f,%f), (%f,%f), (%f,%f) ...\n "
, (float)cout[0].r , (float)cout[0].i
, (float)cout[1].r , (float)cout[1].i
, (float)cout[2].r , (float)cout[2].i);
printf(" results from kiss_fftr: (%f,%f), (%f,%f), (%f,%f) ...\n "
, (float)sout[0].r , (float)sout[0].i
, (float)sout[1].r , (float)sout[1].i
, (float)sout[2].r , (float)sout[2].i);
*/
printf( "nfft=%d, inverse=%d, snr=%g\n",
nfft,0, snr_compare(cout,sout,(nfft/2)+1) );
ts = cputime();
for (i=0;i<NUMFFTS;++i) {
kiss_fft(kiss_fft_state,cin,cout);
}
tfft = cputime() - ts;
ts = cputime();
for (i=0;i<NUMFFTS;++i) {
kiss_fftr( kiss_fftr_state, rin, cout );
/* kiss_fftri(kiss_fftr_state,cout,rin); */
}
trfft = cputime() - ts;
printf("%d complex ffts took %gs, real took %gs\n",NUMFFTS,tfft,trfft);
free(kiss_fft_state);
free(kiss_fftr_state);
kiss_fft_state = kiss_fft_alloc(nfft,1,0,0);
kiss_fftr_state = kiss_fftr_alloc(nfft,1,0,0);
memset(cin,0,sizeof(cin));
#if 1
for (i=1;i< nfft/2;++i) {
//cin[i].r = (kiss_fft_scalar)(rand()-RAND_MAX/2);
cin[i].r = rand_scalar();
cin[i].i = rand_scalar();
}
#else
cin[0].r = 12000;
cin[3].r = 12000;
cin[nfft/2].r = 12000;
#endif
// conjugate symmetry of real signal
for (i=1;i< nfft/2;++i) {
cin[nfft-i].r = cin[i].r;
cin[nfft-i].i = - cin[i].i;
}
kiss_fft(kiss_fft_state,cin,cout);
kiss_fftri(kiss_fftr_state,cin,rout);
/*
printf(" results from inverse kiss_fft : (%f,%f), (%f,%f), (%f,%f), (%f,%f), (%f,%f) ...\n "
, (float)cout[0].r , (float)cout[0].i , (float)cout[1].r , (float)cout[1].i , (float)cout[2].r , (float)cout[2].i , (float)cout[3].r , (float)cout[3].i , (float)cout[4].r , (float)cout[4].i
);
printf(" results from inverse kiss_fftr: %f,%f,%f,%f,%f ... \n"
,(float)rout[0] ,(float)rout[1] ,(float)rout[2] ,(float)rout[3] ,(float)rout[4]);
*/
for (i=0;i<nfft;++i) {
sout[i].r = rout[i];
sout[i].i = zero;
}
printf( "nfft=%d, inverse=%d, snr=%g\n",
nfft,1, snr_compare(cout,sout,nfft/2) );
free(kiss_fft_state);
free(kiss_fftr_state);
return 0;
}

View File

@ -0,0 +1,25 @@
#include <kiss_fftnd.h>
static void test1(void)
{
int is_inverse = 1;
int n[2] = {256,256};
size_t nbytes = sizeof(kiss_fft_cpx)*n[0]*n[1];
kiss_fft_cpx * inbuf = _mm_malloc(nbytes,16);
kiss_fft_cpx * outbuf = _mm_malloc(nbytes,16);
memset(inbuf,0,nbytes);
memset(outbuf,0,nbytes);
kiss_fftnd_cfg cfg = kiss_fftnd_alloc(n,2,is_inverse,0,0);
kiss_fftnd(cfg,inbuf,outbuf);
kiss_fft_free(cfg);
_mm_free(inbuf);
_mm_free(outbuf);
}
int main(void)
{
test1();
return 0;
}

View File

@ -0,0 +1,80 @@
/*
* Copyright (c) 2003-2010, Mark Borgerding. All rights reserved.
* This file is part of KISS FFT - https://github.com/mborgerding/kissfft
*
* SPDX-License-Identifier: BSD-3-Clause
* See COPYING file for more information.
*/
#include "kissfft.hh"
#include <iostream>
#include <cstdlib>
#include <typeinfo>
#include <sys/time.h>
static inline
double curtime(void)
{
struct timeval tv;
gettimeofday(&tv, NULL);
return (double)tv.tv_sec + (double)tv.tv_usec*.000001;
}
using namespace std;
template <class T>
void dotest(int nfft)
{
typedef kissfft<T> FFT;
typedef std::complex<T> cpx_type;
cout << "type:" << typeid(T).name() << " nfft:" << nfft;
FFT fft(nfft,false);
vector<cpx_type> inbuf(nfft);
vector<cpx_type> outbuf(nfft);
for (int k=0;k<nfft;++k)
inbuf[k]= cpx_type(
(T)(rand()/(double)RAND_MAX - .5),
(T)(rand()/(double)RAND_MAX - .5) );
fft.transform( &inbuf[0] , &outbuf[0] );
long double totalpower=0;
long double difpower=0;
for (int k0=0;k0<nfft;++k0) {
complex<long double> acc = 0;
long double phinc = 2*k0* M_PIl / nfft;
for (int k1=0;k1<nfft;++k1) {
complex<long double> x(inbuf[k1].real(),inbuf[k1].imag());
acc += x * exp( complex<long double>(0,-k1*phinc) );
}
totalpower += norm(acc);
complex<long double> x(outbuf[k0].real(),outbuf[k0].imag());
complex<long double> dif = acc - x;
difpower += norm(dif);
}
cout << " RMSE:" << sqrt(difpower/totalpower) << "\t";
double t0 = curtime();
int nits=20e6/nfft;
for (int k=0;k<nits;++k) {
fft.transform( &inbuf[0] , &outbuf[0] );
}
double t1 = curtime();
cout << " MSPS:" << ( (nits*nfft)*1e-6/ (t1-t0) ) << endl;
}
int main(int argc,char ** argv)
{
if (argc>1) {
for (int k=1;k<argc;++k) {
int nfft = atoi(argv[k]);
dotest<float>(nfft); dotest<double>(nfft); dotest<long double>(nfft);
}
}else{
dotest<float>(32); dotest<double>(32); dotest<long double>(32);
dotest<float>(1024); dotest<double>(1024); dotest<long double>(1024);
dotest<float>(840); dotest<double>(840); dotest<long double>(840);
}
return 0;
}

View File

@ -0,0 +1,143 @@
#!/usr/bin/env python
# Copyright (c) 2003-2019, Mark Borgerding. All rights reserved.
# This file is part of KISS FFT - https://github.com/mborgerding/kissfft
#
# SPDX-License-Identifier: BSD-3-Clause
# See COPYING file for more information.
from __future__ import absolute_import, division, print_function
import math
import sys
import os
import random
import struct
import getopt
import numpy as np
po = math.pi
e = math.e
do_real = False
datatype = os.environ.get('KISSFFT_DATATYPE', 'float')
openmp = os.environ.get('KISSFFT_OPENMP', 'float')
util = './fastfilt-' + datatype
if openmp == '1' or openmp == 'ON':
util = util + '-openmp'
minsnr = 90
if datatype == 'double':
dtype = np.float64
elif datatype == 'float':
dtype = np.float32
elif datatype == 'int16_t':
dtype = np.int16
minsnr = 10
elif datatype == 'int32_t':
dtype = np.int32
elif datatype == 'simd':
sys.stderr.write('testkiss.py does not yet test simd')
sys.exit(0)
else:
sys.stderr.write('unrecognized datatype {0}\n'.format(datatype))
sys.exit(1)
def dopack(x):
if np.iscomplexobj(x):
x = x.astype(np.complex128).view(np.float64)
else:
x = x.astype(np.float64)
return x.astype(dtype).tobytes()
def dounpack(x, cpx):
x = np.frombuffer(x, dtype).astype(np.float64)
if cpx:
x = x[::2] + 1j * x[1::2]
return x
def make_random(shape):
'create random uniform (-1,1) data of the given shape'
if do_real:
return np.random.uniform(-1, 1, shape)
else:
return (np.random.uniform(-1, 1, shape) + 1j * np.random.uniform(-1, 1, shape))
def randmat(ndim):
'create a random multidimensional array in range (-1,1)'
dims = np.random.randint(2, 5, ndim)
if do_real:
dims[-1] = (dims[-1] // 2) * 2 # force even last dimension if real
return make_random(dims)
def test_fft(ndim):
x = randmat(ndim)
if do_real:
xver = np.fft.rfftn(x)
else:
xver = np.fft.fftn(x)
x2 = dofft(x, do_real)
err = xver - x2
errf = err.ravel()
xverf = xver.ravel()
errpow = np.vdot(errf, errf) + 1e-10
sigpow = np.vdot(xverf, xverf) + 1e-10
snr = 10 * math.log10(abs(sigpow / errpow))
print('SNR (compared to NumPy) : {0:.1f}dB'.format(float(snr)))
if snr < minsnr:
print('xver=', xver)
print('x2=', x2)
print('err', err)
sys.exit(1)
def dofft(x, isreal):
dims = list(np.shape(x))
x = x.ravel()
scale = 1
if datatype == 'int16_t':
x = 32767 * x
scale = len(x) / 32767.0
elif datatype == 'int32_t':
x = 2147483647.0 * x
scale = len(x) / 2147483647.0
cmd = util + ' -n '
cmd += ','.join([str(d) for d in dims])
if do_real:
cmd += ' -R '
print(cmd)
from subprocess import Popen, PIPE
p = Popen(cmd, shell=True, stdin=PIPE, stdout=PIPE)
p.stdin.write(dopack(x))
p.stdin.close()
res = dounpack(p.stdout.read(), 1)
if do_real:
dims[-1] = (dims[-1] // 2) + 1
res = scale * res
p.wait()
return np.reshape(res, dims)
def main():
opts, args = getopt.getopt(sys.argv[1:], 'r')
opts = dict(opts)
global do_real
do_real = '-r' in opts
if do_real:
print('Testing multi-dimensional real FFTs')
else:
print('Testing multi-dimensional FFTs')
for dim in range(1, 4):
test_fft(dim)
if __name__ == "__main__":
main()

View File

@ -0,0 +1,101 @@
/*
* Copyright (c) 2003-2010, Mark Borgerding. All rights reserved.
* This file is part of KISS FFT - https://github.com/mborgerding/kissfft
*
* SPDX-License-Identifier: BSD-3-Clause
* See COPYING file for more information.
*/
#include <stdlib.h>
#include <string.h>
#include <stdio.h>
#include "kiss_fft.h"
#include "kiss_fftr.h"
#include <limits.h>
static
double two_tone_test( int nfft, int bin1,int bin2)
{
kiss_fftr_cfg cfg = NULL;
kiss_fft_cpx *kout = NULL;
kiss_fft_scalar *tbuf = NULL;
int i;
double f1 = bin1*2*M_PI/nfft;
double f2 = bin2*2*M_PI/nfft;
double sigpow=0;
double noisepow=0;
#if FIXED_POINT==32
long maxrange = LONG_MAX;
#else
long maxrange = SHRT_MAX;/* works fine for float too*/
#endif
cfg = kiss_fftr_alloc(nfft , 0, NULL, NULL);
tbuf = KISS_FFT_MALLOC(nfft * sizeof(kiss_fft_scalar));
kout = KISS_FFT_MALLOC(nfft * sizeof(kiss_fft_cpx));
/* generate a signal with two tones*/
for (i = 0; i < nfft; i++) {
#ifdef USE_SIMD
tbuf[i] = _mm_set1_ps( (maxrange>>1)*cos(f1*i)
+ (maxrange>>1)*cos(f2*i) );
#else
tbuf[i] = (maxrange>>1)*cos(f1*i)
+ (maxrange>>1)*cos(f2*i);
#endif
}
kiss_fftr(cfg, tbuf, kout);
for (i=0;i < (nfft/2+1);++i) {
#ifdef USE_SIMD
double tmpr = (double)*(float*)&kout[i].r / (double)maxrange;
double tmpi = (double)*(float*)&kout[i].i / (double)maxrange;
#else
double tmpr = (double)kout[i].r / (double)maxrange;
double tmpi = (double)kout[i].i / (double)maxrange;
#endif
double mag2 = tmpr*tmpr + tmpi*tmpi;
if (i!=0 && i!= nfft/2)
mag2 *= 2; /* all bins except DC and Nyquist have symmetric counterparts implied*/
/* if there is power in one of the expected bins, it is signal, otherwise noise*/
if ( i!=bin1 && i != bin2 )
noisepow += mag2;
else
sigpow += mag2;
}
kiss_fft_cleanup();
/*printf("TEST %d,%d,%d noise @ %fdB\n",nfft,bin1,bin2,10*log10(noisepow/sigpow +1e-30) );*/
return 10*log10(sigpow/(noisepow+1e-50) );
}
int main(int argc,char ** argv)
{
int nfft = 4*2*2*3*5;
if (argc>1) nfft = atoi(argv[1]);
int i,j;
double minsnr = 500;
double maxsnr = -500;
double snr;
for (i=0;i<nfft/2;i+= (nfft>>4)+1) {
for (j=i;j<nfft/2;j+=(nfft>>4)+7) {
snr = two_tone_test(nfft,i,j);
if (snr<minsnr) {
minsnr=snr;
}
if (snr>maxsnr) {
maxsnr=snr;
}
}
}
snr = two_tone_test(nfft,nfft/2,nfft/2);
if (snr<minsnr) minsnr=snr;
if (snr>maxsnr) maxsnr=snr;
printf("TwoToneTest: snr ranges from %ddB to %ddB\n",(int)minsnr,(int)maxsnr);
printf("sizeof(kiss_fft_scalar) = %d\n",(int)sizeof(kiss_fft_scalar) );
return 0;
}

View File

@ -0,0 +1,29 @@
add_kissfft_executable(fastconvr kiss_fastfir.c)
target_compile_definitions(fastconvr PRIVATE REAL_FASTFIR FAST_FILT_UTIL)
add_kissfft_executable(fastconv kiss_fastfir.c)
target_compile_definitions(fastconv PRIVATE FAST_FILT_UTIL)
add_kissfft_executable(fft fftutil.c)
install(TARGETS fastconv fastconvr fft
ARCHIVE DESTINATION ${CMAKE_INSTALL_LIBDIR}
LIBRARY DESTINATION ${CMAKE_INSTALL_LIBDIR}
RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR}
PUBLIC_HEADER DESTINATION ${PKGINCLUDEDIR})
# psdpng does not build with "simd" datatype
if(NOT KISSFFT_DATATYPE MATCHES "simd")
include(FindPkgConfig)
pkg_check_modules(libpng REQUIRED IMPORTED_TARGET libpng)
add_kissfft_executable(psdpng psdpng.c)
target_link_libraries(psdpng PRIVATE PkgConfig::libpng)
install(TARGETS psdpng
ARCHIVE DESTINATION ${CMAKE_INSTALL_LIBDIR}
LIBRARY DESTINATION ${CMAKE_INSTALL_LIBDIR}
RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR}
PUBLIC_HEADER DESTINATION ${PKGINCLUDEDIR})
endif()
#FIXME: dumphdr.c is not available
#add_kissfft_executable(dumphdr dumphdr.c)

View File

@ -0,0 +1,202 @@
/*
* Copyright (c) 2003-2004, Mark Borgerding. All rights reserved.
* This file is part of KISS FFT - https://github.com/mborgerding/kissfft
*
* SPDX-License-Identifier: BSD-3-Clause
* See COPYING file for more information.
*/
#include <stdlib.h>
#include <math.h>
#include <stdio.h>
#include <string.h>
#include <unistd.h>
#include "kiss_fft.h"
#include "kiss_fftndr.h"
static
void fft_file(FILE * fin,FILE * fout,int nfft,int isinverse)
{
kiss_fft_cfg st;
kiss_fft_cpx * buf;
kiss_fft_cpx * bufout;
buf = (kiss_fft_cpx*)malloc(sizeof(kiss_fft_cpx) * nfft );
bufout = (kiss_fft_cpx*)malloc(sizeof(kiss_fft_cpx) * nfft );
st = kiss_fft_alloc( nfft ,isinverse ,0,0);
while ( fread( buf , sizeof(kiss_fft_cpx) * nfft ,1, fin ) > 0 ) {
kiss_fft( st , buf ,bufout);
fwrite( bufout , sizeof(kiss_fft_cpx) , nfft , fout );
}
free(st);
free(buf);
free(bufout);
}
static
void fft_filend(FILE * fin,FILE * fout,int *dims,int ndims,int isinverse)
{
kiss_fftnd_cfg st;
kiss_fft_cpx *buf;
int dimprod=1,i;
for (i=0;i<ndims;++i)
dimprod *= dims[i];
buf = (kiss_fft_cpx *) malloc (sizeof (kiss_fft_cpx) * dimprod);
st = kiss_fftnd_alloc (dims, ndims, isinverse, 0, 0);
while (fread (buf, sizeof (kiss_fft_cpx) * dimprod, 1, fin) > 0) {
kiss_fftnd (st, buf, buf);
fwrite (buf, sizeof (kiss_fft_cpx), dimprod, fout);
}
free (st);
free (buf);
}
static
void fft_filend_real(FILE * fin,FILE * fout,int *dims,int ndims,int isinverse)
{
int dimprod=1,i;
kiss_fftndr_cfg st;
void *ibuf;
void *obuf;
int insize,outsize; // size in bytes
for (i=0;i<ndims;++i)
dimprod *= dims[i];
insize = outsize = dimprod;
int rdim = dims[ndims-1];
if (isinverse)
insize = insize*2*(rdim/2+1)/rdim;
else
outsize = outsize*2*(rdim/2+1)/rdim;
ibuf = malloc(insize*sizeof(kiss_fft_scalar));
obuf = malloc(outsize*sizeof(kiss_fft_scalar));
st = kiss_fftndr_alloc(dims, ndims, isinverse, 0, 0);
while ( fread (ibuf, sizeof(kiss_fft_scalar), insize, fin) > 0) {
if (isinverse) {
kiss_fftndri(st,
(kiss_fft_cpx*)ibuf,
(kiss_fft_scalar*)obuf);
}else{
kiss_fftndr(st,
(kiss_fft_scalar*)ibuf,
(kiss_fft_cpx*)obuf);
}
fwrite (obuf, sizeof(kiss_fft_scalar), outsize,fout);
}
free(st);
free(ibuf);
free(obuf);
}
static
void fft_file_real(FILE * fin,FILE * fout,int nfft,int isinverse)
{
kiss_fftr_cfg st;
kiss_fft_scalar * rbuf;
kiss_fft_cpx * cbuf;
rbuf = (kiss_fft_scalar*)malloc(sizeof(kiss_fft_scalar) * nfft );
cbuf = (kiss_fft_cpx*)malloc(sizeof(kiss_fft_cpx) * (nfft/2+1) );
st = kiss_fftr_alloc( nfft ,isinverse ,0,0);
if (isinverse==0) {
while ( fread( rbuf , sizeof(kiss_fft_scalar) * nfft ,1, fin ) > 0 ) {
kiss_fftr( st , rbuf ,cbuf);
fwrite( cbuf , sizeof(kiss_fft_cpx) , (nfft/2 + 1) , fout );
}
}else{
while ( fread( cbuf , sizeof(kiss_fft_cpx) * (nfft/2+1) ,1, fin ) > 0 ) {
kiss_fftri( st , cbuf ,rbuf);
fwrite( rbuf , sizeof(kiss_fft_scalar) , nfft , fout );
}
}
free(st);
free(rbuf);
free(cbuf);
}
static
int get_dims(char * arg,int * dims)
{
char *p0;
int ndims=0;
do{
p0 = strchr(arg,',');
if (p0)
*p0++ = '\0';
dims[ndims++] = atoi(arg);
// fprintf(stderr,"dims[%d] = %d\n",ndims-1,dims[ndims-1]);
arg = p0;
}while (p0);
return ndims;
}
int main(int argc,char ** argv)
{
int isinverse=0;
int isreal=0;
FILE *fin=stdin;
FILE *fout=stdout;
int ndims=1;
int dims[32];
dims[0] = 1024; /*default fft size*/
while (1) {
int c=getopt(argc,argv,"n:iR");
if (c==-1) break;
switch (c) {
case 'n':
ndims = get_dims(optarg,dims);
break;
case 'i':isinverse=1;break;
case 'R':isreal=1;break;
case '?':
fprintf(stderr,"usage options:\n"
"\t-n d1[,d2,d3...]: fft dimension(s)\n"
"\t-i : inverse\n"
"\t-R : real input samples, not complex\n");
exit (1);
default:fprintf(stderr,"bad %c\n",c);break;
}
}
if ( optind < argc ) {
if (strcmp("-",argv[optind]) !=0)
fin = fopen(argv[optind],"rb");
++optind;
}
if ( optind < argc ) {
if ( strcmp("-",argv[optind]) !=0 )
fout = fopen(argv[optind],"wb");
++optind;
}
if (ndims==1) {
if (isreal)
fft_file_real(fin,fout,dims[0],isinverse);
else
fft_file(fin,fout,dims[0],isinverse);
}else{
if (isreal)
fft_filend_real(fin,fout,dims,ndims,isinverse);
else
fft_filend(fin,fout,dims,ndims,isinverse);
}
if (fout!=stdout) fclose(fout);
if (fin!=stdin) fclose(fin);
return 0;
}

View File

@ -0,0 +1,464 @@
/*
* Copyright (c) 2003-2004, Mark Borgerding. All rights reserved.
* This file is part of KISS FFT - https://github.com/mborgerding/kissfft
*
* SPDX-License-Identifier: BSD-3-Clause
* See COPYING file for more information.
*/
#include "_kiss_fft_guts.h"
/*
Some definitions that allow real or complex filtering
*/
#ifdef REAL_FASTFIR
#define MIN_FFT_LEN 2048
#include "kiss_fftr.h"
typedef kiss_fft_scalar kffsamp_t;
typedef kiss_fftr_cfg kfcfg_t;
#define FFT_ALLOC kiss_fftr_alloc
#define FFTFWD kiss_fftr
#define FFTINV kiss_fftri
#else
#define MIN_FFT_LEN 1024
typedef kiss_fft_cpx kffsamp_t;
typedef kiss_fft_cfg kfcfg_t;
#define FFT_ALLOC kiss_fft_alloc
#define FFTFWD kiss_fft
#define FFTINV kiss_fft
#endif
typedef struct kiss_fastfir_state *kiss_fastfir_cfg;
kiss_fastfir_cfg kiss_fastfir_alloc(const kffsamp_t * imp_resp,size_t n_imp_resp,
size_t * nfft,void * mem,size_t*lenmem);
/* see do_file_filter for usage */
size_t kiss_fastfir( kiss_fastfir_cfg cfg, kffsamp_t * inbuf, kffsamp_t * outbuf, size_t n, size_t *offset);
static int verbose=0;
struct kiss_fastfir_state{
size_t nfft;
size_t ngood;
kfcfg_t fftcfg;
kfcfg_t ifftcfg;
kiss_fft_cpx * fir_freq_resp;
kiss_fft_cpx * freqbuf;
size_t n_freq_bins;
kffsamp_t * tmpbuf;
};
kiss_fastfir_cfg kiss_fastfir_alloc(
const kffsamp_t * imp_resp,size_t n_imp_resp,
size_t *pnfft, /* if <= 0, an appropriate size will be chosen */
void * mem,size_t*lenmem)
{
kiss_fastfir_cfg st = NULL;
size_t len_fftcfg,len_ifftcfg;
size_t memneeded = sizeof(struct kiss_fastfir_state);
char * ptr;
size_t i;
size_t nfft=0;
float scale;
int n_freq_bins;
if (pnfft)
nfft=*pnfft;
if (nfft<=0) {
/* determine fft size as next power of two at least 2x
the impulse response length*/
i=n_imp_resp-1;
nfft=2;
do{
nfft<<=1;
}while (i>>=1);
#ifdef MIN_FFT_LEN
if ( nfft < MIN_FFT_LEN )
nfft=MIN_FFT_LEN;
#endif
}
if (pnfft)
*pnfft = nfft;
#ifdef REAL_FASTFIR
n_freq_bins = nfft/2 + 1;
#else
n_freq_bins = nfft;
#endif
/*fftcfg*/
FFT_ALLOC (nfft, 0, NULL, &len_fftcfg);
memneeded += len_fftcfg;
/*ifftcfg*/
FFT_ALLOC (nfft, 1, NULL, &len_ifftcfg);
memneeded += len_ifftcfg;
/* tmpbuf */
memneeded += sizeof(kffsamp_t) * nfft;
/* fir_freq_resp */
memneeded += sizeof(kiss_fft_cpx) * n_freq_bins;
/* freqbuf */
memneeded += sizeof(kiss_fft_cpx) * n_freq_bins;
if (lenmem == NULL) {
st = (kiss_fastfir_cfg) malloc (memneeded);
} else {
if (*lenmem >= memneeded)
st = (kiss_fastfir_cfg) mem;
*lenmem = memneeded;
}
if (!st)
return NULL;
st->nfft = nfft;
st->ngood = nfft - n_imp_resp + 1;
st->n_freq_bins = n_freq_bins;
ptr=(char*)(st+1);
st->fftcfg = (kfcfg_t)ptr;
ptr += len_fftcfg;
st->ifftcfg = (kfcfg_t)ptr;
ptr += len_ifftcfg;
st->tmpbuf = (kffsamp_t*)ptr;
ptr += sizeof(kffsamp_t) * nfft;
st->freqbuf = (kiss_fft_cpx*)ptr;
ptr += sizeof(kiss_fft_cpx) * n_freq_bins;
st->fir_freq_resp = (kiss_fft_cpx*)ptr;
ptr += sizeof(kiss_fft_cpx) * n_freq_bins;
FFT_ALLOC (nfft,0,st->fftcfg , &len_fftcfg);
FFT_ALLOC (nfft,1,st->ifftcfg , &len_ifftcfg);
memset(st->tmpbuf,0,sizeof(kffsamp_t)*nfft);
/*zero pad in the middle to left-rotate the impulse response
This puts the scrap samples at the end of the inverse fft'd buffer */
st->tmpbuf[0] = imp_resp[ n_imp_resp - 1 ];
for (i=0;i<n_imp_resp - 1; ++i) {
st->tmpbuf[ nfft - n_imp_resp + 1 + i ] = imp_resp[ i ];
}
FFTFWD(st->fftcfg,st->tmpbuf,st->fir_freq_resp);
/* TODO: this won't work for fixed point */
scale = 1.0 / st->nfft;
for ( i=0; i < st->n_freq_bins; ++i ) {
#ifdef USE_SIMD
st->fir_freq_resp[i].r *= _mm_set1_ps(scale);
st->fir_freq_resp[i].i *= _mm_set1_ps(scale);
#else
st->fir_freq_resp[i].r *= scale;
st->fir_freq_resp[i].i *= scale;
#endif
}
return st;
}
static void fastconv1buf(const kiss_fastfir_cfg st,const kffsamp_t * in,kffsamp_t * out)
{
size_t i;
/* multiply the frequency response of the input signal by
that of the fir filter*/
FFTFWD( st->fftcfg, in , st->freqbuf );
for ( i=0; i<st->n_freq_bins; ++i ) {
kiss_fft_cpx tmpsamp;
C_MUL(tmpsamp,st->freqbuf[i],st->fir_freq_resp[i]);
st->freqbuf[i] = tmpsamp;
}
/* perform the inverse fft*/
FFTINV(st->ifftcfg,st->freqbuf,out);
}
/* n : the size of inbuf and outbuf in samples
return value: the number of samples completely processed
n-retval samples should be copied to the front of the next input buffer */
static size_t kff_nocopy(
kiss_fastfir_cfg st,
const kffsamp_t * inbuf,
kffsamp_t * outbuf,
size_t n)
{
size_t norig=n;
while (n >= st->nfft ) {
fastconv1buf(st,inbuf,outbuf);
inbuf += st->ngood;
outbuf += st->ngood;
n -= st->ngood;
}
return norig - n;
}
static
size_t kff_flush(kiss_fastfir_cfg st,const kffsamp_t * inbuf,kffsamp_t * outbuf,size_t n)
{
size_t zpad=0,ntmp;
ntmp = kff_nocopy(st,inbuf,outbuf,n);
n -= ntmp;
inbuf += ntmp;
outbuf += ntmp;
zpad = st->nfft - n;
memset(st->tmpbuf,0,sizeof(kffsamp_t)*st->nfft );
memcpy(st->tmpbuf,inbuf,sizeof(kffsamp_t)*n );
fastconv1buf(st,st->tmpbuf,st->tmpbuf);
memcpy(outbuf,st->tmpbuf,sizeof(kffsamp_t)*( st->ngood - zpad ));
return ntmp + st->ngood - zpad;
}
size_t kiss_fastfir(
kiss_fastfir_cfg vst,
kffsamp_t * inbuf,
kffsamp_t * outbuf,
size_t n_new,
size_t *offset)
{
size_t ntot = n_new + *offset;
if (n_new==0) {
return kff_flush(vst,inbuf,outbuf,ntot);
}else{
size_t nwritten = kff_nocopy(vst,inbuf,outbuf,ntot);
*offset = ntot - nwritten;
/*save the unused or underused samples at the front of the input buffer */
memcpy( inbuf , inbuf+nwritten , *offset * sizeof(kffsamp_t) );
return nwritten;
}
}
#ifdef FAST_FILT_UTIL
#include <unistd.h>
#include <sys/types.h>
#include <sys/mman.h>
#include <assert.h>
static
void direct_file_filter(
FILE * fin,
FILE * fout,
const kffsamp_t * imp_resp,
size_t n_imp_resp)
{
size_t nlag = n_imp_resp - 1;
const kffsamp_t *tmph;
kffsamp_t *buf, *circbuf;
kffsamp_t outval;
size_t nread;
size_t nbuf;
size_t oldestlag = 0;
size_t k, tap;
#ifndef REAL_FASTFIR
kffsamp_t tmp;
#endif
nbuf = 4096;
buf = (kffsamp_t *) malloc ( sizeof (kffsamp_t) * nbuf);
circbuf = (kffsamp_t *) malloc (sizeof (kffsamp_t) * nlag);
if (!circbuf || !buf) {
perror("circbuf allocation");
exit(1);
}
if ( fread (circbuf, sizeof (kffsamp_t), nlag, fin) != nlag ) {
perror ("insufficient data to overcome transient");
exit (1);
}
do {
nread = fread (buf, sizeof (kffsamp_t), nbuf, fin);
if (nread <= 0)
break;
for (k = 0; k < nread; ++k) {
tmph = imp_resp+nlag;
#ifdef REAL_FASTFIR
# ifdef USE_SIMD
outval = _mm_set1_ps(0);
#else
outval = 0;
#endif
for (tap = oldestlag; tap < nlag; ++tap)
outval += circbuf[tap] * *tmph--;
for (tap = 0; tap < oldestlag; ++tap)
outval += circbuf[tap] * *tmph--;
outval += buf[k] * *tmph;
#else
# ifdef USE_SIMD
outval.r = outval.i = _mm_set1_ps(0);
#else
outval.r = outval.i = 0;
#endif
for (tap = oldestlag; tap < nlag; ++tap){
C_MUL(tmp,circbuf[tap],*tmph);
--tmph;
C_ADDTO(outval,tmp);
}
for (tap = 0; tap < oldestlag; ++tap) {
C_MUL(tmp,circbuf[tap],*tmph);
--tmph;
C_ADDTO(outval,tmp);
}
C_MUL(tmp,buf[k],*tmph);
C_ADDTO(outval,tmp);
#endif
circbuf[oldestlag++] = buf[k];
buf[k] = outval;
if (oldestlag == nlag)
oldestlag = 0;
}
if (fwrite (buf, sizeof (buf[0]), nread, fout) != nread) {
perror ("short write");
exit (1);
}
} while (nread);
free (buf);
free (circbuf);
}
static
void do_file_filter(
FILE * fin,
FILE * fout,
const kffsamp_t * imp_resp,
size_t n_imp_resp,
size_t nfft )
{
int fdout;
size_t n_samps_buf;
kiss_fastfir_cfg cfg;
kffsamp_t *inbuf,*outbuf;
int nread,nwrite;
size_t idx_inbuf;
fdout = fileno(fout);
cfg=kiss_fastfir_alloc(imp_resp,n_imp_resp,&nfft,0,0);
/* use length to minimize buffer shift*/
n_samps_buf = 8*4096/sizeof(kffsamp_t);
n_samps_buf = nfft + 4*(nfft-n_imp_resp+1);
if (verbose) fprintf(stderr,"bufsize=%d\n",(int)(sizeof(kffsamp_t)*n_samps_buf) );
/*allocate space and initialize pointers */
inbuf = (kffsamp_t*)malloc(sizeof(kffsamp_t)*n_samps_buf);
outbuf = (kffsamp_t*)malloc(sizeof(kffsamp_t)*n_samps_buf);
idx_inbuf=0;
do{
/* start reading at inbuf[idx_inbuf] */
nread = fread( inbuf + idx_inbuf, sizeof(kffsamp_t), n_samps_buf - idx_inbuf,fin );
/* If nread==0, then this is a flush.
The total number of samples in input is idx_inbuf + nread . */
nwrite = kiss_fastfir(cfg, inbuf, outbuf,nread,&idx_inbuf) * sizeof(kffsamp_t);
/* kiss_fastfir moved any unused samples to the front of inbuf and updated idx_inbuf */
if ( write(fdout, outbuf, nwrite) != nwrite ) {
perror("short write");
exit(1);
}
}while ( nread );
free(cfg);
free(inbuf);
free(outbuf);
}
int main(int argc,char**argv)
{
kffsamp_t * h;
int use_direct=0;
size_t nh,nfft=0;
FILE *fin=stdin;
FILE *fout=stdout;
FILE *filtfile=NULL;
while (1) {
int c=getopt(argc,argv,"n:h:i:o:vd");
if (c==-1) break;
switch (c) {
case 'v':
verbose=1;
break;
case 'n':
nfft=atoi(optarg);
break;
case 'i':
fin = fopen(optarg,"rb");
if (fin==NULL) {
perror(optarg);
exit(1);
}
break;
case 'o':
fout = fopen(optarg,"w+b");
if (fout==NULL) {
perror(optarg);
exit(1);
}
break;
case 'h':
filtfile = fopen(optarg,"rb");
if (filtfile==NULL) {
perror(optarg);
exit(1);
}
break;
case 'd':
use_direct=1;
break;
case '?':
fprintf(stderr,"usage options:\n"
"\t-n nfft: fft size to use\n"
"\t-d : use direct FIR filtering, not fast convolution\n"
"\t-i filename: input file\n"
"\t-o filename: output(filtered) file\n"
"\t-n nfft: fft size to use\n"
"\t-h filename: impulse response\n");
exit (1);
default:fprintf(stderr,"bad %c\n",c);break;
}
}
if (filtfile==NULL) {
fprintf(stderr,"You must supply the FIR coeffs via -h\n");
exit(1);
}
fseek(filtfile,0,SEEK_END);
nh = ftell(filtfile) / sizeof(kffsamp_t);
if (verbose) fprintf(stderr,"%d samples in FIR filter\n",(int)nh);
h = (kffsamp_t*)malloc(sizeof(kffsamp_t)*nh);
fseek(filtfile,0,SEEK_SET);
if (fread(h,sizeof(kffsamp_t),nh,filtfile) != nh)
fprintf(stderr,"short read on filter file\n");
fclose(filtfile);
if (use_direct)
direct_file_filter( fin, fout, h,nh);
else
do_file_filter( fin, fout, h,nh,nfft);
if (fout!=stdout) fclose(fout);
if (fin!=stdin) fclose(fin);
return 0;
}
#endif

View File

@ -0,0 +1,232 @@
/*
* Copyright (c) 2003-2004, Mark Borgerding. All rights reserved.
* This file is part of KISS FFT - https://github.com/mborgerding/kissfft
*
* SPDX-License-Identifier: BSD-3-Clause
* See COPYING file for more information.
*/
#include <stdlib.h>
#include <math.h>
#include <stdio.h>
#include <string.h>
#include <unistd.h>
#include <png.h>
#include "kiss_fft.h"
#include "kiss_fftr.h"
int nfft=1024;
FILE * fin=NULL;
FILE * fout=NULL;
int navg=20;
int remove_dc=0;
int nrows=0;
float * vals=NULL;
int stereo=0;
static
void config(int argc,char** argv)
{
while (1) {
int c = getopt (argc, argv, "n:r:as");
if (c == -1)
break;
switch (c) {
case 'n': nfft=(int)atoi(optarg);break;
case 'r': navg=(int)atoi(optarg);break;
case 'a': remove_dc=1;break;
case 's': stereo=1;break;
case '?':
fprintf (stderr, "usage options:\n"
"\t-n d: fft dimension(s) [1024]\n"
"\t-r d: number of rows to average [20]\n"
"\t-a : remove average from each fft buffer\n"
"\t-s : input is stereo, channels will be combined before fft\n"
"16 bit machine format real input is assumed\n"
);
break;
default:
fprintf (stderr, "bad %c\n", c);
exit (1);
break;
}
}
if ( optind < argc ) {
if (strcmp("-",argv[optind]) !=0)
fin = fopen(argv[optind],"rb");
++optind;
}
if ( optind < argc ) {
if ( strcmp("-",argv[optind]) !=0 )
fout = fopen(argv[optind],"wb");
++optind;
}
if (fin==NULL)
fin=stdin;
if (fout==NULL)
fout=stdout;
}
#define CHECKNULL(p) if ( (p)==NULL ) do { fprintf(stderr,"CHECKNULL failed @ %s(%d): %s\n",__FILE__,__LINE__,#p );exit(1);} while(0)
typedef struct
{
png_byte r;
png_byte g;
png_byte b;
} rgb_t;
static
void val2rgb(float x,rgb_t *p)
{
const double pi = 3.14159265358979;
p->g = (int)(255*sin(x*pi));
p->r = (int)(255*abs(sin(x*pi*3/2)));
p->b = (int)(255*abs(sin(x*pi*5/2)));
//fprintf(stderr,"%.2f : %d,%d,%d\n",x,(int)p->r,(int)p->g,(int)p->b);
}
static
void cpx2pixels(rgb_t * res,const float * fbuf,size_t n)
{
size_t i;
float minval,maxval,valrange;
minval=maxval=fbuf[0];
for (i = 0; i < n; ++i) {
if (fbuf[i] > maxval) maxval = fbuf[i];
if (fbuf[i] < minval) minval = fbuf[i];
}
fprintf(stderr,"min ==%f,max=%f\n",minval,maxval);
valrange = maxval-minval;
if (valrange == 0) {
fprintf(stderr,"min == max == %f\n",minval);
exit (1);
}
for (i = 0; i < n; ++i)
val2rgb( (fbuf[i] - minval)/valrange , res+i );
}
static
void transform_signal(void)
{
short *inbuf;
kiss_fftr_cfg cfg=NULL;
kiss_fft_scalar *tbuf;
kiss_fft_cpx *fbuf;
float *mag2buf;
int i;
int n;
int avgctr=0;
int nfreqs=nfft/2+1;
CHECKNULL( cfg=kiss_fftr_alloc(nfft,0,0,0) );
CHECKNULL( inbuf=(short*)malloc(sizeof(short)*2*nfft ) );
CHECKNULL( tbuf=(kiss_fft_scalar*)malloc(sizeof(kiss_fft_scalar)*nfft ) );
CHECKNULL( fbuf=(kiss_fft_cpx*)malloc(sizeof(kiss_fft_cpx)*nfreqs ) );
CHECKNULL( mag2buf=(float*)calloc(nfreqs,sizeof(float) ) );
while (1) {
if (stereo) {
n = fread(inbuf,sizeof(short)*2,nfft,fin);
if (n != nfft )
break;
for (i=0;i<nfft;++i)
tbuf[i] = inbuf[2*i] + inbuf[2*i+1];
}else{
n = fread(inbuf,sizeof(short),nfft,fin);
if (n != nfft )
break;
for (i=0;i<nfft;++i)
tbuf[i] = inbuf[i];
}
if (remove_dc) {
float avg = 0;
for (i=0;i<nfft;++i) avg += tbuf[i];
avg /= nfft;
for (i=0;i<nfft;++i) tbuf[i] -= (kiss_fft_scalar)avg;
}
/* do FFT */
kiss_fftr(cfg,tbuf,fbuf);
for (i=0;i<nfreqs;++i)
mag2buf[i] += fbuf[i].r * fbuf[i].r + fbuf[i].i * fbuf[i].i;
if (++avgctr == navg) {
avgctr=0;
++nrows;
CHECKNULL( vals = (float*)realloc(vals,sizeof(float)*nrows*nfreqs) );
float eps = 1;
for (i=0;i<nfreqs;++i)
vals[(nrows - 1) * nfreqs + i] = 10 * log10 ( mag2buf[i] / navg + eps );
memset(mag2buf,0,sizeof(mag2buf[0])*nfreqs);
}
}
free(cfg);
free(inbuf);
free(tbuf);
free(fbuf);
free(mag2buf);
}
static
void make_png(void)
{
png_bytepp row_pointers=NULL;
rgb_t * row_data=NULL;
int i;
int nfreqs = nfft/2+1;
png_structp png_ptr=NULL;
png_infop info_ptr=NULL;
CHECKNULL( png_ptr = png_create_write_struct (PNG_LIBPNG_VER_STRING,0,0,0) );
CHECKNULL( info_ptr = png_create_info_struct(png_ptr) );
png_init_io(png_ptr, fout );
png_set_IHDR(png_ptr, info_ptr ,nfreqs,nrows,8,PNG_COLOR_TYPE_RGB,PNG_INTERLACE_NONE,PNG_COMPRESSION_TYPE_DEFAULT,PNG_FILTER_TYPE_DEFAULT );
CHECKNULL( row_data = (rgb_t*)malloc(sizeof(rgb_t) * nrows * nfreqs) );
cpx2pixels(row_data, vals, nfreqs*nrows );
CHECKNULL( row_pointers = malloc(nrows*sizeof(png_bytep)) );
for (i=0;i<nrows;++i) {
row_pointers[i] = (png_bytep)(row_data + i*nfreqs);
}
png_set_rows(png_ptr, info_ptr, row_pointers);
fprintf(stderr,"creating %dx%d png\n",nfreqs,nrows);
fprintf(stderr,"bitdepth %d \n",png_get_bit_depth(png_ptr,info_ptr ) );
png_write_png(png_ptr, info_ptr, PNG_TRANSFORM_IDENTITY , NULL);
free(row_pointers);
free(row_data);
png_destroy_write_struct(&png_ptr, &info_ptr);
}
int main(int argc,char ** argv)
{
config(argc,argv);
transform_signal();
make_png();
if (fout!=stdout) fclose(fout);
if (fin!=stdin) fclose(fin);
free(vals);
return 0;
}

Some files were not shown because too many files have changed in this diff Show More