razzaline/tahoma2d
1
0
Fork 0
Code Issues Pull requests Packages Projects Releases 1 Wiki Activity

bokeh fx iwa

Browse source
This commit is contained in:
shun_iwasawa 2017-03-16 15:29:47 +09:00
parent 54b714a0b6
commit a35b8f8c48
54 changed files with 6115 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

20
stuff/config/current.txt
View file

@ -1197,6 +1197,26 @@
<item>"STD_iwa_SoapBubbleFx.noiseDepthMixRatio" "Noise to Depth" </item>
<item>"STD_iwa_SoapBubbleFx.noiseThicknessMixRatio" "Noise to Thickness" </item>
<item>"STD_iwa_BokehFx" "Bokeh Iwa"</item>
<item>"STD_iwa_BokehFx.on_focus_distance" "On-Focus Distance"</item>
<item>"STD_iwa_BokehFx.bokeh_amount" "Bokeh Amount"</item>
<item>"STD_iwa_BokehFx.hardness" "Hardness"</item>
<item>"STD_iwa_BokehFx.premultiply1" "Layer1 Premultiply"</item>
<item>"STD_iwa_BokehFx.distance1" "Layer1 Distance"</item>
<item>"STD_iwa_BokehFx.bokeh_adjustment1" "Layer1 Bokeh Adjustment"</item>
<item>"STD_iwa_BokehFx.premultiply2" "Layer2 Premultiply"</item>
<item>"STD_iwa_BokehFx.distance2" "Layer2 Distance"</item>
<item>"STD_iwa_BokehFx.bokeh_adjustment2" "Layer2 Bokeh Adjustment"</item>
<item>"STD_iwa_BokehFx.premultiply3" "Layer3 Premultiply"</item>
<item>"STD_iwa_BokehFx.distance3" "Layer3 Distance"</item>
<item>"STD_iwa_BokehFx.bokeh_adjustment3" "Layer3 Bokeh Adjustment"</item>
<item>"STD_iwa_BokehFx.premultiply4" "Layer4 Premultiply"</item>
<item>"STD_iwa_BokehFx.distance4" "Layer4 Distance"</item>
<item>"STD_iwa_BokehFx.bokeh_adjustment4" "Layer4 Bokeh Adjustment"</item>
<item>"STD_iwa_BokehFx.premultiply5" "Layer5 Premultiply"</item>
<item>"STD_iwa_BokehFx.distance5" "Layer5 Distance"</item>
<item>"STD_iwa_BokehFx.bokeh_adjustment5" "Layer5 Bokeh Adjustment"</item>
<!------------------------------ Tiled Particles Iwa ------------------------------------------->
<item>STD_iwa_TiledParticlesFx "Tiled Particles Iwa" </item>

63
stuff/doc/BokehIwa.html Normal file
View file

@ -0,0 +1,63 @@
<html lang="ja">
<head>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
<title>Bokeh Fx Iwa</title>
</head>
<body bgcolor="#f5f5f5" text="#220011">
<h1><img src = ".\img\fx_iwa_bokeh.png" width = 30 height = 30 > Bokeh Fx Iwa</h1>
<h4>● 概要</h4>
レンズのボケを再現するエフェクトです。各レイヤのRGB値を露光値に変換して、絞り形状でボカし、合成します。<br>
フィルタ処理にフーリエ変換を用いて高速化を図っています。
<h4>● 入力ポート</h4>
<UL>
<LI><b>Iris</b> : 絞り画像を接続します。入力された画像の輝度値がフィルタに用いられます。Irisポートに何も接続されていない場合は、計算が行われません。RGBA8bit又はRGBA16bit画像が入力できます。
<LI><b>Source[1～5]</b> : レイヤー画像を接続します。ここでの接続の順番は、レイヤーの重ね順に影響しません。全てのSourceポートに何も接続されていない場合は、計算が行われません。
</UL>
<h4>● パラメータ</h4>
共通パラメータ
<UL>
<LI><b>On-Focus Distance</b> : フォーカス位置。この位置にレイヤーがあると、
そのレイヤーはボカされず、通常合成されます。0がカメラ位置です。(範囲 0.0～1.0)
<LI><b>Bokeh Amount</b> : ボケの最大サイズ（単位 Unit）。フォーカス位置とレイヤー位置が 1.0 離れていて、Bokeh Adjustmentが 1 のとき、絞り画像の横幅がこの値になるまで拡大されて用いられます。
<LI><b>Hardness</b> : フィルムのガンマ値。RGB値と露光量の変換に用います。露光量が10倍増えた時の、RGB値（0.0～1.0）の増加量に相当します。この値が小さいほど、ハイライトが強調されます。(範囲 0.05～3.0)
</UL>
レイヤー毎のパラメータ
<UL>
<LI><b>Premultiply</b> : DigiBook等、Premultiplyされていない素材を直接このFxに接続するときは、このチェックを入れて下さい。
<LI><b>Distance</b> : レイヤーのカメラからの距離。この値に合わせ、レイヤーの重なる順序が自動的にソートされます。(範囲 0.0～1.0)
<LI><b>Bokeh Adjustment</b> : ボケサイズの補正値。レイヤーの重なる順序はそのままに、ボケのサイズが N 倍されます。この値が 0 なら、どの距離にレイヤーを置いても、ボケずに通常合成されます。(範囲 0.0～2.0)
</UL>
<h4>● 注意点</h4>
<UL>
<LI>複数フレームをレンダリングする際、レイヤーだけでなくIrisの素材も、レンダリングされるすべてのフレームの範囲に入っている必要があります。
<LI>メモリを多く使います。
</UL>
<h4>● ライセンス情報</h4>
<UL>
<LI>このエフェクトは、フーリエ変換にKiss FFTというオープンソース・ライブラリを用いています。<br>
</UL>
<font size = "-1">
<blockquote>
This is the BSD-style license for the KissFFT.<br>
<br>
Copyright (c) 2003-2010 Mark Borgerding<br>
<br>
All rights reserved.<br>
<br>
Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:<br>
<br>
* Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.<br>
* 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.<br>
* 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.<br>
<br>
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.
</blockquote>
</font>
</body>
</html>

BIN
stuff/doc/img/fx_iwa_bokeh.png Normal file
View file

Binary file not shown.
Side by side

After

Width:  |  Height:  |  Size: 782 B

44
stuff/profiles/layouts/fxs/STD_iwa_BokehFx.xml Normal file
View file

@ -0,0 +1,44 @@
<fxlayout help_command="iexplore" help_file="BokehIwa.html">
<page name="Bokeh Iwa">
<vbox>
<control>on_focus_distance</control>
<control>bokeh_amount</control>
<control>hardness</control>
</vbox>
<separator label="Layer1"/>
<vbox>
<control>premultiply1</control>
<control>distance1</control>
<control>bokeh_adjustment1</control>
</vbox>
<separator label="Layer2"/>
<vbox>
<control>premultiply2</control>
<control>distance2</control>
<control>bokeh_adjustment2</control>
</vbox>
<separator label="Layer3"/>
<vbox>
<control>premultiply3</control>
<control>distance3</control>
<control>bokeh_adjustment3</control>
</vbox>
<separator label="Layer4"/>
<vbox>
<control>premultiply4</control>
<control>distance4</control>
<control>bokeh_adjustment4</control>
</vbox>
<separator label="Layer5"/>
<vbox>
<control>premultiply5</control>
<control>distance5</control>
<control>bokeh_adjustment5</control>
</vbox>
</page>
</fxlayout>

1
stuff/profiles/layouts/fxs/fxs.lst
View file

@ -20,6 +20,7 @@
STD_inoRadialBlurFx
STD_rotationalBlurFx
STD_inoSpinBlurFx
STD_iwa_BokehFx
</Blur>
<Distort>
STD_freeDistortFx

2
thirdparty/kiss_fft130/.gitignore vendored Normal file
View file

@ -0,0 +1,2 @@
*
!.gitignore

4
thirdparty/kiss_fft130/.hg_archival.txt vendored Normal file
View file

@ -0,0 +1,4 @@
repo: d844c818f599aea64fe86745cdd2ef9b3d1910dc
node: b354a59534b0a77c43c67deb1eb1bc39eb99b487
branch: default
tag: v130

10
thirdparty/kiss_fft130/.hgignore vendored Normal file
View file

@ -0,0 +1,10 @@
syntax:glob
test/bm_*
test/st_*
test/tkfc_*
test/tr_*
tools/fastconv_*
tools/fastconvr_*
tools/fft_*
*.swp
*~

20
thirdparty/kiss_fft130/.hgtags vendored Normal file
View file

@ -0,0 +1,20 @@
32061530d4353ced38386a4578153954f46e5f8e v04_prehg
52a9ca23c9d9600794388912103e0f7d5f4c92ee v1_2_2_prehg
68e0375761a5789d72c4baa2f8265cec26fb546e v110_prehg
69a218df7458f0269a84b07f64678392edbff941 v111_prehg
702f90b11f060d45b23d563764a3b1bd8f851b7d about2do_real_multi_d_prehg
9c3041fb0677c071d7ca2797e22a63fe681d311b b4simd_prehg
9e532c64d324a0844f1efa7c32908fc2bc0cf350 v1_2_6_prehg
a9797b79bf2c61f7c8259ec417a0189caa48b17a v1_2_3_prehg
b1b2739c82378d6e04977440b2c31b7d1b34c266 aftersimd_prehg
b8c210e0bccdeb66930423a528b11b3660491af8 v120_prehg
c16172181d6aef63712a3f7bc31ef555ce7178bd help_prehg
c16172181d6aef63712a3f7bc31ef555ce7178bd v1_2_3a_prehg
c83c1ec6a5f21c40edf3bb9cbfd8ba1f862f1e0f half_bottle_o_wine_prehg
dec01bc9c4c5f807e2e04aef3449031bc5c6b69e v1_2_5_prehg
e6a840a1383ffc83c8ba17ed4b0bf5d6aefc9a34 v127_prehg
ea3a56794d9326d5505b5bf02f98a70217a1b86d v101_prehg
edd6236ecb25904b385856ffae608548ce3df4e5 v011_prehg
f19b7bcef3f322d277d5b51b2b132657cf08c1a7 v1_2_8_prehg
f73f8d58c37d52379e9d7f13d81f5391226209c1 v1_2_1_prehg
9e0bf8478cc337da0de18b7413d51c714aa0db46 v129

123
thirdparty/kiss_fft130/CHANGELOG vendored Normal file
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

11
thirdparty/kiss_fft130/COPYING vendored Normal file
View file

@ -0,0 +1,11 @@
Copyright (c) 2003-2010 Mark Borgerding
All rights reserved.
Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
* 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.
* 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.
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.

33
thirdparty/kiss_fft130/Makefile vendored Normal file
View file

@ -0,0 +1,33 @@
KFVER=130
doc:
@echo "Start by reading the README file. If you want to build and test lots of stuff, do a 'make testall'"
@echo "but be aware that 'make testall' has dependencies that the basic kissfft software does not."
@echo "It is generally unneeded to run these tests yourself, unless you plan on changing the inner workings"
@echo "of kissfft and would like to make use of its regression tests."
testall:
# The simd and int32_t types may or may not work on your machine
make -C test DATATYPE=simd CFLAGADD="$(CFLAGADD)" test
make -C test DATATYPE=int32_t CFLAGADD="$(CFLAGADD)" test
make -C test DATATYPE=int16_t CFLAGADD="$(CFLAGADD)" test
make -C test DATATYPE=float CFLAGADD="$(CFLAGADD)" test
make -C test DATATYPE=double CFLAGADD="$(CFLAGADD)" test
echo "all tests passed"
tarball: clean
hg archive -r v$(KFVER) -t tgz kiss_fft$(KFVER).tar.gz
hg archive -r v$(KFVER) -t zip kiss_fft$(KFVER).zip
clean:
cd test && make clean
cd tools && make clean
rm -f kiss_fft*.tar.gz *~ *.pyc kiss_fft*.zip
asm: kiss_fft.s
kiss_fft.s: kiss_fft.c kiss_fft.h _kiss_fft_guts.h
[ -e kiss_fft.s ] && mv kiss_fft.s kiss_fft.s~ || true
gcc -S kiss_fft.c -O3 -mtune=native -ffast-math -fomit-frame-pointer -unroll-loops -dA -fverbose-asm
gcc -o kiss_fft_short.s -S kiss_fft.c -O3 -mtune=native -ffast-math -fomit-frame-pointer -dA -fverbose-asm -DFIXED_POINT
[ -e kiss_fft.s~ ] && diff kiss_fft.s~ kiss_fft.s || true

134
thirdparty/kiss_fft130/README vendored Normal file
View file

@ -0,0 +1,134 @@
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:
#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
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.
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?
A: Yes. See LICENSE below.
Q: Why don't I get the output I expect?
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?
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
A commercial license is available which removes the requirement for attribution. Contact me for details.
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

78
thirdparty/kiss_fft130/README.simd vendored Normal file
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;
}
}

39
thirdparty/kiss_fft130/TIPS vendored Normal file
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.

164
thirdparty/kiss_fft130/_kiss_fft_guts.h vendored Normal file
View file

@ -0,0 +1,164 @@
/*
Copyright (c) 2003-2010, Mark Borgerding
All rights reserved.
Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
* 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.
* 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.
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.
*/
/* 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; */
#include "kiss_fft.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
#if (FIXED_POINT==32)
# define FRACBITS 31
# define SAMPPROD int64_t
#define SAMP_MAX 2147483647
#else
# define FRACBITS 15
# define SAMPPROD int32_t
#define SAMP_MAX 32767
#endif
#define SAMP_MIN -SAMP_MAX
#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 ) { \
fprintf(stderr,"WARNING:overflow @ " __FILE__ "(%d): (%d " #op" %d) = %ld\n",__LINE__,(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)*.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)\
fprintf(stderr,"%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

408
thirdparty/kiss_fft130/kiss_fft.c vendored Normal file
View file

@ -0,0 +1,408 @@
/*
Copyright (c) 2003-2010, Mark Borgerding
All rights reserved.
Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
* 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.
* 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.
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.
*/
#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);
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)
{
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_cfg st=NULL;
size_t memneeded = 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
kiss_fft_cpx * tmpbuf = (kiss_fft_cpx*)KISS_FFT_TMP_ALLOC( sizeof(kiss_fft_cpx)*st->nfft);
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;
}

124
thirdparty/kiss_fft130/kiss_fft.h vendored Normal file
View file

@ -0,0 +1,124 @@
#ifndef KISS_FFT_H
#define KISS_FFT_H
#include <stdlib.h>
#include <stdio.h>
#include <math.h>
#include <string.h>
#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.
*/
#ifdef USE_SIMD
# include <xmmintrin.h>
# define kiss_fft_scalar __m128
#define KISS_FFT_MALLOC(nbytes) _mm_malloc(nbytes,16)
#define KISS_FFT_FREE _mm_free
#else
#define KISS_FFT_MALLOC malloc
#define KISS_FFT_FREE free
#endif
#ifdef FIXED_POINT
#include <sys/types.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_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(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_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 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_cleanup(void);
/*
* Returns the smallest integer k, such that k>=n and k has only "fast" factors (2,3,5)
*/
int 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

299
thirdparty/kiss_fft130/kissfft.hh vendored Normal file
View file

@ -0,0 +1,299 @@
#ifndef KISSFFT_CLASS_HH
#include <complex>
#include <vector>
namespace kissfft_utils {
template <typename T_scalar>
struct traits
{
typedef T_scalar scalar_type;
typedef std::complex<scalar_type> cpx_type;
void fill_twiddles( std::complex<T_scalar> * dst ,int nfft,bool inverse)
{
T_scalar phinc = (inverse?2:-2)* acos( (T_scalar) -1) / nfft;
for (int i=0;i<nfft;++i)
dst[i] = exp( std::complex<T_scalar>(0,i*phinc) );
}
void prepare(
std::vector< std::complex<T_scalar> > & dst,
int nfft,bool inverse,
std::vector<int> & stageRadix,
std::vector<int> & stageRemainder )
{
_twiddles.resize(nfft);
fill_twiddles( &_twiddles[0],nfft,inverse);
dst = _twiddles;
//factorize
//start factoring out 4's, then 2's, then 3,5,7,9,...
int n= nfft;
int 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);
}
std::vector<cpx_type> _twiddles;
const cpx_type twiddle(int i) { return _twiddles[i]; }
};
}
template <typename T_Scalar,
typename T_traits=kissfft_utils::traits<T_Scalar>
>
class kissfft
{
public:
typedef T_traits traits_type;
typedef typename traits_type::scalar_type scalar_type;
typedef typename traits_type::cpx_type cpx_type;
kissfft(int nfft,bool inverse,const traits_type & traits=traits_type() )
:_nfft(nfft),_inverse(inverse),_traits(traits)
{
_traits.prepare(_twiddles, _nfft,_inverse ,_stageRadix, _stageRemainder);
}
void transform(const cpx_type * src , cpx_type * dst)
{
kf_work(0, dst, src, 1,1);
}
private:
void kf_work( int stage,cpx_type * Fout, const cpx_type * f, size_t fstride,size_t in_stride)
{
int p = _stageRadix[stage];
int m = _stageRemainder[stage];
cpx_type * Fout_beg = Fout;
cpx_type * Fout_end = Fout + p*m;
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(stage+1, Fout , f, fstride*p,in_stride);
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,m); break;
case 3: kf_bfly3(Fout,fstride,m); break;
case 4: kf_bfly4(Fout,fstride,m); break;
case 5: kf_bfly5(Fout,fstride,m); break;
default: kf_bfly_generic(Fout,fstride,m,p); break;
}
}
// these were #define macros in the original kiss_fft
void C_ADD( cpx_type & c,const cpx_type & a,const cpx_type & b) { c=a+b;}
void C_MUL( cpx_type & c,const cpx_type & a,const cpx_type & b) { c=a*b;}
void C_SUB( cpx_type & c,const cpx_type & a,const cpx_type & b) { c=a-b;}
void C_ADDTO( cpx_type & c,const cpx_type & a) { c+=a;}
void C_FIXDIV( cpx_type & ,int ) {} // NO-OP for float types
scalar_type S_MUL( const scalar_type & a,const scalar_type & b) { return a*b;}
scalar_type HALF_OF( const scalar_type & a) { return a*.5;}
void C_MULBYSCALAR(cpx_type & c,const scalar_type & a) {c*=a;}
void kf_bfly2( cpx_type * Fout, const size_t fstride, int m)
{
for (int k=0;k<m;++k) {
cpx_type t = Fout[m+k] * _traits.twiddle(k*fstride);
Fout[m+k] = Fout[k] - t;
Fout[k] += t;
}
}
void kf_bfly4( cpx_type * Fout, const size_t fstride, const size_t m)
{
cpx_type scratch[7];
int negative_if_inverse = _inverse * -2 +1;
for (size_t k=0;k<m;++k) {
scratch[0] = Fout[k+m] * _traits.twiddle(k*fstride);
scratch[1] = Fout[k+2*m] * _traits.twiddle(k*fstride*2);
scratch[2] = Fout[k+3*m] * _traits.twiddle(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_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_bfly3( cpx_type * Fout, const size_t fstride, const size_t m)
{
size_t k=m;
const size_t m2 = 2*m;
cpx_type *tw1,*tw2;
cpx_type scratch[5];
cpx_type epi3;
epi3 = _twiddles[fstride*m];
tw1=tw2=&_twiddles[0];
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] = cpx_type( Fout->real() - HALF_OF(scratch[3].real() ) , Fout->imag() - HALF_OF(scratch[3].imag() ) );
C_MULBYSCALAR( scratch[0] , epi3.imag() );
C_ADDTO(*Fout,scratch[3]);
Fout[m2] = cpx_type( Fout[m].real() + scratch[0].imag() , Fout[m].imag() - scratch[0].real() );
C_ADDTO( Fout[m] , cpx_type( -scratch[0].imag(),scratch[0].real() ) );
++Fout;
}while(--k);
}
void kf_bfly5( cpx_type * Fout, const size_t fstride, const size_t m)
{
cpx_type *Fout0,*Fout1,*Fout2,*Fout3,*Fout4;
size_t u;
cpx_type scratch[13];
cpx_type * twiddles = &_twiddles[0];
cpx_type *tw;
cpx_type 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=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]);
C_ADDTO( *Fout0, scratch[7]);
C_ADDTO( *Fout0, scratch[8]);
scratch[5] = scratch[0] + cpx_type(
S_MUL(scratch[7].real(),ya.real() ) + S_MUL(scratch[8].real() ,yb.real() ),
S_MUL(scratch[7].imag(),ya.real()) + S_MUL(scratch[8].imag(),yb.real())
);
scratch[6] = cpx_type(
S_MUL(scratch[10].imag(),ya.imag()) + S_MUL(scratch[9].imag(),yb.imag()),
-S_MUL(scratch[10].real(),ya.imag()) - S_MUL(scratch[9].real(),yb.imag())
);
C_SUB(*Fout1,scratch[5],scratch[6]);
C_ADD(*Fout4,scratch[5],scratch[6]);
scratch[11] = scratch[0] +
cpx_type(
S_MUL(scratch[7].real(),yb.real()) + S_MUL(scratch[8].real(),ya.real()),
S_MUL(scratch[7].imag(),yb.real()) + S_MUL(scratch[8].imag(),ya.real())
);
scratch[12] = cpx_type(
-S_MUL(scratch[10].imag(),yb.imag()) + S_MUL(scratch[9].imag(),ya.imag()),
S_MUL(scratch[10].real(),yb.imag()) - S_MUL(scratch[9].real(),ya.imag())
);
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 */
void kf_bfly_generic(
cpx_type * Fout,
const size_t fstride,
int m,
int p
)
{
int u,k,q1,q;
cpx_type * twiddles = &_twiddles[0];
cpx_type t;
int Norig = _nfft;
cpx_type scratchbuf[p];
for ( u=0; u<m; ++u ) {
k=u;
for ( q1=0 ; q1<p ; ++q1 ) {
scratchbuf[q1] = Fout[ k ];
C_FIXDIV(scratchbuf[q1],p);
k += m;
}
k=u;
for ( q1=0 ; q1<p ; ++q1 ) {
int twidx=0;
Fout[ k ] = scratchbuf[0];
for (q=1;q<p;++q ) {
twidx += fstride * k;
if (twidx>=Norig) twidx-=Norig;
C_MUL(t,scratchbuf[q] , twiddles[twidx] );
C_ADDTO( Fout[ k ] ,t);
}
k += m;
}
}
}
int _nfft;
bool _inverse;
std::vector<cpx_type> _twiddles;
std::vector<int> _stageRadix;
std::vector<int> _stageRemainder;
traits_type _traits;
};
#endif

108
thirdparty/kiss_fft130/test/Makefile vendored Normal file
View file

@ -0,0 +1,108 @@
WARNINGS=-W -Wall -Wstrict-prototypes -Wmissing-prototypes -Waggregate-return \
-Wcast-align -Wcast-qual -Wnested-externs -Wshadow -Wbad-function-cast \
-Wwrite-strings
CFLAGS=-O3 -I.. -I../tools $(WARNINGS)
CFLAGS+=-ffast-math -fomit-frame-pointer
#CFLAGS+=-funroll-loops
#CFLAGS+=-march=prescott
#CFLAGS+= -mtune=native
# TIP: try adding -openmp or -fopenmp to enable OPENMP directives and use of multiple cores
#CFLAGS+=-fopenmp
CFLAGS+= $(CFLAGADD)
ifeq "$(NFFT)" ""
NFFT=1800
endif
ifeq "$(NUMFFTS)" ""
NUMFFTS=10000
endif
ifeq "$(DATATYPE)" ""
DATATYPE=float
endif
BENCHKISS=bm_kiss_$(DATATYPE)
BENCHFFTW=bm_fftw_$(DATATYPE)
SELFTEST=st_$(DATATYPE)
TESTREAL=tr_$(DATATYPE)
TESTKFC=tkfc_$(DATATYPE)
FASTFILTREAL=ffr_$(DATATYPE)
SELFTESTSRC=twotonetest.c
TYPEFLAGS=-Dkiss_fft_scalar=$(DATATYPE)
ifeq "$(DATATYPE)" "int16_t"
TYPEFLAGS=-DFIXED_POINT=16
endif
ifeq "$(DATATYPE)" "int32_t"
TYPEFLAGS=-DFIXED_POINT=32
endif
ifeq "$(DATATYPE)" "simd"
TYPEFLAGS=-DUSE_SIMD=1 -msse
endif
ifeq "$(DATATYPE)" "float"
# fftw needs to be built with --enable-float to build this lib
FFTWLIB=-lfftw3f
else
FFTWLIB=-lfftw3
endif
FFTWLIBDIR=-L/usr/local/lib/
SRCFILES=../kiss_fft.c ../tools/kiss_fftnd.c ../tools/kiss_fftr.c pstats.c ../tools/kfc.c ../tools/kiss_fftndr.c
all: tools $(BENCHKISS) $(SELFTEST) $(BENCHFFTW) $(TESTREAL) $(TESTKFC)
tools:
cd ../tools && make all
$(SELFTEST): $(SELFTESTSRC) $(SRCFILES)
$(CC) -o $@ $(CFLAGS) $(TYPEFLAGS) $+ -lm
$(TESTKFC): $(SRCFILES)
$(CC) -o $@ $(CFLAGS) -DKFC_TEST $(TYPEFLAGS) $+ -lm
$(TESTREAL): test_real.c $(SRCFILES)
$(CC) -o $@ $(CFLAGS) $(TYPEFLAGS) $+ -lm
$(BENCHKISS): benchkiss.c $(SRCFILES)
$(CC) -o $@ $(CFLAGS) $(TYPEFLAGS) $+ -lm
$(BENCHFFTW): benchfftw.c pstats.c
@echo "======attempting to build FFTW benchmark"
@$(CC) -o $@ $(CFLAGS) -DDATATYPE$(DATATYPE) $+ $(FFTWLIB) $(FFTWLIBDIR) -lm || echo "FFTW not available for comparison"
test: all
@./$(TESTKFC)
@echo "======1d & 2-d complex fft self test (type= $(DATATYPE) )"
@./$(SELFTEST)
@echo "======real FFT (type= $(DATATYPE) )"
@./$(TESTREAL)
@echo "======timing test (type=$(DATATYPE))"
@./$(BENCHKISS) -x $(NUMFFTS) -n $(NFFT)
@[ -x ./$(BENCHFFTW) ] && ./$(BENCHFFTW) -x $(NUMFFTS) -n $(NFFT) ||true
@echo "======higher dimensions type=$(DATATYPE))"
@./testkiss.py
selftest.c:
./mk_test.py 10 12 14 > selftest.c
selftest_short.c:
./mk_test.py -s 10 12 14 > selftest_short.c
CXXFLAGS=-O3 -ffast-math -fomit-frame-pointer -I.. -I../tools -W -Wall
testcpp: testcpp.cc ../kissfft.hh
$(CXX) -o $@ $(CXXFLAGS) testcpp.cc -lm
clean:
rm -f *~ bm_* st_* tr_* kf_* tkfc_* ff_* ffr_* *.pyc *.pyo *.dat testcpp

94
thirdparty/kiss_fft130/test/benchfftw.c vendored Normal file
View file

@ -0,0 +1,94 @@
#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

122
thirdparty/kiss_fft130/test/benchkiss.c vendored Normal file
View file

@ -0,0 +1,122 @@
#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;
}

92
thirdparty/kiss_fft130/test/compfft.py vendored Normal file
View file

@ -0,0 +1,92 @@
#!/usr/bin/env python
# use FFTPACK as a baseline
import FFT
from Numeric import *
import math
import random
import sys
import struct
import fft
pi=math.pi
e=math.e
j=complex(0,1)
lims=(-32768,32767)
def randbuf(n,cpx=1):
res = array( [ random.uniform( lims[0],lims[1] ) for i in range(n) ] )
if cpx:
res = res + j*randbuf(n,0)
return res
def main():
from getopt import getopt
import popen2
opts,args = getopt( sys.argv[1:],'u:n:Rt:' )
opts=dict(opts)
exitcode=0
util = opts.get('-u','./kf_float')
try:
dims = [ int(d) for d in opts['-n'].split(',')]
cpx = opts.get('-R') is None
fmt=opts.get('-t','f')
except KeyError:
sys.stderr.write("""
usage: compfft.py
-n d1[,d2,d3...] : FFT dimension(s)
-u utilname : see sample_code/fftutil.c, default = ./kf_float
-R : real-optimized version\n""")
sys.exit(1)
x = fft.make_random( dims )
cmd = '%s -n %s ' % ( util, ','.join([ str(d) for d in dims]) )
if cpx:
xout = FFT.fftnd(x)
xout = reshape(xout,(size(xout),))
else:
cmd += '-R '
xout = FFT.real_fft(x)
proc = popen2.Popen3( cmd , bufsize=len(x) )
proc.tochild.write( dopack( x , fmt ,cpx ) )
proc.tochild.close()
xoutcomp = dounpack( proc.fromchild.read( ) , fmt ,1 )
#xoutcomp = reshape( xoutcomp , dims )
sig = xout * conjugate(xout)
sigpow = sum( sig )
diff = xout-xoutcomp
noisepow = sum( diff * conjugate(diff) )
snr = 10 * math.log10(abs( sigpow / noisepow ) )
if snr<100:
print xout
print xoutcomp
exitcode=1
print 'NFFT=%s,SNR = %f dB' % (str(dims),snr)
sys.exit(exitcode)
def dopack(x,fmt,cpx):
x = reshape( x, ( size(x),) )
if cpx:
s = ''.join( [ struct.pack('ff',c.real,c.imag) for c in x ] )
else:
s = ''.join( [ struct.pack('f',c) for c in x ] )
return s
def dounpack(x,fmt,cpx):
uf = fmt * ( len(x) / 4 )
s = struct.unpack(uf,x)
if cpx:
return array(s[::2]) + array( s[1::2] )*j
else:
return array(s )
if __name__ == "__main__":
main()

129
thirdparty/kiss_fft130/test/doit.c vendored Normal file
View file

@ -0,0 +1,129 @@
/* this program is in the public domain
A program that helps the authors of the fine fftw library benchmark kiss
*/
#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);
}

102
thirdparty/kiss_fft130/test/fastfir.py vendored Normal file
View file

@ -0,0 +1,102 @@
#!/usr/bin/env python
from Numeric import *
from FFT import *
def make_random(len):
import random
res=[]
for i in range(int(len)):
r=random.uniform(-1,1)
i=random.uniform(-1,1)
res.append( complex(r,i) )
return res
def slowfilter(sig,h):
translen = len(h)-1
return convolve(sig,h)[translen:-translen]
def nextpow2(x):
return 2 ** math.ceil(math.log(x)/math.log(2))
def fastfilter(sig,h,nfft=None):
if nfft is None:
nfft = int( nextpow2( 2*len(h) ) )
H = fft( h , nfft )
scraplen = len(h)-1
keeplen = nfft-scraplen
res=[]
isdone = 0
lastidx = nfft
idx0 = 0
while not isdone:
idx1 = idx0 + nfft
if idx1 >= len(sig):
idx1 = len(sig)
lastidx = idx1-idx0
if lastidx <= scraplen:
break
isdone = 1
Fss = fft(sig[idx0:idx1],nfft)
fm = Fss * H
m = inverse_fft(fm)
res.append( m[scraplen:lastidx] )
idx0 += keeplen
return concatenate( res )
def main():
import sys
from getopt import getopt
opts,args = getopt(sys.argv[1:],'rn:l:')
opts=dict(opts)
siglen = int(opts.get('-l',1e4 ) )
hlen =50
nfft = int(opts.get('-n',128) )
usereal = opts.has_key('-r')
print 'nfft=%d'%nfft
# make a signal
sig = make_random( siglen )
# make an impulse response
h = make_random( hlen )
#h=[1]*2+[0]*3
if usereal:
sig=[c.real for c in sig]
h=[c.real for c in h]
# perform MAC filtering
yslow = slowfilter(sig,h)
#print '<YSLOW>',yslow,'</YSLOW>'
#yfast = fastfilter(sig,h,nfft)
yfast = utilfastfilter(sig,h,nfft,usereal)
#print yfast
print 'len(yslow)=%d'%len(yslow)
print 'len(yfast)=%d'%len(yfast)
diff = yslow-yfast
snr = 10*log10( abs( vdot(yslow,yslow) / vdot(diff,diff) ) )
print 'snr=%s' % snr
if snr < 10.0:
print 'h=',h
print 'sig=',sig[:5],'...'
print 'yslow=',yslow[:5],'...'
print 'yfast=',yfast[:5],'...'
def utilfastfilter(sig,h,nfft,usereal):
import compfft
import os
open( 'sig.dat','w').write( compfft.dopack(sig,'f',not usereal) )
open( 'h.dat','w').write( compfft.dopack(h,'f',not usereal) )
if usereal:
util = './fastconvr'
else:
util = './fastconv'
cmd = 'time %s -n %d -i sig.dat -h h.dat -o out.dat' % (util, nfft)
print cmd
ec = os.system(cmd)
print 'exited->',ec
return compfft.dounpack(open('out.dat').read(),'f',not usereal)
if __name__ == "__main__":
main()

196
thirdparty/kiss_fft130/test/fft.py vendored Normal file
View file

@ -0,0 +1,196 @@
#!/usr/bin/env python
import math
import sys
import random
pi=math.pi
e=math.e
j=complex(0,1)
def fft(f,inv):
n=len(f)
if n==1:
return f
for p in 2,3,5:
if n%p==0:
break
else:
raise Exception('%s not factorable ' % n)
m = n/p
Fout=[]
for q in range(p): # 0,1
fp = f[q::p] # every p'th time sample
Fp = fft( fp ,inv)
Fout.extend( Fp )
for u in range(m):
scratch = Fout[u::m] # u to end in strides of m
for q1 in range(p):
k = q1*m + u # indices to Fout above that became scratch
Fout[ k ] = scratch[0] # cuz e**0==1 in loop below
for q in range(1,p):
if inv:
t = e ** ( j*2*pi*k*q/n )
else:
t = e ** ( -j*2*pi*k*q/n )
Fout[ k ] += scratch[q] * t
return Fout
def rifft(F):
N = len(F) - 1
Z = [0] * (N)
for k in range(N):
Fek = ( F[k] + F[-k-1].conjugate() )
Fok = ( F[k] - F[-k-1].conjugate() ) * e ** (j*pi*k/N)
Z[k] = Fek + j*Fok
fp = fft(Z , 1)
f = []
for c in fp:
f.append(c.real)
f.append(c.imag)
return f
def real_fft( f,inv ):
if inv:
return rifft(f)
N = len(f) / 2
res = f[::2]
ims = f[1::2]
fp = [ complex(r,i) for r,i in zip(res,ims) ]
print 'fft input ', fp
Fp = fft( fp ,0 )
print 'fft output ', Fp
F = [ complex(0,0) ] * ( N+1 )
F[0] = complex( Fp[0].real + Fp[0].imag , 0 )
for k in range(1,N/2+1):
tw = e ** ( -j*pi*(.5+float(k)/N ) )
F1k = Fp[k] + Fp[N-k].conjugate()
F2k = Fp[k] - Fp[N-k].conjugate()
F2k *= tw
F[k] = ( F1k + F2k ) * .5
F[N-k] = ( F1k - F2k ).conjugate() * .5
#F[N-k] = ( F1kp + e ** ( -j*pi*(.5+float(N-k)/N ) ) * F2kp ) * .5
#F[N-k] = ( F1k.conjugate() - tw.conjugate() * F2k.conjugate() ) * .5
F[N] = complex( Fp[0].real - Fp[0].imag , 0 )
return F
def main():
#fft_func = fft
fft_func = real_fft
tvec = [0.309655,0.815653,0.768570,0.591841,0.404767,0.637617,0.007803,0.012665]
Ftvec = [ complex(r,i) for r,i in zip(
[3.548571,-0.378761,-0.061950,0.188537,-0.566981,0.188537,-0.061950,-0.378761],
[0.000000,-1.296198,-0.848764,0.225337,0.000000,-0.225337,0.848764,1.296198] ) ]
F = fft_func( tvec,0 )
nerrs= 0
for i in range(len(Ftvec)/2 + 1):
if abs( F[i] - Ftvec[i] )> 1e-5:
print 'F[%d]: %s != %s' % (i,F[i],Ftvec[i])
nerrs += 1
print '%d errors in forward fft' % nerrs
if nerrs:
return
trec = fft_func( F , 1 )
for i in range(len(trec) ):
trec[i] /= len(trec)
for i in range(len(tvec) ):
if abs( trec[i] - tvec[i] )> 1e-5:
print 't[%d]: %s != %s' % (i,tvec[i],trec[i])
nerrs += 1
print '%d errors in reverse fft' % nerrs
def make_random(dims=[1]):
import Numeric
res = []
for i in range(dims[0]):
if len(dims)==1:
r=random.uniform(-1,1)
i=random.uniform(-1,1)
res.append( complex(r,i) )
else:
res.append( make_random( dims[1:] ) )
return Numeric.array(res)
def flatten(x):
import Numeric
ntotal = Numeric.product(Numeric.shape(x))
return Numeric.reshape(x,(ntotal,))
def randmat( ndims ):
dims=[]
for i in range( ndims ):
curdim = int( random.uniform(2,4) )
dims.append( curdim )
return make_random(dims )
def test_fftnd(ndims=3):
import FFT
import Numeric
x=randmat( ndims )
print 'dimensions=%s' % str( Numeric.shape(x) )
#print 'x=%s' %str(x)
xver = FFT.fftnd(x)
x2=myfftnd(x)
err = xver - x2
errf = flatten(err)
xverf = flatten(xver)
errpow = Numeric.vdot(errf,errf)+1e-10
sigpow = Numeric.vdot(xverf,xverf)+1e-10
snr = 10*math.log10(abs(sigpow/errpow) )
if snr<80:
print xver
print x2
print 'SNR=%sdB' % str( snr )
def myfftnd(x):
import Numeric
xf = flatten(x)
Xf = fftndwork( xf , Numeric.shape(x) )
return Numeric.reshape(Xf,Numeric.shape(x) )
def fftndwork(x,dims):
import Numeric
dimprod=Numeric.product( dims )
for k in range( len(dims) ):
cur_dim=dims[ k ]
stride=dimprod/cur_dim
next_x = [complex(0,0)]*len(x)
for i in range(stride):
next_x[i*cur_dim:(i+1)*cur_dim] = fft(x[i:(i+cur_dim)*stride:stride],0)
x = next_x
return x
if __name__ == "__main__":
try:
nd = int(sys.argv[1])
except:
nd=None
if nd:
test_fftnd( nd )
else:
sys.exit(0)

117
thirdparty/kiss_fft130/test/mk_test.py vendored Normal file
View file

@ -0,0 +1,117 @@
#!/usr/bin/env python
import FFT
import sys
import random
import re
j=complex(0,1)
def randvec(n,iscomplex):
if iscomplex:
return [
int(random.uniform(-32768,32767) ) + j*int(random.uniform(-32768,32767) )
for i in range(n) ]
else:
return [ int(random.uniform(-32768,32767) ) for i in range(n) ]
def c_format(v,round=0):
if round:
return ','.join( [ '{%d,%d}' %(int(c.real),int(c.imag) ) for c in v ] )
else:
s= ','.join( [ '{%.60f ,%.60f }' %(c.real,c.imag) for c in v ] )
return re.sub(r'\.?0+ ',' ',s)
def test_cpx( n,inverse ,short):
v = randvec(n,1)
scale = 1
if short:
minsnr=30
else:
minsnr=100
if inverse:
tvecout = FFT.inverse_fft(v)
if short:
scale = 1
else:
scale = len(v)
else:
tvecout = FFT.fft(v)
if short:
scale = 1.0/len(v)
tvecout = [ c * scale for c in tvecout ]
s="""#define NFFT %d""" % len(v) + """
{
double snr;
kiss_fft_cpx test_vec_in[NFFT] = { """ + c_format(v) + """};
kiss_fft_cpx test_vec_out[NFFT] = {""" + c_format( tvecout ) + """};
kiss_fft_cpx testbuf[NFFT];
void * cfg = kiss_fft_alloc(NFFT,%d,0,0);""" % inverse + """
kiss_fft(cfg,test_vec_in,testbuf);
snr = snr_compare(test_vec_out,testbuf,NFFT);
printf("DATATYPE=" xstr(kiss_fft_scalar) ", FFT n=%d, inverse=%d, snr = %g dB\\n",NFFT,""" + str(inverse) + """,snr);
if (snr<""" + str(minsnr) + """)
exit_code++;
free(cfg);
}
#undef NFFT
"""
return s
def compare_func():
s="""
#define xstr(s) str(s)
#define str(s) #s
double snr_compare( kiss_fft_cpx * test_vec_out,kiss_fft_cpx * testbuf, int n)
{
int k;
double sigpow,noisepow,err,snr,scale=0;
kiss_fft_cpx err;
sigpow = noisepow = .000000000000000000000000000001;
for (k=0;k<n;++k) {
sigpow += test_vec_out[k].r * test_vec_out[k].r +
test_vec_out[k].i * test_vec_out[k].i;
C_SUB(err,test_vec_out[k],testbuf[k].r);
noisepow += err.r * err.r + err.i + err.i;
if (test_vec_out[k].r)
scale += testbuf[k].r / test_vec_out[k].r;
}
snr = 10*log10( sigpow / noisepow );
scale /= n;
if (snr<10)
printf( "\\npoor snr, try a scaling factor %f\\n" , scale );
return snr;
}
"""
return s
def main():
from getopt import getopt
opts,args = getopt(sys.argv[1:],'s')
opts = dict(opts)
short = int( opts.has_key('-s') )
fftsizes = args
if not fftsizes:
fftsizes = [ 1800 ]
print '#include "kiss_fft.h"'
print compare_func()
print "int main() { int exit_code=0;\n"
for n in fftsizes:
n = int(n)
print test_cpx(n,0,short)
print test_cpx(n,1,short)
print """
return exit_code;
}
"""
if __name__ == "__main__":
main()

49
thirdparty/kiss_fft130/test/pstats.c vendored Normal file
View file

@ -0,0 +1,49 @@
#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();
}

7
thirdparty/kiss_fft130/test/pstats.h vendored Normal file
View file

@ -0,0 +1,7 @@
#ifndef PSTATS_H
#define PSTATS_H
void pstats_init(void);
void pstats_report(void);
#endif

26
thirdparty/kiss_fft130/test/tailscrap.m vendored Normal file
View file

@ -0,0 +1,26 @@
function maxabsdiff=tailscrap()
% test code for circular convolution with the scrapped portion
% at the tail of the buffer, rather than the front
%
% The idea is to rotate the zero-padded h (impulse response) buffer
% to the left nh-1 samples, rotating the junk samples as well.
% This could be very handy in avoiding buffer copies during fast filtering.
nh=10;
nfft=256;
h=rand(1,nh);
x=rand(1,nfft);
hpad=[ h(nh) zeros(1,nfft-nh) h(1:nh-1) ];
% baseline comparison
y1 = filter(h,1,x);
y1_notrans = y1(nh:nfft);
% fast convolution
y2 = ifft( fft(hpad) .* fft(x) );
y2_notrans=y2(1:nfft-nh+1);
maxabsdiff = max(abs(y2_notrans - y1_notrans))
end

172
thirdparty/kiss_fft130/test/test_real.c vendored Normal file
View file

@ -0,0 +1,172 @@
#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;
}

74
thirdparty/kiss_fft130/test/test_vs_dft.c vendored Normal file
View file

@ -0,0 +1,74 @@
#include "kiss_fft.h"
void check(kiss_fft_cpx * in,kiss_fft_cpx * out,int nfft,int isinverse)
{
int bin,k;
double errpow=0,sigpow=0;
for (bin=0;bin<nfft;++bin) {
double ansr = 0;
double ansi = 0;
double difr;
double difi;
for (k=0;k<nfft;++k) {
double phase = -2*M_PI*bin*k/nfft;
double re = cos(phase);
double im = sin(phase);
if (isinverse)
im = -im;
#ifdef FIXED_POINT
re /= nfft;
im /= nfft;
#endif
ansr += in[k].r * re - in[k].i * im;
ansi += in[k].r * im + in[k].i * re;
}
difr = ansr - out[bin].r;
difi = ansi - out[bin].i;
errpow += difr*difr + difi*difi;
sigpow += ansr*ansr+ansi*ansi;
}
printf("nfft=%d inverse=%d,snr = %f\n",nfft,isinverse,10*log10(sigpow/errpow) );
}
void test1d(int nfft,int isinverse)
{
size_t buflen = sizeof(kiss_fft_cpx)*nfft;
kiss_fft_cpx * in = (kiss_fft_cpx*)malloc(buflen);
kiss_fft_cpx * out= (kiss_fft_cpx*)malloc(buflen);
kiss_fft_cfg cfg = kiss_fft_alloc(nfft,isinverse,0,0);
int k;
for (k=0;k<nfft;++k) {
in[k].r = (rand() % 65536) - 32768;
in[k].i = (rand() % 65536) - 32768;
}
kiss_fft(cfg,in,out);
check(in,out,nfft,isinverse);
free(in);
free(out);
free(cfg);
}
int main(int argc,char ** argv)
{
if (argc>1) {
int k;
for (k=1;k<argc;++k) {
test1d(atoi(argv[k]),0);
test1d(atoi(argv[k]),1);
}
}else{
test1d(32,0);
test1d(32,1);
}
return 0;
}

73
thirdparty/kiss_fft130/test/testcpp.cc vendored Normal file
View file

@ -0,0 +1,73 @@
#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;
}

162
thirdparty/kiss_fft130/test/testkiss.py vendored Normal file
View file

@ -0,0 +1,162 @@
#!/usr/bin/env python
import math
import sys
import os
import random
import struct
import popen2
import getopt
import numpy
pi=math.pi
e=math.e
j=complex(0,1)
doreal=0
datatype = os.environ.get('DATATYPE','float')
util = '../tools/fft_' + datatype
minsnr=90
if datatype == 'double':
fmt='d'
elif datatype=='int16_t':
fmt='h'
minsnr=10
elif datatype=='int32_t':
fmt='i'
elif datatype=='simd':
fmt='4f'
sys.stderr.write('testkiss.py does not yet test simd')
sys.exit(0)
elif datatype=='float':
fmt='f'
else:
sys.stderr.write('unrecognized datatype %s\n' % datatype)
sys.exit(1)
def dopack(x,cpx=1):
x = numpy.reshape( x, ( numpy.size(x),) )
if cpx:
s = ''.join( [ struct.pack(fmt*2,c.real,c.imag) for c in x ] )
else:
s = ''.join( [ struct.pack(fmt,c.real) for c in x ] )
return s
def dounpack(x,cpx):
uf = fmt * ( len(x) / struct.calcsize(fmt) )
s = struct.unpack(uf,x)
if cpx:
return numpy.array(s[::2]) + numpy.array( s[1::2] )*j
else:
return numpy.array(s )
def make_random(dims=[1]):
res = []
for i in range(dims[0]):
if len(dims)==1:
r=random.uniform(-1,1)
if doreal:
res.append( r )
else:
i=random.uniform(-1,1)
res.append( complex(r,i) )
else:
res.append( make_random( dims[1:] ) )
return numpy.array(res)
def flatten(x):
ntotal = numpy.size(x)
return numpy.reshape(x,(ntotal,))
def randmat( ndims ):
dims=[]
for i in range( ndims ):
curdim = int( random.uniform(2,5) )
if doreal and i==(ndims-1):
curdim = int(curdim/2)*2 # force even last dimension if real
dims.append( curdim )
return make_random(dims )
def test_fft(ndims):
x=randmat( ndims )
if doreal:
xver = numpy.fft.rfftn(x)
else:
xver = numpy.fft.fftn(x)
open('/tmp/fftexp.dat','w').write(dopack( flatten(xver) , True ) )
x2=dofft(x,doreal)
err = xver - x2
errf = flatten(err)
xverf = flatten(xver)
errpow = numpy.vdot(errf,errf)+1e-10
sigpow = numpy.vdot(xverf,xverf)+1e-10
snr = 10*math.log10(abs(sigpow/errpow) )
print 'SNR (compared to NumPy) : %.1fdB' % float(snr)
if snr<minsnr:
print 'xver=',xver
print 'x2=',x2
print 'err',err
sys.exit(1)
def dofft(x,isreal):
dims=list( numpy.shape(x) )
x = flatten(x)
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='%s -n ' % util
cmd += ','.join([str(d) for d in dims])
if doreal:
cmd += ' -R '
print cmd
p = popen2.Popen3(cmd )
open('/tmp/fftin.dat','w').write(dopack( x , isreal==False ) )
p.tochild.write( dopack( x , isreal==False ) )
p.tochild.close()
res = dounpack( p.fromchild.read() , 1 )
open('/tmp/fftout.dat','w').write(dopack( flatten(res) , True ) )
if doreal:
dims[-1] = int( dims[-1]/2 ) + 1
res = scale * res
p.wait()
return numpy.reshape(res,dims)
def main():
opts,args = getopt.getopt(sys.argv[1:],'r')
opts=dict(opts)
global doreal
doreal = opts.has_key('-r')
if doreal:
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()

94
thirdparty/kiss_fft130/test/twotonetest.c vendored Normal file
View file

@ -0,0 +1,94 @@
#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;
}

62
thirdparty/kiss_fft130/tools/Makefile vendored Normal file
View file

@ -0,0 +1,62 @@
WARNINGS=-W -Wall -Wstrict-prototypes -Wmissing-prototypes -Waggregate-return \
-Wcast-align -Wcast-qual -Wnested-externs -Wshadow -Wbad-function-cast \
-Wwrite-strings
ifeq "$(DATATYPE)" ""
DATATYPE=float
endif
ifeq "$(DATATYPE)" "int32_t"
TYPEFLAGS=-DFIXED_POINT=32
endif
ifeq "$(DATATYPE)" "int16_t"
TYPEFLAGS=-DFIXED_POINT=16
endif
ifeq "$(DATATYPE)" "simd"
TYPEFLAGS=-DUSE_SIMD=1 -msse
endif
ifeq "$(TYPEFLAGS)" ""
TYPEFLAGS=-Dkiss_fft_scalar=$(DATATYPE)
endif
ifneq ("$(KISS_FFT_USE_ALLOCA)","")
CFLAGS+= -DKISS_FFT_USE_ALLOCA=1
endif
CFLAGS+= $(CFLAGADD)
FFTUTIL=fft_$(DATATYPE)
FASTFILT=fastconv_$(DATATYPE)
FASTFILTREAL=fastconvr_$(DATATYPE)
PSDPNG=psdpng_$(DATATYPE)
DUMPHDR=dumphdr_$(DATATYPE)
all: $(FFTUTIL) $(FASTFILT) $(FASTFILTREAL)
# $(PSDPNG)
# $(DUMPHDR)
#CFLAGS=-Wall -O3 -pedantic -march=pentiumpro -ffast-math -fomit-frame-pointer $(WARNINGS)
# If the above flags do not work, try the following
CFLAGS=-Wall -O3 $(WARNINGS)
# tip: try -openmp or -fopenmp to use multiple cores
$(FASTFILTREAL): ../kiss_fft.c kiss_fastfir.c kiss_fftr.c
$(CC) -o $@ $(CFLAGS) -I.. $(TYPEFLAGS) -DREAL_FASTFIR $+ -DFAST_FILT_UTIL -lm
$(FASTFILT): ../kiss_fft.c kiss_fastfir.c
$(CC) -o $@ $(CFLAGS) -I.. $(TYPEFLAGS) $+ -DFAST_FILT_UTIL -lm
$(FFTUTIL): ../kiss_fft.c fftutil.c kiss_fftnd.c kiss_fftr.c kiss_fftndr.c
$(CC) -o $@ $(CFLAGS) -I.. $(TYPEFLAGS) $+ -lm
$(PSDPNG): ../kiss_fft.c psdpng.c kiss_fftr.c
$(CC) -o $@ $(CFLAGS) -I.. $(TYPEFLAGS) $+ -lpng -lm
$(DUMPHDR): ../kiss_fft.c dumphdr.c
$(CC) -o $@ $(CFLAGS) -I.. $(TYPEFLAGS) $+ -lm
clean:
rm -f *~ fft fft_* fastconv fastconv_* fastconvr fastconvr_* psdpng psdpng_*

208
thirdparty/kiss_fft130/tools/fftutil.c vendored Normal file
View file

@ -0,0 +1,208 @@
/*
Copyright (c) 2003-2004, Mark Borgerding
All rights reserved.
Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
* 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.
* 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.
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.
*/
#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;
}

116
thirdparty/kiss_fft130/tools/kfc.c vendored Normal file
View file

@ -0,0 +1,116 @@
#include "kfc.h"
/*
Copyright (c) 2003-2004, Mark Borgerding
All rights reserved.
Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
* 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.
* 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.
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.
*/
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

46
thirdparty/kiss_fft130/tools/kfc.h vendored Normal file
View file

@ -0,0 +1,46 @@
#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 kfc_fft(int nfft, const kiss_fft_cpx * fin,kiss_fft_cpx * fout);
/*reverse complex FFT */
void kfc_ifft(int nfft, const kiss_fft_cpx * fin,kiss_fft_cpx * fout);
/*free all cached objects*/
void kfc_cleanup(void);
#ifdef __cplusplus
}
#endif
#endif

470
thirdparty/kiss_fft130/tools/kiss_fastfir.c vendored Normal file
View file

@ -0,0 +1,470 @@
/*
Copyright (c) 2003-2004, Mark Borgerding
All rights reserved.
Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
* 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.
* 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.
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.
*/
#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

193
thirdparty/kiss_fft130/tools/kiss_fftnd.c vendored Normal file
View file

@ -0,0 +1,193 @@
/*
Copyright (c) 2003-2004, Mark Borgerding
All rights reserved.
Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
* 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.
* 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.
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.
*/
#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_fftnd_cfg st = NULL;
int i;
int dimprod=1;
size_t memneeded = sizeof(struct kiss_fftnd_state);
char * ptr;
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 += sizeof(int) * ndims;/* st->dims */
memneeded += sizeof(void*) * ndims;/* st->states */
memneeded += sizeof(kiss_fft_cpx) * dimprod; /* st->tmpbuf */
if (lenmem == NULL) {/* allocate for the caller*/
st = (kiss_fftnd_cfg) malloc (memneeded);
} else { /* initialize supplied buffer if big enough */
if (*lenmem >= memneeded)
st = (kiss_fftnd_cfg) mem;
*lenmem = memneeded; /*tell caller how big struct is (or would be) */
}
if (!st)
return NULL; /*malloc failed or buffer too small */
st->dimprod = dimprod;
st->ndims = ndims;
ptr=(char*)(st+1);
st->states = (kiss_fft_cfg *)ptr;
ptr += sizeof(void*) * ndims;
st->dims = (int*)ptr;
ptr += sizeof(int) * ndims;
st->tmpbuf = (kiss_fft_cpx*)ptr;
ptr += 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 effect,
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;
}
}
}

18
thirdparty/kiss_fft130/tools/kiss_fftnd.h vendored Normal file
View file

@ -0,0 +1,18 @@
#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_fftnd_alloc(const int *dims,int ndims,int inverse_fft,void*mem,size_t*lenmem);
void kiss_fftnd(kiss_fftnd_cfg cfg,const kiss_fft_cpx *fin,kiss_fft_cpx *fout);
#ifdef __cplusplus
}
#endif
#endif

118
thirdparty/kiss_fft130/tools/kiss_fftndr.c vendored Normal file
View file

@ -0,0 +1,118 @@
/*
Copyright (c) 2003-2004, Mark Borgerding
All rights reserved.
Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
* 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.
* 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.
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.
*/
#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_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;
(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 = sizeof( struct kiss_fftndr_state ) + nr + nd + ntmp;
if (lenmem==NULL) {
st = (kiss_fftndr_cfg) malloc(memneeded);
}else{
if (*lenmem >= memneeded)
st = (kiss_fftndr_cfg)mem;
*lenmem = memneeded;
}
if (st==NULL)
return NULL;
memset( st , 0 , memneeded);
st->dimReal = dimReal;
st->dimOther = dimOther;
st->cfg_r = kiss_fftr_alloc( dimReal,inverse_fft,st+1,&nr);
st->cfg_nd = kiss_fftnd_alloc(dims,ndims-1,inverse_fft, ((char*) st->cfg_r)+nr,&nd);
st->tmpbuf = (char*)st->cfg_nd + nd;
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);
}
}

47
thirdparty/kiss_fft130/tools/kiss_fftndr.h vendored Normal file
View file

@ -0,0 +1,47 @@
#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_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_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_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_fftr_free free
#ifdef __cplusplus
}
#endif
#endif

159
thirdparty/kiss_fft130/tools/kiss_fftr.c vendored Normal file
View file

@ -0,0 +1,159 @@
/*
Copyright (c) 2003-2004, Mark Borgerding
All rights reserved.
Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
* 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.
* 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.
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.
*/
#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)
{
int i;
kiss_fftr_cfg st = NULL;
size_t subsize, memneeded;
if (nfft & 1) {
fprintf(stderr,"Real FFT optimization must be even.\n");
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) {
fprintf(stderr,"kiss fft usage error: improper alloc\n");
exit(1);
}
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) {
fprintf (stderr, "kiss fft usage error: improper alloc\n");
exit (1);
}
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);
}

46
thirdparty/kiss_fft130/tools/kiss_fftr.h vendored Normal file
View file

@ -0,0 +1,46 @@
#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_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_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_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 free
#ifdef __cplusplus
}
#endif
#endif

235
thirdparty/kiss_fft130/tools/psdpng.c vendored Normal file
View file

@ -0,0 +1,235 @@
/*
Copyright (c) 2003-2004, Mark Borgerding
All rights reserved.
Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
* 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.
* 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.
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.
*/
#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"
);
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*)malloc(sizeof(float)*nfreqs ) );
memset(mag2buf,0,sizeof(mag2buf)*nfreqs);
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;
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 );
row_data = (rgb_t*)malloc(sizeof(rgb_t) * nrows * nfreqs) ;
cpx2pixels(row_data, vals, nfreqs*nrows );
row_pointers = realloc(row_pointers, 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);
}
int main(int argc,char ** argv)
{
config(argc,argv);
transform_signal();
make_png();
if (fout!=stdout) fclose(fout);
if (fin!=stdin) fclose(fin);
return 0;
}

2
toonz/sources/include/toonzqt/fxiconmanager.h
View file

@ -130,6 +130,8 @@ const struct {
{"STD_iwa_MotionBlurFx", "fx_iwa_motionblur"},
{"STD_iwa_SpectrumFx", "fx_iwa_spectrum"},
{"STD_iwa_PerspectiveDistortFx", "fx_iwa_perspective_distort"},
{"STD_iwa_BokehFx", "fx_iwa_bokeh"},
{"STD_iwa_SoapBubbleFx", "fx_iwa_soapbubble"},
{0, 0}};
};

8
toonz/sources/stdfx/CMakeLists.txt
View file

@ -72,6 +72,7 @@ set(HEADERS
iwa_fresnel.h
iwa_pnperspectivefx.h
iwa_soapbubblefx.h
iwa_bokehfx.h
)
set(SOURCES
@ -246,6 +247,9 @@ set(SOURCES
iwa_noise1234.cpp
iwa_pnperspectivefx.cpp
iwa_soapbubblefx.cpp
${SDKROOT}/kiss_fft130/kiss_fft.c
${SDKROOT}/kiss_fft130/tools/kiss_fftnd.c
iwa_bokehfx.cpp
)
set(OBJCSOURCES
@ -276,6 +280,10 @@ else()
)
endif()
include_directories(
${SDKROOT}/kiss_fft130
)
_find_toonz_library(TNZLIBS "tnzcore;tnzbase;toonzlib")
if(APPLE)

1042
toonz/sources/stdfx/iwa_bokehfx.cpp Normal file
View file

File diff suppressed because it is too large Load diff

149
toonz/sources/stdfx/iwa_bokehfx.h Normal file
View file

@ -0,0 +1,149 @@
#pragma once
/*------------------------------------
Iwa_BokehFx
Apply an off-focus effect to the source image, using user input iris image.
It considers characteristics of films (which is known as Hurter–Driffield
curves)
or human eye's perception (which is known as Weber–Fechner law).
For filtering process I used KissFFT, an FFT library by Mark Borgerding,
distributed with a 3-clause BSD-style license.
------------------------------------*/
#ifndef IWA_BOKEHFX_H
#define IWA_BOKEHFX_H
#include "stdfx.h"
#include "tfxparam.h"
#include "traster.h"
#include <QList>
#include <QThread>
#include "tools/kiss_fftnd.h"
const int LAYER_NUM = 5;
struct double2 {
double x, y;
};
struct int2 {
int x, y;
};
class MyThread : public QThread {
public:
enum Channel { Red = 0, Green, Blue };
private:
Channel m_channel;
volatile bool m_finished;
TRasterP m_layerTileRas;
TRasterP m_outTileRas;
TRasterP m_tmpAlphaRas;
kiss_fft_cpx *m_kissfft_comp_iris;
float m_filmGamma; // keep the film gamma in each thread as it is refered so
// often
TRasterGR8P m_kissfft_comp_in_ras, m_kissfft_comp_out_ras;
kiss_fft_cpx *m_kissfft_comp_in, *m_kissfft_comp_out;
kiss_fftnd_cfg m_kissfft_plan_fwd, m_kissfft_plan_bkwd;
bool m_isTerminated;
// not used for now
bool m_doLightenComp;
public:
MyThread(Channel channel, TRasterP layerTileRas, TRasterP outTileRas,
TRasterP tmpAlphaRas, kiss_fft_cpx *kissfft_comp_iris,
float m_filmGamma,
bool doLightenComp = false); // not used for now
// Convert the pixels from RGB values to exposures and multiply it by alpha
// channel value.
// Store the results in the real part of kiss_fft_cpx.
template <typename RASTER, typename PIXEL>
void setLayerRaster(const RASTER srcRas, kiss_fft_cpx *dstMem,
TDimensionI dim);
// Composite the bokeh layer to the result
template <typename RASTER, typename PIXEL, typename A_RASTER,
typename A_PIXEL>
void compositLayerToTile(const RASTER layerRas, const RASTER outTileRas,
const A_RASTER alphaRas, TDimensionI dim,
int2 margin);
void run();
bool isFinished() { return m_finished; }
// RGB value <--> Exposure
float valueToExposure(float value);
float exposureToValue(float exposure);
// memory allocation
bool init();
void terminateThread() { m_isTerminated = true; }
bool checkTerminationAndCleanupThread();
};
class Iwa_BokehFx : public TStandardRasterFx {
FX_PLUGIN_DECLARATION(Iwa_BokehFx)
protected:
TRasterFxPort m_iris;
TDoubleParamP m_onFocusDistance; // Focus Distance (0-1)
TDoubleParamP m_bokehAmount; // The maximum bokeh size. The size of bokeh at
// the layer separated by 1.0 from the focal
// position
TDoubleParamP m_hardness; // Film gamma
struct LAYERPARAM {
TRasterFxPort m_source;
TBoolParamP m_premultiply;
TDoubleParamP m_distance; // The layer distance from the camera (0-1)
TDoubleParamP m_bokehAdjustment; // Factor for adjusting distance (= focal
// distance - layer distance) (0-2.0)
} m_layerParams[LAYER_NUM];
// Sort source layers by distance
QList<int> getSortedSourceIndices(double frame);
// Get the pixel size of bokehAmount ( referenced ino_blur.cpp )
float getBokehPixelAmount(const double frame, const TAffine affine);
// Compute the bokeh size for each layer. The source tile will be enlarged by
// the largest size of them.
QVector<float> getIrisSizes(const double frame,
const QList<int> sourceIndices,
const float bokehPixelAmount, float &maxIrisSize);
//"Over" composite the layer to the output raster.
void compositLayerAsIs(TTile &tile, TTile &layerTile, const double frame,
const TRenderSettings &settings, const int index);
// Resize / flip the iris image according to the size ratio.
// Normalize the brightness of the iris image.
// Enlarge the iris to the output size.
void convertIris(const float irisSize, kiss_fft_cpx *kissfft_comp_iris_before,
const TDimensionI &dimOut, const TRectD &irisBBox,
const TTile &irisTile);
// Do FFT the alpha channel.
// Forward FFT -> Multiply by the iris data -> Backward FFT
void calcAlfaChannelBokeh(kiss_fft_cpx *kissfft_comp_iris, TTile &layerTile,
TRasterP tmpAlphaRas);
public:
Iwa_BokehFx();
void doCompute(TTile &tile, double frame, const TRenderSettings &settings);
bool doGetBBox(double frame, TRectD &bBox, const TRenderSettings &info);
bool canHandle(const TRenderSettings &info, double frame);
};
#endif

BIN
toonz/sources/toonzqt/Resources/fxicons/fx_iwa_bokeh.png Normal file
View file

Binary file not shown.
Side by side

After

Width:  |  Height:  |  Size: 782 B

BIN
toonz/sources/toonzqt/Resources/fxicons/fx_iwa_soapbubble.png Normal file
View file

Binary file not shown.
Side by side

After

Width:  |  Height:  |  Size: 15 KiB

2
toonz/sources/toonzqt/toonzqt.qrc
View file

@ -296,6 +296,8 @@
<file>Resources/fxicons/fx_iwa_spectrum.png</file>
<file>Resources/fxicons/fx_iwa_perspective_distort.png</file>
<file>Resources/fxicons/fx_iwa_pn_perspective.png</file>
<file>Resources/fxicons/fx_iwa_bokeh.png</file>
<file>Resources/fxicons/fx_iwa_soapbubble.png</file>
<file>Resources/fxicons/fx_unidentified.png</file>
<file>Resources/keyframe_noanim.png</file>
<file>Resources/keyframe_key.png</file>