Sponsored Content
Full Discussion: Slow FFT in ksh93 and awk.
Operating Systems OS X (Apple) Slow FFT in ksh93 and awk. Post 303023145 by wisecracker on Wednesday 12th of September 2018 12:11:57 PM
Old 09-12-2018
Hi all...

Well guys, it took a while because of the serious limitations of ARexx for a stock AMIGA A1200 but here is a working FFT for it. ARexx has no ARRAY facility but there is a workaround. ;o)
Boy oh boy, ARexx's arithmetic is so convoluted and floating point precision is not the best.
This was a challenge. I could have generated SIN, COS and SQRT in pure ARexx but decided to use an external ARexx math(s) library instead...
It would be just as easy to create SIN, COS and SQRT in ksh too eliminating awk entirely but awk just made it easier...
Just for the record the AMIGA shell/terminal has a large subset of xterm's escape codes.
An exercise in futility? Maybe but serious fun nevertheless...
This is through FS-UAE AMIGA emulation and my real A1200 setup for easy transfer from this Macbook Pro.
ARexx code:
Code:
/* Simple_FFT_DEMO.rexx */

/* This ARexx version needs this dependency: */
/* http://aminet.net/package/util/rexx/RexxMathLib */
CALL ADDLIB('rexxmathlib.library',0,-30,0)

NUMERIC DIGITS 14

SAY ""
SAY "An experimental method to do an FFT in ARexx using the standard test values."
SAY "It requires rexxmathlib.library for the SIN and COS."
SAY "$VER Simple_FFT_DEMO.rexx_(C)06-09-2018_B.Walker_issued_under_GPL2."
SAY ""

/* Create standard test _ARRAYS_ REAL and IMAGINARY. */
/* These MUST be powers of 2 and greater than 2 in size. */
/* Real values, '1.0 1.0 1.0 1.0 0.0 0.0 0.0 0.0' */
/* Imag values, '0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0' */
/* Not really ARRAYS but good enough for this DEMO! */
DO N=0 TO 3 BY 1
	REAL_ARRAY.N=1.0
END
DO N=4 TO 7 BY 1
	REAL_ARRAY.N=0.0
END
DO N=0 TO 7 BY 1
	IMAG_ARRAY.N=0.0
END

/* Display them. */
SAY "Input, REAL: IMAG:"
DO N=0 TO 7 BY 1
	SAY "        "||REAL_ARRAY.N||"   "||IMAG_ARRAY.N
END

/* Only constant required, 14 decimal places. */
PI=3.14159265358979 
/* N should be 8 by default in this DEMO. */
/* SAY "Number of elements = "||N||"..." */
N=N-1

CALL FFT

SAY ""
DO N=0 TO 7 BY 1
	SAY "REAL: "||+OUT_REAL.N||",    IMAG: "||+OUT_IMAG.N
END

SAY ""
SAY "Final FFT values to 5 decimal places:"
STR=""
DO N=0 TO 7 BY 1
	CALL ABS_COMPLEX
	NUMERIC DIGITS 14
	IF +ABS <= 0.000000000001
	THEN
		ABS=0
	ENDIF
	NUMERIC DIGITS 6
	STR=STR||+ABS||" "
END
SAY STR
SAY ""
EXIT

/* Create the subroutines required. */
/* FFT subroutine. */
FFT:
DO K=0 TO N BY 1
	SUMREAL=0.0
	SUMIMAG=0.0
	DO T=0 TO N BY 1
		ANGLE=(2.0*PI*T*K)/(N+1)
		SUMREAL=SUMREAL+(REAL_ARRAY.T*COS(ANGLE))+(IMAG_ARRAY.T*SIN(ANGLE))
		SUMIMAG=SUMIMAG-(REAL_ARRAY.T*SIN(ANGLE))+(IMAG_ARRAY.T*COS(ANGLE))
	END
	OUT_REAL.K=SUMREAL
	OUT_IMAG.K=SUMIMAG
END
RETURN
/* FFT subroutine end. */

/* Absolute value of complex number subroutine. */
ABS_COMPLEX:
	ABS=SQRT( (OUT_REAL.N**2)+(OUT_IMAG.N**2) )
RETURN
/* Absolute value of complex number subroutine end. */

 

10 More Discussions You Might Find Interesting

1. UNIX for Dummies Questions & Answers

echo is too slow. HELP with Awk

Hello All, Below is a simple script i worte to find the 208th char in a file. If the char = "C" then I re-direct the line to a file called change.txt. If it is not "C" then I re-direct it to a file called delete.txt. My problem is I have a file 0f 500K lines. this script is very slow. I am... (4 Replies)
Discussion started by: eja
4 Replies

2. Shell Programming and Scripting

ksh93 deprecation...

Any means of running ksh93 in a ksh88-mode? Might sound odd, but I want/need to restrict U/Win-developed scripts to correspond to the ksh88 version on my Solaris environment(s). Thanks. (2 Replies)
Discussion started by: curleb
2 Replies

3. Shell Programming and Scripting

ksh88 or ksh93

Hi all! Does anybody know how can I check if any UNIX installation has implemented ksh88 or ksh93? Thanks in advance. Néstor. (3 Replies)
Discussion started by: Nestor
3 Replies

4. Shell Programming and Scripting

The builtin split function in AWK is too slow

I have a text file that contains 4 million lines, each line contains 2 fields(colon as field separator). as shown: 123:444,555,666,777,888,345 233:5444,555,666,777,888,345 623:454,585,664,773,888,345 ...... Here I have to split the second field(can be up to 40,000 fields) by comma into an... (14 Replies)
Discussion started by: kevintse
14 Replies

5. Shell Programming and Scripting

cut, sed, awk too slow to retrieve line - other options?

Hi, I have a script that, basically, has two input files of this type: file1 key1=value1_1_1 key2=value1_2_1 key4=value1_4_1 ... file2 key2=value2_2_1 key2=value2_2_2 key3=value2_3_1 key4=value2_4_1 ... My files are 10k lines big each (approx). The keys are strings that don't... (7 Replies)
Discussion started by: fzd
7 Replies

6. Shell Programming and Scripting

Making a faster alternative to a slow awk command

Hi, I have a large number of input files with two columns of numbers. For example: 83 1453 99 3255 99 8482 99 7372 83 175 I only wish to retain lines where the numbers fullfil two requirements. E.g: =83 1000<=<=2000 To do this I use the following... (10 Replies)
Discussion started by: s052866
10 Replies

7. UNIX for Advanced & Expert Users

Ksh93 on Linux compatible with ksh93 on AIX

Hi Experts, I have several shell scripts that have been developed on a Linux box for korn ksh93. If we want to run this software on an AIX 6.1 box that runs ksh88 by default can we just change the she-bang line to reference /bin/ksh93 which ships with AIX as its "enhanced shell" to ensure... (6 Replies)
Discussion started by: Keith Turley
6 Replies

8. OS X (Apple)

FFT for the AMIGA through ksh88 shell.

I don't know if anyone is interested but I have been meddling with FFT for the AMIGA. (Sadly we AMIGAns don't have these luxuries through any scripting language. Below is a Python snippet that uses the builtin 'cmath' module to work with the lowly Python 2.0.1 for the AMIGA. It is part of a... (0 Replies)
Discussion started by: wisecracker
0 Replies

9. AIX

Ksh93/AIX compatibility

Hi everyone ! Im trying to know from wich version of AIX KSH93 is available ? Internet tell me 6.x and 7.x AIX are available, bue what about 5.x ? Is KSH93 available on AIX 5.x ? Is it the same way to manipulate variables as KSH93 on 7.x ? Thanks for your support and have a nice day ! (2 Replies)
Discussion started by: majinfrede
2 Replies

10. OS X (Apple)

FFT for Python 2.0.x to 3.7.0.

Hi guys... This is code that was originally designed to work on an upgraded AMIGA A1200 using Python 2.0.x. Unfortunately it broke inside much later versions, NOT because of the print statement/function but other minor subtleties. So this is the final result tested on various machines including... (0 Replies)
Discussion started by: wisecracker
0 Replies
FFTW(3pm)						User Contributed Perl Documentation						 FFTW(3pm)

NAME
PDL::FFTW - PDL interface to the Fastest Fourier Transform in the West v2.x DESCRIPTION
This is a means to interface PDL with the FFTW library. It's similar to the standard FFT routine but it's usually faster and has support for real transforms. It works well for the types of PDLs for which was the library was compiled (otherwise it must do conversions). SYNOPSIS
use PDL::FFTW load_wisdom("file_name"); out_cplx_pdl = fftw(in_cplx_pdl); out_cplx_pdl = ifftw(in_cplx_pdl); out_cplx_pdl = rfftw(in_real_pdl); out_real_pdl = irfftw(in_cplx_pdl); cplx_pdl = nfftw(cplx_pdl); cplx_pdl = infftw(cplx_pdl); cplx_pdl = Cmul(a_cplx_pdl, b_cplx_pdl); cplx_pdl = Cconj(a_cplx_pdl); real_pdl = Cmod(a_cplx_pdl); real_pdl = Cmod2(a_cplx_pdl); FFTW documentation Please refer to the FFTW documentation for a better understanding of these functions. Note that complex numbers are represented as piddles with leading dimension size 2 (real/imaginary pairs). FUNCTIONS
load_wisdom Loads the wisdom from a file for better FFTW performance. The wisdom is automatically saved when the program ends. It will be automagically called when the variable $PDL::FFT::wisdom is set to a file name. For example, the following is a useful idiom to have in your .perldlrc file: $PDL::FFT::wisdom = "$ENV{HOME}/.fftwisdom"; # save fftw wisdom in this file Explicit usage: load_wisdom($fname); fftw calculate the complex FFT of a real piddle (complex input, complex output) $pdl_cplx = fftw $pdl_cplx; ifftw Complex inverse FFT (complex input, complex output). $pdl_cplx = ifftw $pdl_cplx; nfftw Complex inplace FFT (complex input, complex output). $pdl_cplx = nfftw $pdl_cplx; infftw Complex inplace inverse FFT (complex input, complex output). $pdl_cplx = infftw $pdl_cplx; rfftw Real FFT. For an input piddle of dimensions [n1,n2,...] the output is [2,(n1/2)+1,n2,...] (real input, complex output). $pdl_cplx = fftw $pdl_real; irfftw Real inverse FFT. Have a look at rfftw to understand the format. USE ONLY an even n1! (complex input, real output) $pdl_real = ifftw $pdl_cplx; nrfftw Real inplace FFT. If you want a transformation on a piddle with dimensions [n1,n2,....] you MUST pass in a piddle with dimensions [2*(n1/2+1),n2,...] (real input, complex output). Use with care due to dimension restrictions mentioned below. For details check the html docs that come with the fftw library. $pdl_cplx = nrfftw $pdl_real; inrfftw Real inplace inverse FFT. Have a look at nrfftw to understand the format. USE ONLY an even first dimension size! (complex input, real output) $pdl_real = infftw $pdl_cplx; rfftwconv ND convolution using real ffts from the FFTW library $conv = rfftwconv $im, kernctr $im, $k; # kernctr is from PDL::FFT fftwconv ND convolution using complex ffts from the FFTW library Assumes real input! $conv = fftwconv $im, kernctr $im, $k; # kernctr is from PDL::FFT Cmul Signature: (a(n); b(n); [o]c(n)) Cmul Complex multiplication $out_pdl_cplx = Cmul($a_pdl_cplx,$b_pdl_cplx); Cmul does not process bad values. It will set the bad-value flag of all output piddles if the flag is set for any of the input piddles. Cscale Signature: (a(n); b(); [o]c(n)) Cscale Complex by real multiplation. $out_pdl_cplx = Cscale($a_pdl_cplx,$b_pdl_real); Cscale does not process bad values. It will set the bad-value flag of all output piddles if the flag is set for any of the input piddles. Cdiv Signature: (a(n); b(n); [o]c(n)) Cdiv Complex division. $out_pdl_cplx = Cdiv($a_pdl_cplx,$b_pdl_cplx); Cdiv does not process bad values. It will set the bad-value flag of all output piddles if the flag is set for any of the input piddles. Cbmul Signature: (a(n); b(n)) Cbmul Complex inplace multiplication. Cbmul($a_pdl_cplx,$b_pdl_cplx); Cbmul does not process bad values. It will set the bad-value flag of all output piddles if the flag is set for any of the input piddles. Cbscale Signature: (a(n); b()) Cbscale Complex inplace multiplaction by real. Cbscale($a_pdl_cplx,$b_pdl_real); Cbscale does not process bad values. It will set the bad-value flag of all output piddles if the flag is set for any of the input piddles. Cconj Signature: (a(n); [o]c(n)) Cconj Complex conjugate. $out_pdl_cplx = Cconj($a_pdl_cplx); Cconj does not process bad values. It will set the bad-value flag of all output piddles if the flag is set for any of the input piddles. Cbconj Signature: (a(n)) Cbconj Complex inplace conjugate. Cbconj($a_pdl_cplx); Cbconj does not process bad values. It will set the bad-value flag of all output piddles if the flag is set for any of the input piddles. Cexp Signature: (a(n); [o]c(n)) Cexp Complex exponentation. $out_pdl_cplx = Cexp($a_pdl_cplx); Cexp does not process bad values. It will set the bad-value flag of all output piddles if the flag is set for any of the input piddles. Cbexp Signature: (a(n)) Cbexp Complex inplace exponentation. $out_pdl_cplx = Cbexp($a_pdl_cplx); Cbexp does not process bad values. It will set the bad-value flag of all output piddles if the flag is set for any of the input piddles. Cmod Signature: (a(n); [o]c()) Cmod modulus of a complex piddle. $out_pdl_real = Cmod($a_pdl_cplx); Cmod does not process bad values. It will set the bad-value flag of all output piddles if the flag is set for any of the input piddles. Carg Signature: (a(n); [o]c()) Carg argument of a complex number. $out_pdl_real = Carg($a_pdl_cplx); Carg does not process bad values. It will set the bad-value flag of all output piddles if the flag is set for any of the input piddles. Cmod2 Signature: (a(n); [o]c()) Cmod2 Returns squared modulus of a complex number. $out_pdl_real = Cmod2($a_pdl_cplx); Cmod2 does not process bad values. It will set the bad-value flag of all output piddles if the flag is set for any of the input piddles. AUTHOR
Copyright (C) 1999 Christian Pellegrin, 2000 Christian Soeller. All rights reserved. There is no warranty. You are allowed to redistribute this software / documentation under certain conditions. For details, see the file COPYING in the PDL distribution. If this file is separated from the PDL distribution, the copyright notice should be included in the file. perl v5.14.2 2012-05-30 FFTW(3pm)
All times are GMT -4. The time now is 09:19 AM.
Unix & Linux Forums Content Copyright 1993-2022. All Rights Reserved.
Privacy Policy