From bug-octave-request at bevo dot che dot wisc dot edu Tue Jan 22 13:40:27 2002 Subject: Re: IMPORTANT: Bug in "arg" function From: Paul Kienzle To: "Lippert, Ross A." , "'Paul Kienzle'" , robher@adinet.com.uy, bug-octave@bevo.che.wisc.edu Date: Tue, 22 Jan 2002 14:40:24 -0500 It's an IEEE 754 issue. From the article, "What Every Computer Scientist Should Know About Floating-Point Arithmetic" in the section on signed zero: "If a distinction were made when comparing +0 and -0, simple tests like if (x = 0) would have very unpredictable behavior, depending on the sign of x. Thus the IEEE standard defines comparison so that +0 = -0, rather than -0 < +0." This document is available all over the web, including here: http://cch.loria.fr/documentation/IEEE754/ACM/goldbergSUN.ps scroll down to page 191 for a more complete discussion. It still doesn't explain why they didn't define both -0 = +0 and -0 < +0 as true. Paul Kienzle On Tue, Jan 22, 2002 at 01:27:19PM -0500, Lippert, Ross A. wrote: > > Does any one else besides me consider > > -0.0 < 0.0 == false > > to be a gcc bug? > > Isn't the (a) whole reason IEEE has two zeros is so that correct branch > cuts can be taken for the complex functions? > > > > -r > > -----Original Message----- > From: Paul Kienzle [mailto:pkienzle at jazz dot ncnr dot nist dot gov] > Sent: Tuesday, January 22, 2002 12:17 PM > To: robher at adinet dot com dot uy; bug-octave@bevo.che.wisc.edu > Subject: Re: IMPORTANT: Bug in "arg" function > > > After playing around in the debugger for a while, it seems that the problem > is that (-0 < 0) is not true. > > Looking at the value X(3), the debugger says that this is -0+0i. When > evaluated as a vector, the function complex::arg in the standard > C++ library returns atan2 (imag (x), real (x)), which apparently recognizes > -0 as a negative number. However, when evaluated as a scalar, Octave > automatically converts the type to real(X(3)) since the complex portion > is 0. This is then evaluated in liboctave/lo-mappers.cc using the > following: > > double > arg (double x) > { > if (x < 0.0) > return M_PI; > else > #if defined (HAVE_ISNAN) > return xisnan (x) ? octave_NaN : 0.0; > #else > return 0.0; > #endif > } > > Since x < 0.0 is false for x=-0.0, this evaluates to 0.0 rather than > M_PI. You get consistent behaviour if you replace > if (x < 0.0) > with > double y = copysign(1.0,x); > if (y < 0.0) > > Paul Kienzle > pkienzle at users dot sf dot net > > On Fri, Jan 18, 2002 at 05:11:40PM -0300, Roberto Hernandez wrote: > > Bug report for Octave 2.1.34 configured for i686-pc-linux-gnu > > > > Description: > > ----------- > > > > The "arg" function is returning different values when calculated on a > > vector and a scalar. > > > > Repeat-By: > > --------- > > > > ## Create one period of a sinewave and do the FFT > > t = 0 : pi/4 : 7*pi/4; > > x = sin(t); > > X = fft(x); > > > > ## Truncate the residual real values from the result > > indices = find( abs(real(X)) <= 20*eps ); > > X(indices) = i * imag(X(indices)); > > > > ## Truncate the residual imaginary values > > indices = find( abs(imag(X)) <= 20*eps ); > > X(indices) = real(X(indices)); > > > > ## Compute the phase for the whole vector and element-by-element > > a1 = arg(X); > > for i = 1:length(X) > > a2(i) = arg(X(i)); > > endfor > > > > ## Show the difference between the two > > printf("Arg on the vector\n"); > > a1(:) > > printf("Arg element-by-element\n"); > > a2(:) > > > > Result on my box: > > > > octave:1> bug_demo > > Arg on the vector > > ans = > > > > 0.00000 > > -1.57080 > > 3.14159 <----------- > > 3.14159 <----------- > > 0.00000 > > 3.14159 <----------- > > 0.00000 > > 1.57080 > > > > Arg element-by-element > > ans = > > > > 0.00000 > > -1.57080 > > 0.00000 > > 0.00000 > > 0.00000 > > 0.00000 > > 0.00000 > > 1.57080 > > > > > > Fix: > > --- > > > > ?????? > > > > > > > > Configuration (please do not edit this section): > > ----------------------------------------------- > > > > uname output: Linux ether.localdomain 2.4.17 #1 Thu Dec 27 01:09:52 > > UYT 2001 i686 unknown > > configure opts: --prefix=/usr/local/octave-2.1.34 --enable-shared > > --enable-dl --enable-lite-kernel > > Fortran compiler: g77 > > FFLAGS: -O > > F2C: > > F2CFLAGS: > > FLIBS: -lg2c -lm -L/usr/lib/gcc-lib/i386-redhat-linux/2.96 > > -L/usr/lib/gcc-lib/i386-redhat-linux/2.96/../../.. -lm > > CPPFLAGS: > > INCFLAGS: -I. -I. -I./liboctave -I./src -I./libcruft/misc > > -I./glob -I./glob > > C compiler: gcc, version 2.96 20000731 (Red Hat Linux 7.1 2.96-85) > > CFLAGS: -g -O2 -Wall > > CPICFLAG: -fPIC > > C++ compiler: c++, version 2.96 20000731 (Red Hat Linux 7.1 2.96-85) > > CXXFLAGS: -g -O2 -Wall > > CXXPICFLAG: -fPIC > > LDFLAGS: -g > > LIBFLAGS: -L. > > RLD_FLAG: -Xlinker -rpath -Xlinker > > /usr/local/octave-2.1.34/lib/octave-2.1.34 > > TERMLIBS: -lncurses > > LIBS: -ldl -lz -lm > > LEXLIB: > > LIBPLPLOT: > > LIBDLFCN: > > LIBGLOB: ./glob/glob.o ./glob/fnmatch.o > > DEFS: > > > > -DOCTAVE_SOURCE=1 -DSEPCHAR=':' -DSEPCHAR_STR=":" -DUSE_READLINE=1 > > -D__NO_MATH_INLINES=1 -DCXX_NEW_FRIEND_TEMPLATE_DECL=1 > > -DHAVE_LIBM=1 -DHAVE_LIBZ=1 -DF77_APPEND_UNDERSCORE=1 -DOCTAVE_LITE=1 > > -DHAVE_GETHOSTNAME=1 -DHAVE_GETPWNAM=1 -DSIZEOF_SHORT=2 -DSIZEOF_INT=4 > > -DSIZEOF_LONG=4 -DSIZEOF_LONG_LONG=8 -DHAVE_ALLOCA_H=1 -DHAVE_ALLOCA=1 > > -DNPOS=std::string::npos -DSTDC_HEADERS=1 -DHAVE_DIRENT_H=1 > > -DTIME_WITH_SYS_TIME=1 -DHAVE_SYS_WAIT_H=1 -DHAVE_ASSERT_H=1 > > -DHAVE_CURSES_H=1 -DHAVE_DLFCN_H=1 -DHAVE_FCNTL_H=1 -DHAVE_FLOAT_H=1 > > -DHAVE_FNMATCH_H=1 -DHAVE_GLOB_H=1 -DHAVE_GRP_H=1 -DHAVE_LIMITS_H=1 > > -DHAVE_MEMORY_H=1 -DHAVE_NCURSES_H=1 -DHAVE_POLL_H=1 -DHAVE_PWD_H=1 > > -DHAVE_SGTTY_H=1 -DHAVE_STDLIB_H=1 -DHAVE_STRING_H=1 > > -DHAVE_SYS_IOCTL_H=1 > > -DHAVE_SYS_PARAM_H=1 -DHAVE_SYS_POLL_H=1 -DHAVE_SYS_RESOURCE_H=1 > > -DHAVE_SYS_SELECT_H=1 -DHAVE_SYS_STAT_H=1 -DHAVE_SYS_TIME_H=1 > > -DHAVE_SYS_TIMES_H=1 -DHAVE_SYS_TYPES_H=1 -DHAVE_SYS_UTSNAME_H=1 > > -DHAVE_TERMCAP_H=1 -DHAVE_TERMIO_H=1 -DHAVE_UNISTD_H=1 > > -DHAVE_VARARGS_H=1 -DHAVE_ATEXIT=1 -DHAVE_BCOPY=1 -DHAVE_BZERO=1 > > -DHAVE_DUP2=1 -DHAVE_ENDGRENT=1 -DHAVE_ENDPWENT=1 -DHAVE_EXECVP=1 > > -DHAVE_FCNTL=1 -DHAVE_FORK=1 -DHAVE_GETCWD=1 -DHAVE_GETEGID=1 > > -DHAVE_GETEUID=1 -DHAVE_GETGID=1 -DHAVE_GETGRENT=1 -DHAVE_GETGRGID=1 > > -DHAVE_GETGRNAM=1 -DHAVE_GETPGRP=1 -DHAVE_GETPID=1 -DHAVE_GETPPID=1 > > -DHAVE_GETPWENT=1 -DHAVE_GETPWUID=1 -DHAVE_GETTIMEOFDAY=1 > > -DHAVE_GETUID=1 > > -DHAVE_GETWD=1 -DHAVE_LINK=1 -DHAVE_LOCALTIME_R=1 -DHAVE_LSTAT=1 > > -DHAVE_MEMMOVE=1 -DHAVE_MKDIR=1 -DHAVE_MKFIFO=1 -DHAVE_ON_EXIT=1 > > -DHAVE_PIPE=1 -DHAVE_POLL=1 -DHAVE_PUTENV=1 -DHAVE_READLINK=1 > > -DHAVE_RENAME=1 -DHAVE_RINDEX=1 -DHAVE_RMDIR=1 -DHAVE_SELECT=1 > > -DHAVE_SETGRENT=1 -DHAVE_SETPWENT=1 -DHAVE_SETVBUF=1 > -DHAVE_SIGACTION=1 > > -DHAVE_SIGPENDING=1 -DHAVE_SIGPROCMASK=1 -DHAVE_SIGSUSPEND=1 > > -DHAVE_STAT=1 > > -DHAVE_STRCASECMP=1 -DHAVE_STRDUP=1 -DHAVE_STRERROR=1 > -DHAVE_STRFTIME=1 > > -DHAVE_STRNCASECMP=1 -DHAVE_STRPTIME=1 -DHAVE_SYMLINK=1 > -DHAVE_TEMPNAM=1 > > -DHAVE_UMASK=1 -DHAVE_UNLINK=1 -DHAVE_USLEEP=1 -DHAVE_VFPRINTF=1 > > -DHAVE_VSPRINTF=1 -DHAVE_VSNPRINTF=1 -DHAVE_WAITPID=1 -DHAVE_LIBDL=1 > > -DHAVE_DLOPEN=1 -DHAVE_DLSYM=1 -DHAVE_DLERROR=1 -DHAVE_DLCLOSE=1 > > -DWITH_DL=1 -DWITH_DYNAMIC_LINKING=1 -DHAVE_TIMEVAL=1 -DHAVE_FINITE=1 > > -DHAVE_ISNAN=1 -DHAVE_ISINF=1 -DHAVE_ACOSH=1 -DHAVE_ASINH=1 > > -DHAVE_ATANH=1 -DHAVE_ERF=1 -DHAVE_ERFC=1 -DHAVE_ST_BLKSIZE=1 > > -DHAVE_ST_BLOCKS=1 -DHAVE_ST_RDEV=1 -DHAVE_TM_ZONE=1 > -DHAVE_GR_PASSWD=1 > > -DEXCEPTION_IN_MATH=1 -DRETSIGTYPE=void -DSYS_SIGLIST_DECLARED=1 > > -DHAVE_SYS_SIGLIST=1 -DHAVE_POSIX_SIGNALS=1 -DHAVE_GETRUSAGE=1 > > -DHAVE_TIMES=1 -DGNUPLOT_HAS_MULTIPLOT=1 -DGNUPLOT_HAS_FRAMES=1 > > > > > > > > > > ------------------------------------------------------------- > > Octave is freely available under the terms of the GNU GPL. > > > > Octave's home on the web: http://www.octave.org > > How to fund new projects: http://www.octave.org/funding.html > > Subscription information: http://www.octave.org/archive.html > > ------------------------------------------------------------- > > > > > > ------------------------------------------------------------- > Octave is freely available under the terms of the GNU GPL. > > Octave's home on the web: http://www.octave.org > How to fund new projects: http://www.octave.org/funding.html > Subscription information: http://www.octave.org/archive.html > ------------------------------------------------------------- > > > > ------------------------------------------------------------- > Octave is freely available under the terms of the GNU GPL. > > Octave's home on the web: http://www.octave.org > How to fund new projects: http://www.octave.org/funding.html > Subscription information: http://www.octave.org/archive.html > ------------------------------------------------------------- > ------------------------------------------------------------- Octave is freely available under the terms of the GNU GPL. Octave's home on the web: http://www.octave.org How to fund new projects: http://www.octave.org/funding.html Subscription information: http://www.octave.org/archive.html -------------------------------------------------------------