float calc. strangeness in 5.21.11-ld

[p5pRT]

Migrated from rt.perl.org#124433 (status was 'rejected')

Searchable as RT124433$

[p5pRT]

From [email protected]

$ cat strange_thing_my.pl
use Devel​::Peek;
use POSIX qw/floor ceil/;
my $sum = 0.002*1000;
my $s2 = int($sum/100);
print "sum is $sum, s2 is $s2\n";
print "floor( sum - (s2 * 100) ) is ";
my $sx = floor($sum - ($s2*100));
print "$sx\n";

---

$ perlbrew exec --with perl-5.21.11 perl strange_thing_my.pl
perl-5.21.11

sum is 2, s2 is 0
floor( sum - (s2 * 100) ) is 2

$ perlbrew exec --with perl-5.21.11-ld perl strange_thing_my.pl
perl-5.21.11-ld

sum is 2, s2 is 0
floor( sum - (s2 * 100) ) is 1

$ perlbrew exec --with perl-5.21.2-ld perl strange_thing_my.pl
perl-5.21.2-ld

sum is 2, s2 is 0
floor( sum - (s2 * 100) ) is 2

---

problem that in perl-5.21.11-ld floor( sum - (s2 * 100) ) is 1, not 2.
is this a bug?
found in http​://www.cpantesters.org/cpan/report/1e041e66-ede9-11e4-93de-b9891b12eb46

perl-5.21.11-ld

Summary of my perl5 (revision 5 version 21 subversion 11) configuration​:

  Platform​:
  osname=linux, osvers=3.13.0-49-generic, archname=x86_64-linux-ld
  uname='linux green-u 3.13.0-49-generic #81~precise1-ubuntu smp wed
mar 25 16​:32​:15 utc 2015 x86_64 x86_64 x86_64 gnulinux '
  config_args='-de -Dprefix=/home/perlbrew/perls/perl-5.21.11-ld
-Duselongdouble -Dusedevel
-Aeval​:scriptdir=/home/perlbrew/perls/perl-5.21.11-ld/bin'
  hint=recommended, useposix=true, d_sigaction=define
  useithreads=undef, usemultiplicity=undef
  use64bitint=define, use64bitall=define, uselongdouble=define
  usemymalloc=n, bincompat5005=undef
  Compiler​:
  cc='cc', ccflags ='-fwrapv -fno-strict-aliasing -pipe
-fstack-protector -I/usr/local/include -D_LARGEFILE_SOURCE
-D_FILE_OFFSET_BITS=64',
  optimize='-O2',
  cppflags='-fwrapv -fno-strict-aliasing -pipe -fstack-protector
-I/usr/local/include'
  ccversion='', gccversion='4.6.3', gccosandvers=''
  intsize=4, longsize=8, ptrsize=8, doublesize=8,
byteorder=12345678, doublekind=3
  d_longlong=define, longlongsize=8, d_longdbl=define,
longdblsize=16, longdblkind=3
  ivtype='long', ivsize=8, nvtype='long double', nvsize=16,
Off_t='off_t', lseeksize=8
  alignbytes=16, prototype=define
  Linker and Libraries​:
  ld='cc', ldflags =' -fstack-protector -L/usr/local/lib'
  libpth=/usr/local/lib
/usr/lib/gcc/x86_64-linux-gnu/4.6/include-fixed
/usr/include/x86_64-linux-gnu /usr/lib /lib/x86_64-linux-gnu
/lib/../lib /usr/lib/x86_64-linux-gnu /usr/lib/../lib /lib
  libs=-lpthread -lnsl -lgdbm -ldl -lm -lcrypt -lutil -lc -lgdbm_compat
  perllibs=-lpthread -lnsl -ldl -lm -lcrypt -lutil -lc
  libc=, so=so, useshrplib=false, libperl=libperl.a
  gnulibc_version='2.15'
  Dynamic Linking​:
  dlsrc=dl_dlopen.xs, dlext=so, d_dlsymun=undef, ccdlflags='-Wl,-E'
  cccdlflags='-fPIC', lddlflags='-shared -O2 -L/usr/local/lib
-fstack-protector'

Characteristics of this binary (from libperl)​:
  Compile-time options​: HAS_TIMES PERLIO_LAYERS PERL_DONT_CREATE_GVSV
  PERL_HASH_FUNC_ONE_AT_A_TIME_HARD PERL_MALLOC_WRAP
  PERL_NEW_COPY_ON_WRITE PERL_PRESERVE_IVUV
  PERL_USE_DEVEL USE_64_BIT_ALL USE_64_BIT_INT
  USE_LARGE_FILES USE_LOCALE USE_LOCALE_COLLATE
  USE_LOCALE_CTYPE USE_LOCALE_NUMERIC USE_LOCALE_TIME
  USE_LONG_DOUBLE USE_PERLIO USE_PERL_ATOF
  Built under linux
  Compiled at May 1 2015 13​:35​:33
  %ENV​:
  PERL5LIB=""
  PERLBREW_BASHRC_VERSION="0.69"
  PERLBREW_HOME="/home/vse/.perlbrew"
  PERLBREW_MANPATH="/home/perlbrew/perls/perl-5.21.11-ld/man"
  PERLBREW_PATH="/home/perlbrew/bin​:/home/perlbrew/perls/perl-5.21.11-ld/bin"
  PERLBREW_PERL="perl-5.21.11-ld"
  PERLBREW_ROOT="/home/perlbrew"
  PERLBREW_VERSION="0.69"
  @​INC​:
  /home/perlbrew/perls/perl-5.21.11-ld/lib/site_perl/5.21.11/x86_64-linux-ld
  /home/perlbrew/perls/perl-5.21.11-ld/lib/site_perl/5.21.11
  /home/perlbrew/perls/perl-5.21.11-ld/lib/5.21.11/x86_64-linux-ld
  /home/perlbrew/perls/perl-5.21.11-ld/lib/5.21.11
  .

[p5pRT]

From [email protected]

Victor Efimov wrote​:

problem that in perl-5.21.11-ld floor( sum - (s2 * 100) ) is 1, not 2.
is this a bug?

No.

$ perl-ld -lwe 'printf "%.75f\n", 0.002'
0.001999999999999999999916990771169078566771304394933395087718963623046875000
$ perl-ld -lwe 'printf "%.75f\n", 0.002*1000'
1.999999999999999999891579782751449556599254719913005828857421875000000000000

0.002 does not have a terminating binary representation, so cannot
be represented exactly in floating point. Any computation with it
necessarily suffers rounding error. 0.002*1000 in floating point is
really computing round(round(0.002)*1000)). This produces the above
inexact answer in the x87 80-bit format. As it happens, the same
computation in IEEE double precision produces the mathematically correct
answer, but that's really by accident​:

$ perl-d -lwe 'printf "%.75f\n", 0.002'
0.002000000000000000041633363423443370265886187553405761718750000000000000000
$ perl-d -lwe 'printf "%.75f\n", 0.002*1000'
2.000000000000000000000000000000000000000000000000000000000000000000000000000

-zefram

[p5pRT]

The RT System itself - Status changed from 'new' to 'open'

[p5pRT]

From [email protected]

Ok, indeed. Difference between versions confused me. Ticket can be closed.

On Fri May 01 04​:17​:52 2015, zefram@​fysh.org wrote​:

Victor Efimov wrote​:

problem that in perl-5.21.11-ld floor( sum - (s2 * 100) ) is 1, not
2.
is this a bug?

No.

$ perl-ld -lwe 'printf "%.75f\n", 0.002'
0.001999999999999999999916990771169078566771304394933395087718963623046875000
$ perl-ld -lwe 'printf "%.75f\n", 0.002*1000'
1.999999999999999999891579782751449556599254719913005828857421875000000000000

0.002 does not have a terminating binary representation, so cannot
be represented exactly in floating point. Any computation with it
necessarily suffers rounding error. 0.002*1000 in floating point is
really computing round(round(0.002)*1000)). This produces the above
inexact answer in the x87 80-bit format. As it happens, the same
computation in IEEE double precision produces the mathematically
correct
answer, but that's really by accident​:

$ perl-d -lwe 'printf "%.75f\n", 0.002'
0.002000000000000000041633363423443370265886187553405761718750000000000000000
$ perl-d -lwe 'printf "%.75f\n", 0.002*1000'
2.000000000000000000000000000000000000000000000000000000000000000000000000000

-zefram

[p5pRT]

From @jkeenan

Closing per request of original poster.

--
James E Keenan (jkeenan@​cpan.org)

[p5pRT]

@jkeenan - Status changed from 'open' to 'rejected'