Thanks to visit codestin.com
Credit goes to github.com

Skip to content

BUG: Overflow in tan and tanh for large complex values #5510

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
charris wants to merge 2 commits into from
Closed

BUG: Overflow in tan and tanh for large complex values #5510

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
5b58985
BUG: Overflow in tan and tanh for large complex values
ewmoore Feb 22, 2013
ff5a1c2
STY: C style cleanup.
charris Jan 27, 2015
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Next Next commit
BUG: Overflow in tan and tanh for large complex values 
np.tanh(1000+0j) gives nan+nan*j instead of 1.0+0j.
np.tan(0+1000j) gives  nan+nan*j instead of 1j.

I've imported the implementation for ctanh from FreeBSD's math library
which handles large arguments correctly and fixes this bug and the
equivalent bug in np.tan.

The problem here is that importing the function into npy_math_complex.c
and adding a config check causes us to use the implementation from glibc
on Linux, which also has this bug. Although it seems to have been fixed
in glibc last April
(http://sourceware.org/bugzilla/show_bug.cgi?id=11521).

I see several things that could be done here, the easiest is probably to
use our version of ctanh unconditionally. Although there are a multitude
of ways to go about that, for instance should I remove the
implementation from npy_math_complex.c and just place it directly in
umath/funcs.inc, where the buggy version was? Or should I just remove
the config check and add a note about it somewhere?

Closes #2321.
  • Loading branch information
@ewmoore @charris
ewmoore authored and charris committed Jan 27, 2015
commit 5b58985aa73b36fa717a33247eb66d800aaa57cc
52 changes: 51 additions & 1 deletion numpy/core/include/numpy/npy_math.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -151,6 +151,9 @@ double npy_copysign(double x, double y);
double npy_nextafter(double x, double y);
double npy_spacing(double x);

double npy_ldexp(double x, int exp);
double npy_frexp(double x, int* exp);

/*
* IEEE 754 fpu handling. Those are guaranteed to be macros
*/
Expand Down Expand Up @@ -258,8 +261,11 @@ float npy_copysignf(float x, float y);
float npy_nextafterf(float x, float y);
float npy_spacingf(float x);

float npy_ldexpf(float x, int exp);
Copy link
Copy Markdown
Contributor

@ewmoore ewmoore Jan 29, 2015

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

npy_ldexp and npy_frexp in all three flavors already exist in master. These came in as part of #4852, which pulled the ldexp and frexp changes from this.

float npy_frexpf(float x, int* exp);

/*
* float C99 math functions
* long double C99 math functions
*/

npy_longdouble npy_sinl(npy_longdouble x);
Expand Down Expand Up @@ -302,6 +308,9 @@ npy_longdouble npy_copysignl(npy_longdouble x, npy_longdouble y);
npy_longdouble npy_nextafterl(npy_longdouble x, npy_longdouble y);
npy_longdouble npy_spacingl(npy_longdouble x);

npy_longdouble npy_ldexpl(npy_longdouble x, int exp);
npy_longdouble npy_frexpl(npy_longdouble x, int* exp);

/*
* Non standard functions
*/
Expand Down Expand Up @@ -425,6 +434,19 @@ npy_cdouble npy_csqrt(npy_cdouble z);

npy_cdouble npy_ccos(npy_cdouble z);
npy_cdouble npy_csin(npy_cdouble z);
npy_cdouble npy_ctan(npy_cdouble z);

npy_cdouble npy_ccosh(npy_cdouble z);
npy_cdouble npy_csinh(npy_cdouble z);
npy_cdouble npy_ctanh(npy_cdouble z);

npy_cdouble npy_cacos(npy_cdouble z);
npy_cdouble npy_casin(npy_cdouble z);
npy_cdouble npy_catan(npy_cdouble z);

npy_cdouble npy_cacosh(npy_cdouble z);
npy_cdouble npy_casinh(npy_cdouble z);
npy_cdouble npy_catanh(npy_cdouble z);

/*
* Single precision complex functions
Expand All @@ -440,6 +462,20 @@ npy_cfloat npy_csqrtf(npy_cfloat z);

npy_cfloat npy_ccosf(npy_cfloat z);
npy_cfloat npy_csinf(npy_cfloat z);
npy_cfloat npy_ctanf(npy_cfloat z);

npy_cfloat npy_ccoshf(npy_cfloat z);
npy_cfloat npy_csinhf(npy_cfloat z);
npy_cfloat npy_ctanhf(npy_cfloat z);

npy_cfloat npy_cacosf(npy_cfloat z);
npy_cfloat npy_casinf(npy_cfloat z);
npy_cfloat npy_catanf(npy_cfloat z);

npy_cfloat npy_cacoshf(npy_cfloat z);
npy_cfloat npy_casinhf(npy_cfloat z);
npy_cfloat npy_catanhf(npy_cfloat z);


/*
* Extended precision complex functions
Expand All @@ -455,6 +491,20 @@ npy_clongdouble npy_csqrtl(npy_clongdouble z);

npy_clongdouble npy_ccosl(npy_clongdouble z);
npy_clongdouble npy_csinl(npy_clongdouble z);
npy_clongdouble npy_ctanl(npy_clongdouble z);

npy_clongdouble npy_ccoshl(npy_clongdouble z);
npy_clongdouble npy_csinhl(npy_clongdouble z);
npy_clongdouble npy_ctanhl(npy_clongdouble z);

npy_clongdouble npy_cacosl(npy_clongdouble z);
npy_clongdouble npy_casinl(npy_clongdouble z);
npy_clongdouble npy_catanl(npy_clongdouble z);

npy_clongdouble npy_cacoshl(npy_clongdouble z);
npy_clongdouble npy_casinhl(npy_clongdouble z);
npy_clongdouble npy_catanhl(npy_clongdouble z);


/*
* Functions that set the floating point error
Expand Down
39 changes: 38 additions & 1 deletion numpy/core/setup.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -209,6 +209,32 @@ def check_prec(prec):
else:
priv.extend([(fname2def(f), 1) for f in flist])

flist = [f + prec for f in C99_COMPLEX_FUNCS_CHECKED]
Copy link
Copy Markdown
Member Author

@charris charris Jan 27, 2015

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Makes me wonder if these tests shouldn't be done in python.

Copy link
Copy Markdown
Contributor

@ewmoore ewmoore Jan 28, 2015

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

There was some discussion of using ctypes to call these functions from python. I didn't go down that route since we then need to find libc ourselves, but doing it this way allowed the tool chain to find it for us. I suppose we could make a tiny module that simply exposes them to python and then use that for the tests.

decl = dict([(f, True) for f in flist])
exists = []
if not config.check_funcs_once(flist, call=decl, decl=decl,
libraries=mathlibs):
for f in C99_COMPLEX_FUNCS_CHECKED:
if config.check_func(f + prec, call=True, decl=True,
libraries=mathlibs):
exists.append(f)
else:
exists.extend(C99_COMPLEX_FUNCS_CHECKED)

if len(exists) > 0:
fp = open(join('.', 'numpy', 'core', 'src', 'npymath',
'test_c99complex.c'), 'r')
obody = fp.read()
fp.close()
precname = {'f':'FLOAT', '':'DOUBLE', 'l':'LONGDOUBLE'}[prec]
for f in exists:
body = obody.replace('PYTESTPRECISION', precname) \
.replace('PYTESTFUNC', f.upper())
inc_dir = join('.', 'numpy', 'core', 'src', 'npymath')
if config.try_run(body, libraries=mathlibs,
include_dirs=[inc_dir]):
priv.append((fname2def(f + prec), 1))

check_prec('')
check_prec('f')
check_prec('l')
Expand Down Expand Up @@ -683,7 +709,8 @@ def get_mathlib_info(*args):
npymath_sources = [join('src', 'npymath', 'npy_math.c.src'),
join('src', 'npymath', 'ieee754.c.src'),
join('src', 'npymath', 'npy_math_complex.c.src'),
join('src', 'npymath', 'halffloat.c')]
join('src', 'npymath', 'halffloat.c'),
join('src', 'npymath', 'fpstatus.c')]
config.add_installed_library('npymath',
sources=npymath_sources + [get_mathlib_info],
install_dir='lib')
Expand Down Expand Up @@ -973,6 +1000,16 @@ def generate_umath_c(ext, build_dir):
config.add_extension('operand_flag_tests',
sources = [join('src', 'umath', 'operand_flag_tests.c.src')])

#######################################################################
# npymath_tests module #
#######################################################################

config.add_extension('npymath_tests',
sources = [join('src', 'npymath', 'npymath_tests.c')],
depends = ['test_c99complex.c'],
libraries = ['npymath']
)

config.add_data_dir('tests')
config.add_data_dir('tests/data')

Expand Down
6 changes: 4 additions & 2 deletions numpy/core/setup_common.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -153,8 +153,10 @@ def check_api_version(apiversion, codegen_dir):

C99_COMPLEX_TYPES = ['complex double', 'complex float', 'complex long double']

C99_COMPLEX_FUNCS = ['creal', 'cimag', 'cabs', 'carg', 'cexp', 'csqrt', 'clog',
'ccos', 'csin', 'cpow']
C99_COMPLEX_FUNCS = ['creal', 'cimag', 'cabs', 'carg']
C99_COMPLEX_FUNCS_CHECKED = ['cacos', 'casin', 'catan', 'cacosh', 'casinh',
'catanh', 'ccos', 'csin', 'ctan', 'ccosh', 'csinh',
'ctanh', 'cexp', 'clog', 'cpow', 'csqrt']

def fname2def(name):
return "HAVE_%s" % name.upper()
Expand Down
Loading