Skip to content

Invalid value of longdouble finfo tiny & eps on some platforms (Trac #1059) #1657

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

Jump to bottom
Closed
thouis opened this issue Oct 19, 2012 · 7 comments
Closed

Invalid value of longdouble finfo tiny & eps on some platforms (Trac #1059) #1657

thouis opened this issue Oct 19, 2012 · 7 comments
Labels
00 - Bug component: numpy.lib

Comments

@thouis
Copy link
Contributor

thouis commented Oct 19, 2012

Original ticket http://projects.scipy.org/numpy/ticket/1059 on 2009-03-19 by @pv, assigned to unknown.

On some platforms, finfo apparently reports bogus limits for longdouble:

On the G5 Mac, the user reported:

>>> import numpy as np
>>> info = np.finfo(np.longcomplex)
>>> print "eps:", info.eps, info.eps.dtype
>>> print "tiny:", info.tiny, info.tiny.dtype
>>> print "log10:", np.log10(info.tiny), np.log10(info.tiny/info.eps)
eps: 1.3817869701e-76 float128
tiny: -1.08420217274e-19 float128
log10: nan nan

Which means that both eps and tiny are bogus. The latter is also negative.
@thouis
Copy link
Contributor Author

thouis commented Oct 19, 2012

@pv wrote on 2009-03-19

One downstream failing test punched out in r6697; remove the KNOWNFAIL when this bug is fixed.

@thouis
Copy link
Contributor Author

thouis commented Oct 19, 2012

@pv wrote on 2009-03-19

Also this seems a bit dubious:

Can long double eps really be < 5e-27 on a 64-bit SPARC, when on a x86_64 machine it's only 1e-19?

@thouis
Copy link
Contributor Author

thouis commented Oct 19, 2012

@charris wrote on 2009-03-20

Re SPARC, long double is quad precision with 112 bit mantissa. So eps is probably in the range of 2e-34.

@thouis
Copy link
Contributor Author

thouis commented Oct 19, 2012

@pv wrote on 2009-03-23

The SPARC issue should be fixed in r6715 - r6716. (Though it's not related to this ticket.)

@thouis
Copy link
Contributor Author

thouis commented Oct 19, 2012

Milestone changed to Unscheduled by @mwiebe on 2011-03-24

@charris
Copy link
Member

charris commented Feb 18, 2014

Close this? The PPC long doubles are not ieee, I'm not even sure eps would have the conventional meaning. OTOH, it could be hard coded for that platform if it makes sense.

@charris
Copy link
Member

charris commented Feb 21, 2014

Closing this in favor of #2669.

@charris charris closed this as completed Feb 21, 2014
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
00 - Bug component: numpy.lib
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
2 participants
@charris @thouis