On Tue, Jun 24, 2014 at 05:16:08PM +0200, Martti Kühne wrote:
> True.
> Hence, and following other sources about this topic [0], I suggest
>
> #define CEIL(x) ((int)(x) + ((int)(x) > 0) * ((x - (int)(x)) > FLT_EPSILON))
>
> cheers!
> mar77i
>
> [0]
> http://randomascii.wordpress.com/2012/02/25/comparing-floating-point-numbers-2012-edition/
I think that's still inaccurate. I don't think the standard ceil(3)
takes into account machine epsilon deltas:
~% python
Python 2.7.7 (default, Jun 3 2014, 16:16:56)
[GCC 4.8.3] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import math
>>> import numpy
>>> math.ceil(np.finfo(np.float64).eps) == 1
>>> math.ceil(numpy.finfo(numpy.float64).eps) == 1
True
The clause "np.finfo(np.float64).eps" returns the epsilon value.
Eric
