[SciPy-user] numpy's math library?

David M. Cooke cookedm at physics.mcmaster.ca
Wed Jan 11 05:59:20 CST 2006


On Jan 10, 2006, at 12:10 , Alan G Isaac wrote:

> It was recently claimed on the Gnumeric list
> http://mail.gnome.org/archives/gnumeric-list/2006-January/ 
> msg00006.html
> that if libc is not available (I assume this means at
> compile time) then numpy uses a fallback library that is
> numerically naive.  (See the post for a specific example.)
>
> I suppose this would affect only Windows users,
> but I am one. Can someone tell me how this actually works?

Oh yeah, pick on one of the few functions defined like that :-)

If the inverse hyperbolic functions are not found, replacements are  
used for asinh, acosh, and atanh are used. This was added about 4  
years ago into Numeric.

I know for Numeric that on win32 the setup.py would set  
HAVE_INVERSE_HYPERBOLIC to 0, but with numpy, it actually checks for  
asinh. Have a look at build/src/numpy/core/config.h to see if it's  
been defined or not.

If necessary, we could use the definitions from fdlibm (http:// 
www.netlib.org/fdlibm/).

-- 
|>|\/|<
/------------------------------------------------------------------\
|David M. Cooke              http://arbutus.physics.mcmaster.ca/dmc/
|cookedm at physics.mcmaster.ca



More information about the SciPy-user mailing list