This is the mail archive of the cygwin mailing list for the Cygwin project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

RE: exp() bug?

> -----Original Message-----
> From: cygwin-owner  On Behalf Of Lester Ingber
> Sent: 05 October 2004 17:05

> In reply to:
> : From: cygwin-owner On Behalf Of Lester Ingber
> : > Sent: 05 October 2004 16:36

> : > I first noticed some differences in results in 
> calculations using the
> : > MINGW gcc vs the Cygwin gcc on my ThinkPad XPPro machine. 
>  I applied
> : > the same code on FreeBsd and on SPARC/Solaris9.
> : 
> :   First go and google for "What every computer scientist 
> should know about
> : floating point", and read the entire paper.  Then try 
> running your tests again,
> : but compile them with the "-ffloat-store" option.  Then 
> read "info gcc" about
> : -ffloat-store and its effects.
> : 
> :   Then if you still think there's a bug, come back and 
> describe it.  But if you've
> : taken all that in, you no longer will.
> : 
> :     cheers, 
> :       DaveK
> I disagree about what the default behavior should be.  

  Well, that's just a matter of taste.  There's only one default and whichever way
it's set, it will be suitable for some applications but not for others.

> In any case, here are runs with -ffloat-store.  I see no difference.

  Ok, but you still didn't read that paper, did you?  Because if you had ...

> If true, this is a **very** serious bug, leading to very different
> answers on pretty standard (complex) calculations, as I discovered.

... you would never have written something like that, as you wouldn't _expect_ to
get the same results.  You would also understand about accuracy in the low places
and it's relation to the number of significant figures in your constants,
recurrent binary fractions, and the other issues (all well explained in the paper)
that your program shows.

  If you still insist there's a problem, you'll have to describe it more precisely
than "I expected this program to produce the same results on different systems
with different CPUs and different FP hardware/software", because that's not a
valid assumption.

Can't think of a witty .sigline today.... 

Unsubscribe info:
Problem reports:

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]