RE: Kernel IEEE Math Completion (everything is zeroed!?!)
At 00:07 20.06.2006, Donsbach, Jeff wrote:
Uwe Schindler wrote:
> Same here on EV56 (Alphastation 500/500) with gcc 4.0.4 20060507
>(prerelease) (Debian 4.0.3-3) on Linux 2.6.16-2-alpha-generic:
>
> thetaphi@alpha:~$ ./test
> Please enter the num: 1e-320
> Please enter the den: 2
>
> 9.99989e-321/2 = 0
Actually Uwe, that last one is different from what Bob and I get:
jeff@alpha1 ~ $ ./test
Please enter the num: 1e-320
Please enter the den: 2
9.99989e-321/2 = 4.99994e-321
Now I'm confused. Which is supposed to be correct?
My result is the same as in the initial mail from Tyson. This is what he says:
0/0 = 0 (should be nan)
1/0 = 0 (should be inf)
1e-320/2 = 0 (should be 5e-321: 1e-320 is a denormalized)
So it seems that the last test works correct on your machine. The
difference is:
a) newer kernel on my machine - but I think nothing changed between
.16 and .15, could somebody make a diff in fixup code?
b) gcc 4.0 vs gcc 3.3
-----
Uwe Schindler
H.-H.-Meier-Allee 63, D-28213 Bremen
http://www.thetaphi.de - http://www.schindlers-software.de
eMails: uwe@thetaphi.de (private); info@schindlers-software.de (company)
Tel./Fax: +49 700 PCLATEIN (+49 700 72528346)
Schindlers Software - Home of Schindlers PC-LATEIN 3.20
DIE Software zum Lateinlernen!
Reply to: