Re:No One Peer-Reviews Scientific Software
by spamizbad on Monday November 30, 2009 @01:10PM
(#968322)
Also these climate models are likely calculated on Intel hardware: a company notorious for floating-point bugs. Even if we verify the software is correct, unless we see the RTL that describes the logic behind these chips, it's all just hocus-pocus as far as I'm concerned.
[ Parent
| on HN ]
Re:No One Peer-Reviews Scientific Software
by titusflavius on Monday November 30, 2009 @02:03PM
(#968416)
You can't verify software is correct, it can always fail at time t+1.
Basic configuration testing on PPC or other architectures would highlight a lot of the error-sensitive paths in the math libraries.