| 1 |
On Fri, 9 Jun 2006, Vladimir G. Ivanovic wrote: |
| 2 |
|
| 3 |
> |
| 4 |
> On Fri, 2006-06-09 at 07:47 +0200, Michael Weyershäuser wrote: |
| 5 |
> > I don't recommend -ffast-math on a global basis (or at all), it gets you |
| 6 |
> > right where the Pentium fdiv bug brought us some years ago: floating |
| 7 |
> > point operations will be faster, but they could get a wrong result in |
| 8 |
> > some occasions... |
| 9 |
> |
| 10 |
> Linus seems to use -ffast-math: |
| 11 |
> http://gcc.gnu.org/ml/gcc/2001-07/msg02150.html. At least in 2001 he |
| 12 |
> did. |
| 13 |
|
| 14 |
I use it too, in my numerical simulations of the Schroedigner equation, I |
| 15 |
think it is fairly standard in the numerics community. My understanding |
| 16 |
is that it results in a fairly consistent and predictable loss of |
| 17 |
precision for some operations, nothing like the catastrophic *errors* in |
| 18 |
the the intel fdiv bug. Of course, for algorithms that are sensitive to |
| 19 |
precision, it can (and does) cause problems. |
| 20 |
|
| 21 |
Cheers, |
| 22 |
Ian |
Archives