| 1 |
On Friday 09 June 2006 23:42, Vladimir G. Ivanovic wrote: |
| 2 |
> On Fri, 2006-06-09 at 16:14 -0500, Barry.SCHWARTZ@×××××××××××××.org |
| 3 |
> |
| 4 |
> wrote: |
| 5 |
> > Enough research has gone into writing code for IEEE floating point |
| 6 |
> > that I would not try to bypass it without a good reason. It’s there to |
| 7 |
> > be your friend. |
| 8 |
> |
| 9 |
> Performance is the reason we have hardware FPUs and -ffast-math. |
| 10 |
|
| 11 |
no, ffast-math is for the case, that you |
| 12 |
|
| 13 |
a) don't need accurate results |
| 14 |
|
| 15 |
b) the FPU is not fast enough |
| 16 |
|
| 17 |
|
| 18 |
since you can't say if you need accurate math for an app you did not wrote or |
| 19 |
examined, using ffast-math is highly dangerous. And sicne the hardware gets |
| 20 |
faster on an almost monthly basis, it is even less convincing why anybody |
| 21 |
should use it. |
| 22 |
|
| 23 |
-- |
| 24 |
gentoo-amd64@g.o mailing list |
Archives