| 1 |
On Fri, 2006-07-07 at 15:21 +0200, Simon Stelling wrote: |
| 2 |
> Martin Schlemmer wrote: |
| 3 |
> > Stupid question though ... does the gcc test thingy list __3dNOW__ on |
| 4 |
> > nocona ? I would think that it does, as there is no -march=nocona (or |
| 5 |
> > whatever) yet. |
| 6 |
> |
| 7 |
> There is a -march=nocona, and it doesn't define __3dNOW__. |
| 8 |
> |
| 9 |
|
| 10 |
Missed that, sorry. |
| 11 |
|
| 12 |
> > So now you want to instead of fixing the amd64 profiles to be more |
| 13 |
> > flexible, implement something that will give the green light to users on |
| 14 |
> > x86 to use flag combinations, especially on older gcc's that causes |
| 15 |
> > great pain for themselfs and developers ? |
| 16 |
> |
| 17 |
> I don't understand this. Why is '-march=i686 -m3dnow' bad if it results in the |
| 18 |
> same thing as '-march=athlon-xp'? I guess I'm just lacking facts here, so please |
| 19 |
> give me a hint :) |
| 20 |
> |
| 21 |
|
| 22 |
Check Chris Gianelloni's mail just now. For some compilers with some |
| 23 |
-march's on x86 it did not explicitly turn on some features (or maybe |
| 24 |
not to such a high extend). So where say CFLAGS="-march=pentium3" would |
| 25 |
work, CFLAGS="-march=pentium3 -msse" would fail to build, or generate |
| 26 |
bad code (segfaulting binaries). |
| 27 |
|
| 28 |
|
| 29 |
-- |
| 30 |
Martin Schlemmer |
Archives