| 1 |
Hi, |
| 2 |
|
| 3 |
On Wed, 31 May 2006 18:38:05 +0200 |
| 4 |
"Hemmann, Volker Armin" <volker.armin.hemmann@××××××××××××.de> wrote: |
| 5 |
|
| 6 |
> On Wednesday 31 May 2006 18:29, Hans-Werner Hilse wrote: |
| 7 |
> > On Thu, 1 Jun 2006 00:51:47 +0930 |
| 8 |
> > Raymond Lewis Rebbeck <dystopianray@×××××.com> wrote: |
| 9 |
> > > If it was harmless and beneficial it'd already be included in an -O? |
| 10 |
> > > level. |
| 11 |
> > |
| 12 |
> > Probably. And it seems to be only of interest when using |
| 13 |
> > -fsched2-use-superblocks or -fsched2-use-traces. The man page entry for |
| 14 |
> > the further (included in the latter) option says: "This option is |
| 15 |
> > experimental, as not all machine descriptions used by GCC model the CPU |
| 16 |
> > closely enough to avoid unreliable results from the algorithm." |
| 17 |
> [man page excerpt] |
| 18 |
> where does it say 'experimental'? |
| 19 |
|
| 20 |
You're right. At least for gcc 4.1.1 it doesn't say this anymore. For |
| 21 |
gcc 3.4.6, though, it does. Probably that's the explanation for this |
| 22 |
all. Nervertheless, it's not set by default for optimization options. |
| 23 |
That indicates it's still considered somewhat "beta" or provides |
| 24 |
another optimization strategy than the existing -O? options (and does |
| 25 |
not have some -O? option of its own, yet). |
| 26 |
|
| 27 |
-hwh |
| 28 |
-- |
| 29 |
gentoo-user@g.o mailing list |
Archives