1 |
Daniel Iliev <danny@××××××××.com> posted 453F992B.90803@××××××××.com, |
2 |
excerpted below, on Wed, 25 Oct 2006 20:04:43 +0300: |
3 |
|
4 |
> What about enabling those extensions via C[XX]FLAGS? Is this still a bad |
5 |
> idea? |
6 |
|
7 |
In most cases, not necessarily bad, but redundant. A properly set -march= |
8 |
will take care of the CFLAGS AND allow the best choice in optimizations |
9 |
for your individual sub-arch. The exceptions would be the niches where |
10 |
there's no really correct -march= setting. The newest CPUs from both AMD |
11 |
and Intel have that problem ATM. AMD's newer CPUs (the ones with pni in |
12 |
the CPU flags, IIRC) may benefit from -msse3 in addition to -march=k8 (or |
13 |
alias), while Intel's aren't quite right either, but I've not been |
14 |
following them close enough to know what the best choice is (altho I know |
15 |
there's some disagreement). |
16 |
|
17 |
-- |
18 |
Duncan - List replies preferred. No HTML msgs. |
19 |
"Every nonfree program has a lord, a master -- |
20 |
and if you use the program, he is your master." Richard Stallman |
21 |
|
22 |
-- |
23 |
gentoo-amd64@g.o mailing list |