| 1 |
Le Sat, 30 Sep 2006 22:35:58 +0200, |
| 2 |
Lionel Bouton <lionel-dev@××××××.name> a écrit : |
| 3 |
|
| 4 |
> Hi, I just had an unpleasant experience with -ffast-math and GCC 4.1.1 |
| 5 |
> (it borked my LDAP authentication on several systems which worked with |
| 6 |
> the same CFLAGS as long as GCC 3.4.6 was used). |
| 7 |
> |
| 8 |
> There is a lot of material out there about CFLAGS and Gentoo (google |
| 9 |
> returns 387000 pages) but what's working for someone might not for |
| 10 |
> another. There are flags that work for a GCC version and most ebuilds |
| 11 |
> and don't work with another GCC version (my unfortunate experience) or |
| 12 |
> some ebuilds. Flag combination/architecture/LDFLAGS might be an issue too. |
| 13 |
> |
| 14 |
> There are already good resources (http://gentoo-wiki.com/CFLAGS_matrix |
| 15 |
> was mentioned to me by robbat2) but they may not be advertised enough. |
| 16 |
> I'd like to propose a paragraph to the GWN editor which presents some |
| 17 |
> gotchas and good references on the subject. |
| 18 |
> |
| 19 |
> Here's a draft for review. You're welcomed to expand on the subject. |
| 20 |
> |
| 21 |
> --- Draft BEGIN --- |
| 22 |
> <section> |
| 23 |
> <title>CFLAGS</title> |
| 24 |
> <body> |
| 25 |
> |
| 26 |
> <p> |
| 27 |
> Being able to tune the CFLAGS is part of one of the core principles of |
| 28 |
> Gentoo: let the user be in control. Being in control brings both |
| 29 |
> benefits and problems and CFLAGS tuning is not an exception. |
| 30 |
> </p> |
| 31 |
> <p> |
| 32 |
> The recent upgrade to gcc-4.1.1 for x86 and amd64 users changed the |
| 33 |
> landscape. Users that spent some time tuning their CFLAGS with gcc-3.4.6 |
| 34 |
> might find out that an upgrade to gcc-4.1.1 leaves them with an unstable |
| 35 |
> system. Example of this are : |
| 36 |
> <ul> |
| 37 |
> <li>nss_ldap stopped working with -ffast-math</li> |
| 38 |
> <li>...</li> |
| 39 |
> </ul> |
| 40 |
> </p> |
| 41 |
> <p> |
| 42 |
> Users with unsupported CFLAGS (see the <uri |
| 43 |
> link='http://gentoo-wiki.com/CFLAGS_matrix'>CFLAGS matrix</uri> for |
| 44 |
> example) might want to return to safe CFLAGS (see <uri |
| 45 |
> link='http://gentoo-wiki.com/Safe_Cflags'>Safe CFLAGS</uri>) if recent |
| 46 |
> updates caused them stability problems. On the other hand, more |
| 47 |
> adventurous users might want to experiment with CFLAGS that didn't work |
| 48 |
> properly with gcc-3.4.6... As always, the user is in control. |
| 49 |
> </p> |
| 50 |
> </body> |
| 51 |
> </section> |
| 52 |
> --- Draft END --- |
| 53 |
> |
| 54 |
> If possible, I'd like to expand the list of 3.4.6 -> 4.1.1 upgrade |
| 55 |
> problems which are linked to experimental CFLAGS. If you want to expand |
| 56 |
> the subject to cover other tuning/stability gotchas that recent updates |
| 57 |
> might have brought into the light, please feel free to do so. As English |
| 58 |
> is not my native tongue, feel free to spell check too. |
| 59 |
> |
| 60 |
> Cheers, |
| 61 |
> |
| 62 |
> Lionel. |
| 63 |
My personal experience with other CFLAGS as the ones in the handbook is at |
| 64 |
gcc-4.1.1 have a better optimisation with the default gentoo CFLAGS. Even |
| 65 |
with -O2, the result is a faster system, and -O3 seam to be safer with math |
| 66 |
related applications as with gcc-3.4.*. |
| 67 |
|
| 68 |
But in the other hand, other flags seam to be more problematic as with |
| 69 |
gcc-3.4.*. And the new optimisations flags as the vectorisation flags are not |
| 70 |
easy to use, because the result depend on the code of the program. They can or |
| 71 |
not brake the code, and when the program run well, they can make it faster |
| 72 |
or slower. All depend of the size and complexity of the loops. And I think also |
| 73 |
of the arch. |
| 74 |
|
| 75 |
So my conclusion is: |
| 76 |
For system flags, just keep the default, and if you want to experiment, do |
| 77 |
profiling for each single program you want to optimize. |
| 78 |
|
| 79 |
Cheers, |
| 80 |
Dominique |
| 81 |
|
| 82 |
-- |
| 83 |
Dominique Michel |
| 84 |
-- |
| 85 |
gentoo-dev@g.o mailing list |
Archives