| 1 |
On Thursday 06 March 2003 13:19, Dhruba Bandopadhyay wrote: |
| 2 |
> Hello |
| 3 |
> |
| 4 |
> Today I saw in a gentoo cvs server email the following. |
| 5 |
> |
| 6 |
> 2003-03-05 Wednesday 11:01 method |
| 7 |
> |
| 8 |
> * sys-devel/gcc/: ChangeLog, gcc-3.2.2-r4.ebuild, |
| 9 |
> files/digest-gcc-3.2.2-r4: replace-flags "-march=pentium4" |
| 10 |
> "-march=pentium3" to fix bug #16867 |
| 11 |
> |
| 12 |
> Now, I realise that the pentium4 is prone to errors but I hope this does |
| 13 |
> not begin a spree of changes where -march=pentium4 flags will |
| 14 |
> automatically downgraded to -march=pentium3 by the ebuild or otherwise. |
| 15 |
> |
| 16 |
> I have used pentium4 for some time now with no problems whatsoever and |
| 17 |
> would like the raise the issue of whether downgrading of flags should be |
| 18 |
> manual and left to the user rather than being automatically modified. A |
| 19 |
> choice would be nice too. |
| 20 |
> |
| 21 |
> Also, what exactly is responsible for pentium4 bugs? Is it gcc and is |
| 22 |
> there hope of this being resolved with future versions? And will using |
| 23 |
> pentium3 result in a noticeable decline in performance? |
| 24 |
|
| 25 |
The problem with the pentium4 bugs is entirely in gcc. The packages which |
| 26 |
"downgrade" the pentium4 options do that because either: |
| 27 |
- the program does not compile with pentium4 |
| 28 |
- the program does not run with pentium4 |
| 29 |
- the program is instable with pentium4 |
| 30 |
|
| 31 |
In summary the pentium4 flag makes the specific program instable. The loss in |
| 32 |
speed should be minimal provided that -mcpu=pentium4 is still provided. |
| 33 |
Downgrading flags per ebuild I believe is not an option for most users, as it |
| 34 |
makes automatic merging impossible. For this reason the ebuilds that are |
| 35 |
instable with pentium4 edit the cflags so that the generated code is still |
| 36 |
optimized up to the best stable level. |
| 37 |
|
| 38 |
The only other option would be to have all those ebuilds warn the user about |
| 39 |
the problem with march=pentium4, and exit requesting the user to set his/her |
| 40 |
C(XX)FLAGS better. The thing is that most programs actually do work with |
| 41 |
-march=pentium4 so that people might want to use that. |
| 42 |
|
| 43 |
Paul |
| 44 |
|
| 45 |
ps. This flag downgrading concerns only gcc compilation. Not the code |
| 46 |
generated by this gcc. So other programs are still using -march=pentium4 |
| 47 |
where it is not filtered too. |
| 48 |
|
| 49 |
-- |
| 50 |
Paul de Vrieze |
| 51 |
Researcher |
| 52 |
Mail: pauldv@××××××.nl |
| 53 |
Homepage: http://www.cs.kun.nl/~pauldv |
Archives