| 1 |
On Sun, 2005-03-27 at 12:48 +0900, Chris White wrote: |
| 2 |
> My God.. I love this part: |
| 3 |
> |
| 4 |
> Ok, so I had some users complain about mmx2 being showed as enabled |
| 5 |
> in mplayer despite the lack of mmxext in /proc/cpuinfo. Therefore I |
| 6 |
> decided to get upstream help. Here's what we got...: |
| 7 |
> |
| 8 |
> mmx2 = sse |
| 9 |
> sse = mmx2 |
| 10 |
> |
| 11 |
> but in mplayer's configure script: |
| 12 |
> |
| 13 |
> mmx2 = sse's integer logic |
| 14 |
> sse = sse's floating point stuff |
| 15 |
> |
| 16 |
> so mmx2 is NOT a flag for mmx2 support.. it's a flag for integer portion |
| 17 |
> of sse... |
| 18 |
> |
| 19 |
> ok, so at this point I'm confused as to how to approach this. What |
| 20 |
> about AMD? Do they use mmxext for stuff like that.. or sse2 extensions |
| 21 |
> (as someone mentioned). My plan is to have sse USE flag enable mmx2 |
| 22 |
> only (since it's "the best") and disable sse (since it's the same thing |
| 23 |
> but "worse"). However, that only takes care of intel.. what about AMD |
| 24 |
> processors though? They deal with all this stuff differently. |
| 25 |
> Recomendations are welcome at this point.. I'm going to get some asprin... |
| 26 |
|
| 27 |
That sounds very brain dead by upstream. IMHO they really shouldn't use |
| 28 |
keywords for things which don't do what everybody else would expect them |
| 29 |
to do. Did you already propose them to change this behavior and |
| 30 |
introduce either another keyword for the floating point stuff or merge |
| 31 |
those two, enabling it by default when sse is enabled. |
| 32 |
|
| 33 |
I'm kinda confused now... |
| 34 |
|
| 35 |
-- |
| 36 |
(o_ Markus Nigbur |
| 37 |
//\ Gentoo GNU/Linux Developer |
| 38 |
[ ]/_ http://www.gentoo.org |
| 39 |
GPG Fingerprint: 9E31 187A 6C78 210C F248 EDBB F2FB A77F 707D F8C1 |
| 40 |
|
| 41 |
-- |
| 42 |
gentoo-dev@g.o mailing list |
Archives