1 |
On Thu, 6 Jul 2006 13:49:39 +0100 |
2 |
Ciaran McCreesh <ciaran.mccreesh@×××××××××××××.uk> wrote: |
3 |
|
4 |
> On Thu, 6 Jul 2006 14:29:39 +0200 "Diego 'Flameeyes' Pettenò" |
5 |
> <flameeyes@g.o> wrote: |
6 |
> | On Thursday 06 July 2006 14:19, Ciaran McCreesh wrote: |
7 |
> | > Sounds rather flaky and unreliable... |
8 |
> | Sounds rather vague and without arguments. |
9 |
> |
10 |
> Well, you're assuming that a) everyone's using a C compiler, |
11 |
> b) that gcc has the slightest clue what it's doing, |
12 |
|
13 |
We're mostly talking about specially-written assembler code, not stuff |
14 |
like vectorisation or the intrinsics (which very few packages use, if |
15 |
any). |
16 |
|
17 |
> c) that the user has no |
18 |
> problem using nasty hacks to regain control, |
19 |
|
20 |
Control is easy. Specify the relevant -march in CFLAGS. |
21 |
|
22 |
> d) that this information is only needed at compile time, |
23 |
|
24 |
Where a package does run-time detection, there's no need for any |
25 |
conditional compilation as they build for everything anyway, so such |
26 |
packages wouldn't use mmx/sse/sse2 etc USE flags anyway. |
27 |
|
28 |
> e) that various gcc internal definitions won't change... |
29 |
|
30 |
Unlikely for these macros, as that would break a lot of existing code |
31 |
regardless what we do. |
32 |
|
33 |
> You're adding a lot of complexity, and |
34 |
> thus room for very weird breakages, to something that doesn't need it. |
35 |
|
36 |
I'd argue the current approach is the more complex approach, involving |
37 |
the user having to discover the relationship between their processor |
38 |
(which they've already set via -march in CFLAGS) and the various bits |
39 |
that their processor has. |
40 |
|
41 |
|
42 |
There are relatively few packages affected (<1%), so I think it's worth |
43 |
a try. In the end it may be that a few packages need to deal with |
44 |
stuff manually like with the current USE flags, but they'd be local USE |
45 |
flags at that point. |
46 |
|
47 |
-- |
48 |
Kevin F. Quinn |