1 |
Anthony Metcalf wrote: |
2 |
> Alan McKinnon wrote: |
3 |
>> Now the existing system should work with your new hardware and you can |
4 |
>> update your CFLAGS and 'emerge -e world' at your leisure. |
5 |
>> |
6 |
>> That's the theory at least anyway :-) |
7 |
>> |
8 |
>> |
9 |
> Well, exactly. That is the theory. I want to know the likelihood of |
10 |
> success. I know that using mtune=k6-2 means it won't run on anything |
11 |
> before a k6-2, and most likely not on anything Intel, due to the symbols |
12 |
> and optimisations used. What I want is some idea of the chance it will |
13 |
> run on a *later* AMD processor. Will an Athlon honour the k6-2 |
14 |
> optimisations? |
15 |
|
16 |
There's two points that come to mind. |
17 |
|
18 |
1) mtune is a request for the compiler to make the code more suited to |
19 |
the given processor, but without breaking compatibility. march is |
20 |
telling the compiler, do everything you can to make this code fastest on |
21 |
this processor. |
22 |
|
23 |
From the GCC docs for 4.2.3: |
24 |
"-mtune=cpu-type: Tune to cpu-type everything applicable about the |
25 |
generated code, except for the ABI and the set of available instructions." |
26 |
"-march=cpu-type: Generate instructions for the machine type cpu-type. |
27 |
The choices for cpu-type are the same as for -mtune. Moreover, |
28 |
specifying -march=cpu-type implies -mtune=cpu-type." |
29 |
|
30 |
So mtune shouldn't be using any instructions that are in K-6 that |
31 |
weren't in a 386. |
32 |
|
33 |
2) I believe x86 hardware never goes backwards. That is, if a new |
34 |
feature is added, all future versions of the chip have that feature, |
35 |
just with more added. Of course Intel and AMD both have their separate |
36 |
additions, but since your staying with AMD, moving to a new processor |
37 |
shouldn't break anything (even if you had used march). |
38 |
|
39 |
Disclaimer: I'm not an expert on hardware architectures or compilers, so |
40 |
I might be wrong. |
41 |
|
42 |
Shawn |
43 |
-- |
44 |
gentoo-user@l.g.o mailing list |