1 |
Am Donnerstag 02 August 2007 23:36 schrieb Alexander Skwar: |
2 |
> · Florian Philipp <f.philipp@××××××.de>: |
3 |
> > You see, they are not compatible and even if some code works I wouldn't |
4 |
> > bet multimedia apps will perform well. |
5 |
> > |
6 |
> > With -mtune the instruction set stays the same. It is just "rearranged". |
7 |
> |
8 |
> Hm. Allright. When using just -mtune (ie. without -march), the |
9 |
> docs at |
10 |
> http://gcc.gnu.org/onlinedocs/gcc-4.1.2/gcc/i386-and-x86_002d64-Options.htm |
11 |
>l |
12 |
> |
13 |
> say: |
14 |
> | While picking a specific cpu-type will schedule things appropriately |
15 |
> | for that particular chip, the compiler will not generate any code that |
16 |
> | does not run on the i386 without the -march=cpu-type option being used. |
17 |
> |
18 |
> If -mtune=athlon-xp is used, code is generated which may make |
19 |
> use of 3dNOW!. 3dNOW! is, of course, not to be found on 386 :) |
20 |
> If the instruction set stays the same, code generated with |
21 |
> -mtune=athlon-xp would not be executable on 386 machines, if |
22 |
> I understand you correctly. |
23 |
> |
24 |
> Hm. With -mtune, the set of available instructions (ie. |
25 |
> stuff like 3dNOW!, I suppose?) is NOT changed from the default |
26 |
> of i386, is it? Or what does "Tune to cpu-type everything applicable |
27 |
> about the generated code, except for the ABI and the set of available |
28 |
> instructions." mean - especially note the "except for [...] the set of |
29 |
> available instructions" part. |
30 |
> |
31 |
> So with "-mtune=pentium-m -march=athlon-xp" I'm making the compiler |
32 |
> generate code which is "ordered" the way it's best for pentium-m |
33 |
> machines while allowing it to use athlon-xp instruction set? Is |
34 |
> that what I'm doing? |
35 |
> |
36 |
> If so, then it seems you're right - code will run, but maybe not |
37 |
> so well. |
38 |
> |
39 |
> Is that understanding correct? If so, then I really should think |
40 |
> twice about using "-mtune=pentium-m -march=athlon-xp", shouldn't |
41 |
> I? |
42 |
> |
43 |
> Curious, |
44 |
> |
45 |
> Alexander Skwar |
46 |
> -- |
47 |
|
48 |
At least that's how I understand the issue. At the moment I've got two ideas |
49 |
to solve your problem: |
50 |
|
51 |
1. set march to an inferior target (pentium-3 and pentium-3m seem okay: mmx |
52 |
and sse) and mtune for one (or even both?) of them |
53 |
2. set march to one of them and disable incompatible instruction sets with |
54 |
options like -mno-sse2 or -mno-3dnow |