1 |
Am Freitag 03 August 2007 03:06 schrieb Daniel da Veiga: |
2 |
> On 8/2/07, Florian Philipp <f.philipp@××××××.de> wrote: |
3 |
> > Am Donnerstag 02 August 2007 23:36 schrieb Alexander Skwar: |
4 |
> > > · Florian Philipp <f.philipp@××××××.de>: |
5 |
> > > > You see, they are not compatible and even if some code works I |
6 |
> > > > wouldn't bet multimedia apps will perform well. |
7 |
> > > > |
8 |
> > > > With -mtune the instruction set stays the same. It is just |
9 |
> > > > "rearranged". |
10 |
> > > |
11 |
> > > Hm. Allright. When using just -mtune (ie. without -march), the |
12 |
> > > docs at |
13 |
> > > http://gcc.gnu.org/onlinedocs/gcc-4.1.2/gcc/i386-and-x86_002d64-Options |
14 |
> > >.htm l |
15 |
> > > |
16 |
> > > say: |
17 |
> > > | While picking a specific cpu-type will schedule things appropriately |
18 |
> > > | for that particular chip, the compiler will not generate any code |
19 |
> > > | that does not run on the i386 without the -march=cpu-type option |
20 |
> > > | being used. |
21 |
> > > |
22 |
> > > If -mtune=athlon-xp is used, code is generated which may make |
23 |
> > > use of 3dNOW!. 3dNOW! is, of course, not to be found on 386 :) |
24 |
> > > If the instruction set stays the same, code generated with |
25 |
> > > -mtune=athlon-xp would not be executable on 386 machines, if |
26 |
> > > I understand you correctly. |
27 |
> > > |
28 |
> > > Hm. With -mtune, the set of available instructions (ie. |
29 |
> > > stuff like 3dNOW!, I suppose?) is NOT changed from the default |
30 |
> > > of i386, is it? Or what does "Tune to cpu-type everything applicable |
31 |
> > > about the generated code, except for the ABI and the set of available |
32 |
> > > instructions." mean - especially note the "except for [...] the set of |
33 |
> > > available instructions" part. |
34 |
> > > |
35 |
> > > So with "-mtune=pentium-m -march=athlon-xp" I'm making the compiler |
36 |
> > > generate code which is "ordered" the way it's best for pentium-m |
37 |
> > > machines while allowing it to use athlon-xp instruction set? Is |
38 |
> > > that what I'm doing? |
39 |
> > > |
40 |
> > > If so, then it seems you're right - code will run, but maybe not |
41 |
> > > so well. |
42 |
> > > |
43 |
> > > Is that understanding correct? If so, then I really should think |
44 |
> > > twice about using "-mtune=pentium-m -march=athlon-xp", shouldn't |
45 |
> > > I? |
46 |
> > > |
47 |
> > > Curious, |
48 |
> > > |
49 |
> > > Alexander Skwar |
50 |
> > > -- |
51 |
> > |
52 |
> > At least that's how I understand the issue. At the moment I've got two |
53 |
> > ideas to solve your problem: |
54 |
> > |
55 |
> > 1. set march to an inferior target (pentium-3 and pentium-3m seem okay: |
56 |
> > mmx and sse) and mtune for one (or even both?) of them |
57 |
> > 2. set march to one of them and disable incompatible instruction sets |
58 |
> > with options like -mno-sse2 or -mno-3dnow |
59 |
> |
60 |
> Isn't the -march=i686 valid? |
61 |
> I guess that would be the most "compatible" option for binaries that |
62 |
> will run on AMD and Intel processors... Or simply use no "-march" |
63 |
> setting, only "-mtune"... I did that recently to switch a whole system |
64 |
> from an Athlon XP to a Intel Core Duo... |
65 |
> |
66 |
-march=i686 -mmmx -msse would be the better choice, then. |
67 |
|
68 |
I'd still like to know what gcc does when you enable two mtune settings. Use |
69 |
the last one? Use both? |