Messages in this thread | | | Date | Sat, 7 Nov 2009 00:49:23 +0000 | From | Alan Cox <> | Subject | Re: i686 quirk for AMD Geode |
| |
On Fri, 6 Nov 2009 23:18:06 +0100 Matteo Croce <technoboy85@gmail.com> wrote:
> On Fri, Nov 6, 2009 at 5:44 PM, H. Peter Anvin <hpa@zytor.com> wrote: > > On 11/06/2009 06:59 AM, Matteo Croce wrote: > >> indeed it has MMX, MMXEXT and CMOV, just lacks the long NOP instruction (NOPL). > > > > MMX and MMXEXT are hardly hallmarks of i686, which leaves only cmov.
cmov is not i686 either. Its an optional extension that *should* only be used after you test its available. So gcc "i686" isn't quite "686". There are of course good reasons for that choice.
> yes, I did some test like gzip, bzip2, lame etc and they give more or less > the same results of dhrystone
How does that compare to i486. Certainly the old Nat Semi geode seemed to prefer to be fed i486 code to i586 (and wouldn't of course hack i686). You might also want to play around with -mtune= as well as arch= before assuming why i686 is a win
(I still btw think the patch is a good idea, it simplifies life enormously for users with the CPU)
Alan
| |