Is Gnome2 not supported on the amd64 ARCH?

[fbsdmail at dnswatch.com]

On Mon, January 3, 2011 7:27 am, Luchesar V. ILIEV wrote:
> On 01/03/11 08:45, fbsdmail at dnswatch.com wrote:
>
>> # # # TOP POSTING IS EVIL # # #
>>
>>>> 2011-01-02 15:42, fbsdmail at dnswatch.com skrev:
>>>> <...>
>>>>
>>>>
>>>>> Hello again, and thanks for your response.
>>>>> I commented it out after responding to your response, and
>>>>> it happily built. I just figured I'd use the CPUTYPE?= option to
>>>>> gain better amd64 profiling, but apparently it's only _really_
>>>>> available for the i386 CPU's. I say that because I've always used
>>>>> that option when building on those ARCH types, and never ran into
>>>>> a problem. Oh well, hopefully sometime son, it'll be better
>>>>> supported on the amd64 - fingers crossed. :)
>>>>>
>>>>> Thanks again for taking the time to respond.
>>>>>
>>>>>
>>>>>
>>>>> --Chris
>>>>>
>>>>>
>>>>>
>>>>>
>>>>
>>>> Hi,
>>>>
>>>>
>>>>
>>>> I have an Intel CPU that's amd64 compatible, and I use
>>>> CPUTYPE?=native,
>>>> which never gave me any problems (I use it for all builds, including
>>>>  kernel and world). I can't say whether it works with AMD CPUs
>>>> though. Nor can I really say if it makes a difference, because I've
>>>> never tried without it.
>>>>
>>>> /Rolf
>>>> _______________________________________________
>>>> freebsd-amd64 at freebsd.org mailing list
>>>> http://lists.freebsd.org/mailman/listinfo/freebsd-amd64
>>>> To unsubscribe, send any mail to
>>>> "freebsd-amd64-unsubscribe at freebsd.org"
>>>>
>>>>
>> On Sun, January 2, 2011 9:35 am, Luchesar V. ILIEV wrote:
>>
>>> hw.model: AMD Athlon(tm) 64 X2 Dual Core Processor 6400+
>>> CPUTYPE?=	athlon64
>>>
>>>
>>>
>>> I've never encountered any CPUTYPE specific problems so far, and I
>>> have GNOME2 and almost 1000 ports compiled on my desktop, plus the OS
>>> itself.
>>>
>>>
>>> I wonder if "athlon-mp" is so different to be causing the problems
>>> you've encountered or if there was something else that got fixed
>>> coincidentally when you commented out CPUTYPE from make.conf.
>>
>> Greetings Luchesar, and thank you for your reply.
>> I've got a couple of other threads on the @stable list, and one
>> other on this one related to this. Last time I researched make(1) and
>> make.conf(5), the common consensus was that (open)ssl, and possibly a
>> couple of others were the only things that ever made use of the flag.
>> However, when I experimented heavily on older
>> CPU's, I discovered that CPUTYPE?= _did_ make a difference. In
>> some cases it simply made the difference for correctly recognizing the
>> CPU, in all cases, it added the "feature set" that that CPU
>> possessed - SSE, SSE2, 3DNow, etc... So, I find myself inclined to make
>> use of CPUTYPE?= whenever possible. Problem is, I don't always keep up
>> on gcc(1)'s changes/additions. Which I think is the case here. My
>> _guess_ is that they changed the
>> name(s) - however slightly, and I found out the "hard way". :-\ Bottom
>> line; I need to take the time, and find the difference(s) from
>> then<--to-->now to use it effectively.
>>
>> Thanks again for your reply.
>>
>>
>> --Chris
>>
>
> Hi Chris,
>
>
> Indeed, I was going to comment about those notes in the documentation as
> well. To be honest, I've always been lazy to do my own research in the
> FreeBSD's code -- it's rather typical for the documentation to lag
> behind the code, and that's understandable even for commercial projects.
>
> BTW, an interesting file to check is also /usr/share/mk/bsd.cpu.mk. And
> that's why for the moment I'm a bit reluctant (probably naively) about
> using CPUTYPE=native -- it is understood by GCC itself, but bsd.cpu.mk
> "knows" nothing about it. Well, on amd64 it really does nothing more
> than to add "-march=${CPUTYPE}" to CFLAGS, which should be OK with
> "native", but it also seems to "describe" available CPU capabilities
> (e.g. mmx, sse, 3dnow) for the other mk files -- and that depends on the
> correct value, provided by CPUTYPE. I'm not sure this would work as
> expected with "native", but then again I might be quite wrong.
>
> In any case, I (and likely others) will really appreciate if you share
> your findings with the list -- and thank you for doing it so far. :)

You're correct about bsd.cpu.mk, I researched it in the past, and my
final consensus was that it really only applies to building kernel(s),
and possibly world. In either of these cases it is usually safer to
_not_ use any profiling. So I only consider that file "informational".
I've added CPUTYPE?=athlon64 to my make.conf(5) and have since built
some 250 ports - all w/o incident. I know it was picked up during the
make(1) process, as make(1) echoed (as you already noted) -march=athlon64.
So all in all, I think there is no problems (assuming the _correct_ CPUTYPE
is declared). ;)
So I suppose in the "big picture" /determining/ _which_ CPUTYPE is up
to the user, and given that athlon-mp no longer works for me on AM2,
or AM3 CPU's, it's not always going to be consistent (YMMV). :)

>
> Cheers,
> Luchesar
>
>
> P.S. Oh, and, erm... sorry for that evil top posting... (blush)

LOL No worries. It's an easy thing to do,
but sadly, doing so corrupts the thread.

Best wishes, and thanks for your reply.

--Chris

> _______________________________________________
> freebsd-amd64 at freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-amd64
> To unsubscribe, send any mail to "freebsd-amd64-unsubscribe at freebsd.org"
>
>


-- 
kern:
FreeBSD 8.1-RELEASE amd64