| View previous topic :: View next topic |
| Author |
Message |
Fracoon
Tux's lil' helper
Joined: 18 Mar 2004
Posts: 94
|
 Posted: Mon Nov 19, 2007 4:31 pm Post subject: Installation on Intel Q6600 Core 2 Quad |
 |
|
Hi@all
i just got my new Intel Q6600 Core 2 Quad with 4Gb of RAM... and now i have some questions.
I will install gentoo on this machine and use it with xen. But first of all i want to know how to configure the "base" system.
First Question : What live cd should i use? I want a 64bit system.
Second Question : What settings should be in make.conf? For the quad core and for XEN? I ask this because i dont want to build the system more than once =)
So any answers would be great..
thx... |
|
| Back to top |
|
 |
phsdv
Guru
Joined: 13 Mar 2005
Posts: 372
Location: Europe
|
| Posted: Mon Nov 19, 2007 5:18 pm Post subject: |
|
|
For a 64bit install you could use | Code: | | install-amd64-minimal-2007.0.iso |
or | Code: | | livecd-amd64-installer-2007.0.iso |
It sounds strange but you need amd64 for your intel core2!
Your make.conf could look like this: | Code: | WARNING this is not a complete make.conf, check with the docs for more details
CFLAGS="-O2 -march=nocona -pipe"
CFLAGS="${CFLAGS} -mno-tls-direct-seg-refs" # xen
CXXFLAGS="${CFLAGS}"
CHOST="x86_64-pc-linux-gnu"
FEATURES="sandbox usersandbox distfiles parallel-fetch fixpackages"
MAKEOPTS="-j8" # 8 for quad core
USE="nptl nptlonly" # system
USE="${USE} mmx mmxext sse sse2 ssse3" # processor
USE="${USE} smp" # at least for gimp maybe for others later...
# add your own use flags here...
|
I hope this helps. |
|
| Back to top |
|
|
Fracoon
Tux's lil' helper
Joined: 18 Mar 2004
Posts: 94
|
| Posted: Mon Nov 19, 2007 5:20 pm Post subject: |
|
|
thx alot. I will try this.
Do you think it makes sense to do a stage 1 install? |
|
| Back to top |
|
|
phsdv
Guru
Joined: 13 Mar 2005
Posts: 372
Location: Europe
|
| Posted: Mon Nov 19, 2007 5:24 pm Post subject: |
|
|
| Fracoon wrote: | | Do you think it makes sense to do a stage 1 install? |
I am sorry I do not know, I only did stage3 installs so far. Which means that you have to recompile your system after the install, but with a Q6600 and -j8 it will not take too long  |
|
| Back to top |
|
|
eyoung100
Veteran
Joined: 23 Jan 2004
Posts: 1428
|
|
| Back to top |
|
|
n0ns
n00b
Joined: 02 Aug 2007
Posts: 13
|
| Posted: Tue Nov 20, 2007 3:44 pm Post subject: |
|
|
It will take some time (especially gcc and glibc).
some packages pass to compiler to use j1.
Anyway, it will be much faster to use stage 3. As I know, there is no benefit from using stage 1, but a lot of dirty work and complications. |
|
| Back to top |
|
|
phsdv
Guru
Joined: 13 Mar 2005
Posts: 372
Location: Europe
|
| Posted: Tue Nov 20, 2007 8:57 pm Post subject: |
|
|
I just read in the howto: ( http://gentoo-wiki.com/HOWTO_Xen_and_Gentoo ): | Quote: | | Note: The '-mno-tls-direct-seg-refs' flag does not make sense on any 64bit system. For such systems you can skip the recompilation of the whole world and just recompile glibc |
So that would safe you some work. But I can confirm that including this flag does not harm... |
|
| Back to top |
|
|
Spidey
Apprentice
Joined: 07 Sep 2006
Posts: 269
|
| Posted: Wed Nov 21, 2007 1:20 am Post subject: |
|
|
| What are sandbox and usersandbox FEATURES? I mean, in the past week I had too many troubles which led me to use -sandbox in FEATURES... |
|
| Back to top |
|
|
likewhoa
l33t
Joined: 04 Oct 2006
Posts: 778
Location: Brooklyn, New York
|
| Posted: Wed Nov 21, 2007 1:36 am Post subject: |
|
|
| Spidey wrote: | | What are sandbox and usersandbox FEATURES? I mean, in the past week I had too many troubles which led me to use -sandbox in FEATURES... |
read the definition in /etc/make.conf.example |
|
| Back to top |
|
|
DrAgOnTuX
Apprentice
Joined: 22 Aug 2006
Posts: 176
Location: Switzerland
|
| Posted: Tue Dec 04, 2007 1:29 pm Post subject: |
|
|
why -j8 and not -j5 ?
i thought it was -j((number of cores) + 1)  |
|
| Back to top |
|
|
PaulBredbury
Watchman
Joined: 14 Jul 2005
Posts: 7310
|
| Posted: Tue Dec 04, 2007 4:12 pm Post subject: |
|
|
Because some of those parallel processes may be sleeping, waiting for hard disk activity.
The -j number is just a guesstimate anyway. |
|
| Back to top |
|
|
phsdv
Guru
Joined: 13 Mar 2005
Posts: 372
Location: Europe
|
| Posted: Tue Dec 04, 2007 4:33 pm Post subject: |
|
|
| DrAgOnTuX wrote: | why -j8 and not -j5 ?
i thought it was -j((number of cores) + 1) |
Because I want to keep each core 100% busy! |
|
| Back to top |
|
|
DrAgOnTuX
Apprentice
Joined: 22 Aug 2006
Posts: 176
Location: Switzerland
|
| Posted: Tue Dec 04, 2007 4:46 pm Post subject: |
|
|
| ok nice to know, I'm gonna buy a Q9450 if they're available |
|
| Back to top |
|
|
gm7uac
n00b
Joined: 17 Apr 2004
Posts: 57
Location: Glasgow
|
| Posted: Sun Dec 16, 2007 6:51 pm Post subject: problems with sata |
|
|
Hi
my system is a Core2 Quad 6600, I had problems seeing the sata controller with the gentoo install CD's
Linux localhost 2.6.23-gentoo-r3 #7 SMP PREEMPT Sun Dec 16 12:27:50 GMT 2007 i686 Intel(R) Core(TM)2 Quad CPU @ 2.40GHz GenuineIntel GNU/Linux
still fine tuning... having problems emergeing gimp complains about jpeg plugins.
_________________
If all else fails RTFM. |
|
| Back to top |
|
|
phsdv
Guru
Joined: 13 Mar 2005
Posts: 372
Location: Europe
|
| Posted: Sun Dec 23, 2007 2:42 pm Post subject: |
|
|
| phsdv wrote: | | DrAgOnTuX wrote: | why -j8 and not -j5 ?
i thought it was -j((number of cores) + 1) |
Because I want to keep each core 100% busy! |
I did some test to prove -j8 helps. I choose a kernel compile. First with -j8: | Code: | linux-2.6.23-gentoo-r5 # time make -j8
...(deleted compile stuff)...
Kernel: arch/x86_64/boot/bzImage is ready (#1)
real 1m41.928s
user 5m38.618s
sys 0m52.060s
linux-2.6.23-gentoo-r5 # make clean
linux-2.6.23-gentoo-r5 # time make -j8
.... (repeated to see if cache does something)...
Kernel: arch/x86_64/boot/bzImage is ready (#2)
real 1m40.929s
user 5m37.518s
sys 0m51.850s |
OK pretty much the same, lets try without -j8 | Code: | linux-2.6.23-gentoo-r5 # time make
......
real 5m22.479s
user 5m3.757s
sys 0m41.597s |
user time is actually shorter, 5m3.757s i.s.o 5m37.518s but who cares The real time you have been waiting is now 5m22.479s in stead of 1m41.928s. Lets see what -j5 is doing on my quad core: | Code: | linux-2.6.23-gentoo-r5 # make clean
linux-2.6.23-gentoo-r5 # time make -j5
......
real 1m40.506s
user 5m35.635s
sys 0m52.120s |
OK, wow it is indeed faster! -j5 gives 1m40.506s and -j8 1m40.929s. It is not much, only 0.4s but to my it is enough prove that -j((number of cores) + 1) is enough! |
|
| Back to top |
|
|
Monkeh
Veteran
Joined: 06 Aug 2005
Posts: 1656
Location: England
|
| Posted: Sun Dec 23, 2007 4:22 pm Post subject: |
|
|
| phsdv wrote: | | OK, wow it is indeed faster! -j5 gives 1m40.506s and -j8 1m40.929s. It is not much, only 0.4s but to my it is enough prove that -j((number of cores) + 1) is enough! |
No... It isn't. Run it again, it might well take another second. It very much depends on the software you're compiling. -j8 is good, as long as you have enough RAM. |
|
| Back to top |
|
|
phsdv
Guru
Joined: 13 Mar 2005
Posts: 372
Location: Europe
|
| Posted: Sun Dec 23, 2007 4:45 pm Post subject: |
|
|
| Monkeh wrote: | | phsdv wrote: | | OK, wow it is indeed faster! -j5 gives 1m40.506s and -j8 1m40.929s. It is not much, only 0.4s but to my it is enough prove that -j((number of cores) + 1) is enough! |
No... It isn't. Run it again, it might well take another second. It very much depends on the software you're compiling. -j8 is good, as long as you have enough RAM. |
Some more runs for statistics only (shown in sequence as run, with 4G ram):
| Code: | linux # make clean && time make -j5
real 1m44.332s
user 5m30.832s
sys 0m50.933s
linux # make clean && time make -j5
real 1m38.766s
user 5m34.358s
sys 0m50.713s
linux # make clean && time make -j5
real 1m39.456s
user 5m33.525s
sys 0m50.443s |
| Code: | linux # make clean && time make -j8
real 1m42.260s
user 5m38.278s
sys 0m51.850s
linux # make clean && time make -j8
real 1m40.763s
user 5m38.405s
sys 0m52.037s
linux # make clean && time make -j8
real 1m40.768s
user 5m37.961s
sys 0m51.367s |
User time is always lower with -j5 than -j8. Real time does differ, but on average -j5 is still faster. I think that with 4G I do have enough memory to do a -j8 without swapping, so that will not be the cause for a slower compile. |
|
| Back to top |
|
|
|
Kernel & Hardware
