CFlags for Intel Atom?

Message

pyther · Post by **pyther** » Thu Jun 12, 2008 11:15 pm

Hello.
I was wondering if anyone knows what CFlags I might want to use with the Intel Atom Processor. I just purchased the Intel D945GCLF which has an Atom cpu.

Thanks!

EDIT: Fixed Type on Model

poly_poly-man · Post by **poly_poly-man** » Thu Jun 12, 2008 11:50 pm

this?

no atom in there.... atom is a processor for handhelds - and AFAIK, it's not available anywhere yet.

What kind of processor do you really have (basically, on a livecd, cat /proc/cpuinfo)?

poly-p man

pyther · Post by **pyther** » Fri Jun 13, 2008 12:06 am

My bad I made a huge type the motherboard is Intel D945GCLF and it does have a atom processor.

poly_poly-man · Post by **poly_poly-man** » Fri Jun 13, 2008 12:24 am

can you give the contents of /proc/cpuinfo?

my guess is something like nocona, but I'm not sure what technologies this supports...

poly-p man

pyther · Post by **pyther** » Fri Jun 13, 2008 12:48 am

As I just ordered the hardware today I probably won't have the system until Tuesday or Wednesday of next week so I can't get the contents until then. Sorry

ergobill · Post by **ergobill** » Sat Jun 14, 2008 3:44 am

Pyther is correct this is a new product from Intel going after VIA market. Very Low power mb with 945 chipset in mini-itx form. Ubuntu (canonical) is working on "netbook" version to take advantage of / optimize for the atom processor.

Hun73r · Post by **Hun73r** » Mon Jun 16, 2008 12:07 am

Its basic a celeron with HT so the celeron CFlags probaly work with it ant it out performance the c7 from via

crazycat · Post by **crazycat** » Tue Jun 17, 2008 1:19 pm

Atom has as much in common with "celeron with ht" as a giraffe with a crocodile. The only thing in common is that they are both x86. Atom is in-order cpu. That means compiler optimizations will have much more influence on the speed of execution. Atom's performance is also pretty weak - cflags are more important for the in-order cpu. Here is a very good article about atom architecture.

tab21 · Post by **tab21** » Wed Jun 18, 2008 4:30 am

I have the same board and used the following in my make.conf to install 64 bit Gentoo:

Code: Select all

CFLAGS="-march=nocona -O2 -pipe -fomit-frame-pointer"
CXXFLAGS="${CFLAGS}"
MAKEOPTS="-j3"
CHOST="x86_64-pc-linux-gnu"
ALSA_CARDS="intel_hda"
INPUT_DEVICES="keyboard mouse"
VIDEO_CARDS="i810"

I'm sure you can use -march=prescott and a 32bit stage3 tarball to get up and running as well.

pyther · Post by **pyther** » Fri Jun 20, 2008 2:40 am

tab21: thanks, I'll try those

My /proc/cpuinfo

Code: Select all

processor	: 0
vendor_id	: GenuineIntel
cpu family	: 6
model		: 28
model name	:          Intel(R) Atom(TM) CPU  230   @ 1.60GHz
stepping	: 2
cpu MHz		: 1596.084
cache size	: 512 KB
physical id	: 0
siblings	: 2
core id		: 0
cpu cores	: 1
fpu		: yes
fpu_exception	: yes
cpuid level	: 10
wp		: yes
flags		: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx lm constant_tsc arch_perfmon pebs bts rep_good pni monitor ds_cpl tm2 ssse3 cx16 xtpr lahf_lm
bogomips	: 3196.02
clflush size	: 64
cache_alignment	: 64
address sizes	: 32 bits physical, 48 bits virtual
power management:

processor	: 1
vendor_id	: GenuineIntel
cpu family	: 6
model		: 28
model name	:          Intel(R) Atom(TM) CPU  230   @ 1.60GHz
stepping	: 2
cpu MHz		: 1596.084
cache size	: 512 KB
physical id	: 0
siblings	: 2
core id		: 0
cpu cores	: 1
fpu		: yes
fpu_exception	: yes
cpuid level	: 10
wp		: yes
flags		: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx lm constant_tsc arch_perfmon pebs bts rep_good pni monitor ds_cpl tm2 ssse3 cx16 xtpr lahf_lm
bogomips	: 3193.10
clflush size	: 64
cache_alignment	: 64
address sizes	: 32 bits physical, 48 bits virtual
power management:

wswartzendruber · Post by **wswartzendruber** » Fri Jun 20, 2008 12:06 pm

Intel states that the Atom is ISA compliant with the Merom:

Code: Select all

CFLAGS="-march=core2 -Os -fomit-frame-pointer -pipe"

But that's only if you're using GCC 4.3.

poly_poly-man · Post by **poly_poly-man** » Fri Jun 20, 2008 2:25 pm

wswartzendruber wrote:Intel states that the Atom is ISA compliant with the Merom:
Code: Select all
CFLAGS="-march=core2 -Os -fomit-frame-pointer -pipe"
But that's only if you're using GCC 4.3.

correct....

however, if:

gcc < 4.2.0, 32-bit: prescott
gcc < 4.2.0, 64-bit: nocona
gcc >= 4.2.0: native

no kidding, I just found out about march=native....I'm switching to that now

poly-p man

kernelOfTruth · Post by **kernelOfTruth** » Fri Jun 20, 2008 2:59 pm

poly_poly-man wrote:
wswartzendruber wrote:Intel states that the Atom is ISA compliant with the Merom:
Code: Select all
CFLAGS="-march=core2 -Os -fomit-frame-pointer -pipe"
But that's only if you're using GCC 4.3.
correct....

however, if:

gcc < 4.2.0, 32-bit: prescott
gcc < 4.2.0, 64-bit: nocona
gcc >= 4.2.0: native

no kidding, I just found out about march=native....I'm switching to that now

poly-p man

++

use -march=native, don't use -march=core2 since it breaks several apps like xorg-server, ... (at least for me + hardened)

wswartzendruber · Post by **wswartzendruber** » Sat Jun 21, 2008 5:16 am

kernelOfTruth wrote:
poly_poly-man wrote:
wswartzendruber wrote:Intel states that the Atom is ISA compliant with the Merom:
Code: Select all
CFLAGS="-march=core2 -Os -fomit-frame-pointer -pipe"
But that's only if you're using GCC 4.3.
correct....

however, if:

gcc < 4.2.0, 32-bit: prescott
gcc < 4.2.0, 64-bit: nocona
gcc >= 4.2.0: native

no kidding, I just found out about march=native....I'm switching to that now

poly-p man
++

use -march=native, don't use -march=core2 since it breaks several apps like xorg-server, ... (at least for me + hardened)

Is not march=native just going to determine that you're running a Core 2? And I've never had any problems with march=core2, but I'm not running hardened.

Jinidog · Post by **Jinidog** » Sun Jun 29, 2008 5:12 pm

Well, -march=core2 should work, but is there any evidence that it is the fastest option?
The atom is a in-order CPU. Every CPU since the Pentium-Pro was a Out-Of-Order CPU.

Somebody could do some benchmarks. I would try -march=i586 together mit the -msse -msse2 -msse3 flags.

m0w1337 · Post by **m0w1337** » Wed Jul 09, 2008 8:40 pm

Hi!

I just bought the same mainboard with the Atom on it and wondered, if you already found the right settings for this CPU to run good.

I would be very happy about an answer, because this is going to be my first attempt to install a gentoo linux, so i need every hint i can get

Thanks
m0

Cyker · Post by **Cyker** » Thu Jul 10, 2008 6:24 pm

At the moment, there is no optimisation support for the Atom.

It's a bit tricky, but I'd guess that Pentium MMX with all the SSE flags would be the most efficient setup at the moment...

Martux · Post by **Martux** » Sat Aug 30, 2008 1:39 pm

Pyther, how did this work for you? Everything up and running? Are you satisfied?
Which cflags did you use? I' ll get my Intel D945GCLF on tuesday and am looking for some practical tipps/experiences.
If i get it right it would be possible to use amd64 with this machine?
Also -march=native (GCC-4.3) works great here with a Core2Quad Q9450, i would really like to use it with the atom too. Which brings me to my main question:
I' d like to precompile the system with my Quad for obvious reasons. What exactly would happen with -march=native? It would be optimised for the Quad then, right?
Is precompiling for another machine even possible with that flag?

SDNick484 · Post by **SDNick484** » Thu Sep 11, 2008 1:25 am

Martux wrote:Pyther, how did this work for you? Everything up and running? Are you satisfied?
Which cflags did you use? I' ll get my Intel D945GCLF on tuesday and am looking for some practical tipps/experiences.
If i get it right it would be possible to use amd64 with this machine?
Also -march=native (GCC-4.3) works great here with a Core2Quad Q9450, i would really like to use it with the atom too. Which brings me to my main question:
I' d like to precompile the system with my Quad for obvious reasons. What exactly would happen with -march=native? It would be optimised for the Quad then, right?
Is precompiling for another machine even possible with that flag?

Wow, oddly enough I'm in the exact same boat. I just picked up an Intel C2Q 9300 from Fry's over the weekend ($180, sweet!), and my Acer One with an Atom N270 arrived yesterday. I got rid of it's base OS (Linpus, a Fedora 8 off shoot), and through on Xubuntu, but I'm very tempted to go with Gentoo (which I plan to run on the quad & already run on my gracefully aging laptop).

Anyways, if I get a chance to try it, I'll post back.

EDIT:
Went to look at the online documentation for GCC and noticed:
<<<<<
native - This selects the CPU to tune for at compilation time by determining the processor type of the compiling machine. Using -mtune=native will produce code optimized for the local machine under the constraints of the selected instruction set. Using -march=native will enable all instruction subsets supported by the local machine (hence the result might not run on different machines).
...
prescott - Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction set support.
...
core2 - Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3 instruction set support.
>>>>>

I will probably give native a try, but my hopes are lower. It looks like I may need prescott for builds to work on both systems.

d2_racing · Post by **d2_racing** » Thu Sep 11, 2008 2:04 am

poly_poly-man wrote: however, if:

gcc < 4.2.0, 32-bit: prescott
gcc < 4.2.0, 64-bit: nocona
gcc >= 4.2.0: native

no kidding, I just found out about march=native....I'm switching to that now

poly-p man

Hi Poly, if I use native with my T7200

Code: Select all

gentootux ~ # cat /proc/cpuinfo
processor       : 0
vendor_id       : GenuineIntel
cpu family      : 6
model           : 15
model name      : Intel(R) Core(TM)2 CPU         T7200  @ 2.00GHz
stepping        : 6
cpu MHz         : 2000.000
cache size      : 4096 KB
physical id     : 0
siblings        : 2
core id         : 0
cpu cores       : 2
apicid          : 0
initial apicid  : 0
fdiv_bug        : no
hlt_bug         : no
f00f_bug        : no
coma_bug        : no
fpu             : yes
fpu_exception   : yes
cpuid level     : 10
wp              : yes
flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe nx lm constant_tsc arch_perfmon pebs bts pni monitor ds_cpl vmx est tm2 ssse3 cx16 xtpr lahf_lm
bogomips        : 3993.26
clflush size    : 64
power management:

processor       : 1
vendor_id       : GenuineIntel
cpu family      : 6
model           : 15
model name      : Intel(R) Core(TM)2 CPU         T7200  @ 2.00GHz
stepping        : 6
cpu MHz         : 2000.000
cache size      : 4096 KB
physical id     : 0
siblings        : 2
core id         : 1
cpu cores       : 2
apicid          : 1
initial apicid  : 1
fdiv_bug        : no
hlt_bug         : no
f00f_bug        : no
coma_bug        : no
fpu             : yes
fpu_exception   : yes
cpuid level     : 10
wp              : yes
flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe nx lm constant_tsc arch_perfmon pebs bts pni monitor ds_cpl vmx est tm2 ssse3 cx16 xtpr lahf_lm
bogomips        : 3988.61
clflush size    : 64
power management:

gentootux ~ #

I will endup with a 32 bits or a 64 bits installation ?

Etal · Post by **Etal** » Thu Sep 11, 2008 2:50 am

d2_racing wrote:I will endup with a 32 bits or a 64 bits installation ?

It depends on your chost, not your cflags.

d2_racing · Post by **d2_racing** » Thu Sep 11, 2008 11:16 am

AM088 wrote:
d2_racing wrote:I will endup with a 32 bits or a 64 bits installation ?
It depends on your chost, not your cflags.

Good,that's what I tought

Martux · Post by **Martux** » Thu Sep 11, 2008 2:43 pm

SDNick484, thanks for your answer.
In the meantime, i' ve done the following:
I just copied over my "-march=native" compiled system from the C2Q, 64Bit, with adjusted kernel of course (hyperthreading and the likes) and it booted just fine.
Then i thought the atom would be powerful enough to re-compile itself using "native".
GCC alone took over 8h compile time...

(that's about the time the C2Q take compiling the whole system) So for now i just use it that way and cannot see any performance issues. It does what it should do: play music and videos and serve the net.
No need for more tuning. The only thing is i can't get GLX-extensions to work with the xf86-video-i810 driver whatsoever, but that will soon be another thread.
Thanks again for bringing light into the "mysterious" march=native thing.

d2_racing · Post by **d2_racing** » Fri Sep 12, 2008 2:38 am

Martux wrote:GCC alone took over 8h compile time... (that's about the time the C2Q take compiling the whole system) So for now i just use it that way and cannot see any performance issues. It e thing.

It's insane, 8h, maybe GCC is not optimised for this CPU.

int2str · Post by **int2str** » Fri Sep 12, 2008 11:31 pm

Martux wrote:GCC alone took over 8h compile time...

That's not right at all...

I just did it again to see how mine compares. Command was "time emerge gcc". Result:

* Messages for package sys-devel/gcc-4.1.2:
...
real 82m24.540s
user 75m56.971s
sys 6m51.811s

1h, 20mins... that's a long shot from your 8 hours.