Cflags question for intel core 2 duo

Message

0000000000000 · Post by **0000000000000** » Thu Mar 12, 2009 4:41 am

I have an Intel Core 2 Duo T6400 and am running 64bit on GCC-4.3.3. I currently have as my Cflags "-march=nocona -O2 -pipe" But read that I can use "-march=core2" with this GCC version.

Would there be any reasonable difference in anything if I were to change this and recompile world and system?

Also, what's this -funroll-loops I keep hearing about from trolls on 4chan?

Thanks.

(unrelated, my kernel (2.6.29-r6 from anholt's drm-intel git branch) only allows the sparse memory model as opposed to flat which I'd been using on my old computer, and I notice free -m lists 3814mb of memory as opposed to the 4096 I actually have. Is this related to the memory model, and/or why is sparse the only option?)

desultory · Post by **desultory** » Thu Mar 12, 2009 6:10 am

0000000000000 wrote:Would there be any reasonable difference in anything if I were to change this and recompile world and system?

For less intensive applications the difference would not be noticeable, for encoding or other similarly intensive uses it may well be noticeable.

0000000000000 wrote:Also, what's this -funroll-loops I keep hearing about from trolls on 4chan?

An old joke, since hijacked for some blog.

0000000000000 wrote:(unrelated, my kernel (2.6.29-r6 from anholt's drm-intel git branch) only allows the sparse memory model as opposed to flat which I'd been using on my old computer, and I notice free -m lists 3814mb of memory as opposed to the 4096 I actually have. Is this related to the memory model, and/or why is sparse the only option?)

It could also be due in part to a shared memory video chip set and the memory hole.

digrouz · Post by **digrouz** » Thu Mar 12, 2009 11:57 am

with gcc 4.3.3 you can also use -march=native which should enable the best cflags for you cpu

d2_racing · Post by **d2_racing** » Thu Mar 12, 2009 12:32 pm

In fact, nowadays I always use -march=native and it's GCC that set my Cflags and my CXXFlags.

digrouz · Post by **digrouz** » Thu Mar 12, 2009 1:01 pm

but to do that you need to unmask gcc-4.4.3 because the stable is 4.2.1

depontius · Post by **depontius** » Thu Mar 12, 2009 5:01 pm

digrouz wrote:but to do that you need to unmask gcc-4.4.3 because the stable is 4.2.1

I hope that's a typo, because my x86 stable is 4.1.2.

digrouz · Post by **digrouz** » Thu Mar 12, 2009 5:11 pm

Yes sorry you are right!

FizzyWidget · Post by **FizzyWidget** » Thu Mar 12, 2009 5:36 pm

save's me asking the same thing about my laptop - but i will ask this, is ssse3 a valid useflag?

danomac · Post by **danomac** » Thu Mar 12, 2009 5:57 pm

Carps wrote:save's me asking the same thing about my laptop - but i will ask this, is ssse3 a valid useflag?

Yes. Some packages have specific optimizations for SSSE3.

0000000000000 · Post by **0000000000000** » Fri Mar 13, 2009 12:47 am

> void null # grep sse4 /proc/cpuinfo | tail -n 1 | awk '{print( $49 ) }'
> sse4_1

Anything I can do with that one?

The internets suggest "media-libs/spandsp" is the only package that has such a USE

Edit: nevermind, Desultory's post cleared that up

desultory · Post by **desultory** » Fri Mar 13, 2009 12:48 am

Code: Select all

$ euse -i ssse3
global use flags (searching: ssse3)
************************************************************
no matching entries found

local use flags (searching: ssse3)
************************************************************
[-    ] ssse3 (media-video/ffmpeg):
faster floating point optimization for SSSE3 capable chips (Intel Core 2 and later chips)

[-    ] ssse3 (media-video/mplayer):
faster floating point optimization for SSSE3 capable chips (Intel Core 2 and later chips)