You are not logged in.
dear comrades,
I did a "pacman -Suy" on my arch, and a pretty complicated issue has emerged:
The new gcc is 4.7, but I work with NVidia's CUDA 4.1, and CUDA 4.1 doesn't work with gcc 4.7.
Some people told me to run another pacman command and fall back gcc. The problem is: somo other people said this is dangerous.
The question here is: can I have two gcc versions, like in ubuntu?
Have gcc 4.7 as default gcc and 4.6 as an other gcc that I will call manually?
If the answer is YES, how can I proceed?
thanks.
Offline
Search the AUR
Allan-Volunteer on the (topic being discussed) mailn lists. You never get the people who matters attention on the forums.
jasonwryan-Installing Arch is a measure of your literacy. Maintaining Arch is a measure of your diligence. Contributing to Arch is a measure of your competence.
Griemak-Bleeding edge, not bleeding flat. Edge denotes falls will occur from time to time. Bring your own parachute.
Offline
Dear Carcarah
You can use AUR for this task
in aur exist gcc4.5,4.4,3.x and much more
you care about time for compiling
and dont forget read how use other gcc in your specific program
and remmenber report the bug uptream
you Comrade another Arch User
Lenovo ThinkPad L420 modified
:: Intel i7 2560QM :: 8 GB RAM :: SSD 256 GB ::
:: DVD read+Writter :: 3 USB 3.0 Expresa Card ::
:: a Favulous 1 mins lasting Io-Li battery ::cry::
Offline
hello comrade. Thanks for helping me.
I have these two packages: gcc-4.6.2-7-i686.pkg.tar.xz gcc-libs-4.6.2-7-i686.pkg.tar.xz
I want to do this: keep 4.7 and install gcc-4.6.2 or other gcc that CUDA 4.1 suports.
How should I proceed?
Thank you!
By the way: I reported this bug to NVidia. There is no answer from them.
Offline
One more information:
I thing that CUDA 4.1 only suports gcc 4.4, I dont know about 4.5 and further.
Offline
Dear Carcarah
You need a tool like Packer or Yaourt for download and quasi-"automagicali" install gcc44 from AUR
You cannot use the standard gcc from the Old repos since many packages can break
for information about AUR and AUR-helpers read the archlinux Wikia
beffore you install gcc44 trou AUR you have 2 gcc, and need to specifi that for cuda that gcc use ( edit tha path of gcc for cuda to point to gcc44 default install-site
you Comrade Anoter Archer
Lenovo ThinkPad L420 modified
:: Intel i7 2560QM :: 8 GB RAM :: SSD 256 GB ::
:: DVD read+Writter :: 3 USB 3.0 Expresa Card ::
:: a Favulous 1 mins lasting Io-Li battery ::cry::
Offline
Dear comrade Jristz,
I am doing what you told me: installing gcc-4.5 from Yaourt .
The installation is very verbose and taking to long. hope this strategy works.
Thanks for helping!
Offline
Ok Found informatiuon abouth how set gcc44 and maintain a gcc4x too
http://timelordz.com/wiki/Nvidia_CUDA
http://www.linuxquestions.org/questions … -a-894576/
and in a theme completi random from this i found this ther
sudo ln -s /usr/bin/gcc-4.4 /usr/local/cuda/bin/gcc
sudo ln -s /usr/bin/g++-4.4 /usr/local/cuda/bin/g++in other woerd you can set the "default" gcc after using cuda or simlinking the gcc-cuda-link to the gcc4-path-equivalent
Last edited by Jristz (2012-04-23 03:34:58)
Lenovo ThinkPad L420 modified
:: Intel i7 2560QM :: 8 GB RAM :: SSD 256 GB ::
:: DVD read+Writter :: 3 USB 3.0 Expresa Card ::
:: a Favulous 1 mins lasting Io-Li battery ::cry::
Offline
This is what I am doing, comrade!
I tried gcc 4.5.3, but I got the same error! ![]()
I am trying gcc 4.4.6 in yaourt.
Let's see what will happen.
Thank you, comrade.
Offline
remmember search the path of gcc-4.4
since /usr/bin/gcc for default is gcc4.7
note: cuda is updated now in arch-repo...
Lenovo ThinkPad L420 modified
:: Intel i7 2560QM :: 8 GB RAM :: SSD 256 GB ::
:: DVD read+Writter :: 3 USB 3.0 Expresa Card ::
:: a Favulous 1 mins lasting Io-Li battery ::cry::
Offline
Dear comrades.
I did a gcc downgrade.
But the system is very slow running CUDA codes. I think it is the driver, I dont know. Thank you!
Last edited by carcarah (2012-04-23 13:54:35)
Offline
I successfully worked around this issue by:
- keeping gcc 4.7
- adding the following at the beginning of my .cu file
#undef _GLIBCXX_ATOMIC_BUILTINS
#undef _GLIBCXX_USE_INT128
- putting all of my non-CUDA-related c++ code into a separate .cpp file, and including it in the .cu file.
Offline