CUDA 3.2 incompatible with gcc 4.6.0 ?

Accelerated Computing CUDA CUDA Programming and Performance
1

Hellolo :)!

i am working on a cuda application. everything worked fine so far, but after upgrading gcc to 4.6.0 (i am running archlinux here), the compile fails. not sure, but i think this belongs to cuda-gcc-4.6.0 incompatibility (?). i’ve also got 4.4.0 installed, but don’t know how to force nvcc to use gcc 4.4.0 and gcc-libs 4.4.0 :-/

In file included from /usr/lib/gcc/x86_64-unknown-linux-gnu/4.6.0/…/…/…/…/include/c++/4.6.0/x86_64-unknown-linux-gnu/bits/gthr.h:162:0,
from /usr/lib/gcc/x86_64-unknown-linux-gnu/4.6.0/…/…/…/…/include/c++/4.6.0/ext/atomicity.h:34,
from /usr/lib/gcc/x86_64-unknown-linux-gnu/4.6.0/…/…/…/…/include/c++/4.6.0/bits/ios_base.h:41,
from /usr/lib/gcc/x86_64-unknown-linux-gnu/4.6.0/…/…/…/…/include/c++/4.6.0/ios:43,
from /usr/lib/gcc/x86_64-unknown-linux-gnu/4.6.0/…/…/…/…/include/c++/4.6.0/ostream:40,
from /usr/lib/gcc/x86_64-unknown-linux-gnu/4.6.0/…/…/…/…/include/c++/4.6.0/iostream:40,
from /home/stephan/src/cudaHYPE/problems/mhdGPUPaddle/Linux-x86_64/…/…/…/generic/utils/output.h:7,
from /home/stephan/src/cudaHYPE/problems/mhdGPUPaddle/Linux-x86_64/…/…/…/generic/utils/allocDevice.h:15,
from /home/stephan/src/cudaHYPE/problems/mhdGPUPaddle/Linux-x86_64/…/…/…/generic/utils/vectorHostDevice.h:13,
from /home/stephan/src/cudaHYPE/problems/mhdGPUPaddle/Linux-x86_64/…/paddle.h:5,
from /home/stephan/src/cudaHYPE/problems/mhdGPUPaddle/Linux-x86_64/…/paddle.cu:5:
/usr/lib/gcc/x86_64-unknown-linux-gnu/4.6.0/…/…/…/…/include/c++/4.6.0/x86_64-unknown-linux-gnu/bits/gthr-default.h:251:1: Fehler: das Einfügen von »__gthrw_« und »/* Android’s C library does not provide pthread_cancel, check for
`pthread_create’ instead. */« ergibt kein gültiges Präprozessor-Token

2

Yes, gcc 4.6 isn’t supported. You can force nvcc to use a particular version of gcc by populating a local directory somewhere with symbolic links to the compiler version you want, then passing the --compiler-bindir option with the absolute path of that directory. nvcc will then use the gcc/g++ links found in that directory, which will, in turn, build with the desired compiler version.

3

yes, thank you! that did it! :)