Hello,
I am having the same issue. I am using Linux Red Hat 3.4.6.
I am trying to compile and run some of the programs given with the SDK. I get the following:
bandwidthTest> make
/usr/bin/ld: cannot find -lcutil
collect2: ld returned 1 exit status
make: *** […/…/bin/linux/release/bandwidthTest] Error 1
I have confirmed that /usr/local/cuda-sdk/lib where the libcutil.a is located is in my LD_LIBRARY_PATH. What else do I need to do? I do not have a .so file, do I need that?
Any help anyone can provide would be greatly appreciated.
-lcutil failure means the that directory containing libcutil.a is not listed in the linking list if the file exists.
By default installation, it is should be C/lib, or relative to the project src in SDK, …/…/lib so you can see -L…/…/lib, the libcutil.a can be found and linked.
Your libcutil.a is not in default directory, that’s fine, but guarantee it will appear in the final link command like -L/usr/local/cuda-sdk/lib
elegant way is to change line LIBDIR in common/common.mk, or ROOTDIR, depending on your installation.
LD_LIBRARY does not take any effect during building process, it is effective to search directory for dynamic library loading while the program running.
You’re my hero!! This is the solution!! I spent all afternoon trying to compile a sdk sample <img src=‘The Official NVIDIA Forums | NVIDIA<#EMO_DIR#>/crying.gif’ class=‘bbc_emoticon’ alt=‘:’(’ />
I second that this solved my issue on ArchLinux with cuda 3.2…
I did this…
$ cd /usr/share/cuda-sdk/C/lib
$ sudo ln -s libcutil_x86_64.a libcutil.a
then…
nvcc … blah… -L /usr/share/cuda-sdk/C/lib … blah
I thought maybe I could specify a lib so I wouldn’t have to make the symlink, but, I need to know more about static/dynamic libraries and compilers for that to happen. Hope this helps.