(I originally posted this in the OpenCL Forum, but got no responses, so I’m guessing that it’s more of an issue with CUDA 3.1 and Fedora 12 than something with OpenCL and CUDA 3.1. Thus I’m posting it here to see if anyone has come across this problem before / solved it.)
I just installed OpenCL on my machine (OS = Fedora 12). Previously, I had CUDA SDK 3.0 installed. However, while installing these files, it overwrote some of the stuff from that (it asks you if it’s ok to overwrite the files in the install process). Before installing OpenCL, I could compile all of the CUDA SDK examples. However, that is no longer the case (Interesting Point: I can compile all of the OpenCL examples without problem, and oclDeviceQuery and oclBandwidthTest both passed). Has anyone else had problems like this or solved them?
Because I’m using Fedora 12 (which has gcc 4.4), I needed to do some previous “hacking” of the CUDA files to get everything to compile correctly. These links are the ones I followed:
So, basically, I’ve made all the changes outlined in those already (I didn’t add anything to my common.mk previously, but I did that after installing OpenCL to solve the “stdarg” problem).
Here’s the error I’m getting when I compile the CUDA benchmarks:
make: Entering directory `/home/NVIDIA_GPU_Computing_SDK/C/src/FunctionPointers' In file included from FunctionPointers.cpp:33: FunctionPointers_kernels.h:45: warning: deprecated conversion from string constant to â€˜char*â€™ FunctionPointers_kernels.h:45: warning: deprecated conversion from string constant to â€˜char*â€™ FunctionPointers_kernels.h:45: warning: deprecated conversion from string constant to â€˜char*â€™ FunctionPointers.cpp:50: warning: deprecated conversion from string constant to â€˜char*â€™ nvopencc WARNING: -noinline or -INLINE:=off has been seen, -default_options ignored ### Assertion failure at line 1312 of ../../be/cg/NVISA/cgemit_targ.cxx: ### Compiler Error in file /tmp/tmpxft_00003f80_00000000-7_FunctionPointers_kernels.cpp3.i during Assembly phase: ### NYI initv kind 1 nvopencc INTERNAL ERROR: /usr/local/cuda/open64/lib//be returned non-zero status 1 make: *** [obj/x86_64/release/FunctionPointers_kernels.cu_20.o] Error 1
Any help would be greatly appreciated!