Thank you for your answer.
I get the point on “habits” but i think that as long as won’t exists something like NVCC it will be at least a little bit more difficult for us :)
Anyway now i’m facing new errors such as:
Undefined symbols for architecture i386:
“_oclLoadProgSource”, referenced from:
_OCLconv in OCLconv.cpp.o
“_oclGetFirstDev”, referenced from:
_OCLconv in OCLconv.cpp.o
“_oclLogBuildInfo”, referenced from:
_OCLconv in OCLconv.cpp.o
“_oclLogPtx”, referenced from:
_OCLconv in OCLconv.cpp.o
ld: symbol(s) not found for architecture i386
i linked my code with -framework OpenCL since i’m a mac user. am i doing something wrong? i’m including oclUtils.h.
EDIT: Solved, I swapped two cpp files, forgive me…
In fact, the external compiler in the form of NVCC, and the custom build rules you have to set up for CUDA are exactly the points that I dislike the most in CUDA. In OpenCL, the compiler is part of the driver, and you don’t need any custom build rules.
Anyway, speaking of “something like NVCC” for OpenCL, check out this topic about clcc.