SDK Example Compilation Problem

Accelerated Computing CUDA CUDA Programming and Performance
#1

Hi!

I am a first-time user of Linux (Ubuntu 11.10) and have spent hours in installing CUDA, solving the Optimus problem and changing to gcc 4.4. But now, when I am trying to compile the SDK examples, I get another error. I haven’t found anything on the web about it. Maybe one can tell me how to fix it or whether I have done something wrong.

I think I haven’t done bad everything, since 3 examples are running well. But only 3, the other executables are missing. I would be glad if anyone could help me. I have to work with CUDA for my master thesis.

I have CUDA Toolkit 4.0 and a Geforce GT555m GPU. Sorry for bad english, but I am not a native speaker.

Thank you very much!

Here is the mentioned error:

me@mine:~/NVIDIA_GPU_Computing_SDK/C$ make

make[1]: Entering directory `/home/me/NVIDIA_GPU_Computing_SDK/C/common'

make[1]: Leaving directory `/home/me/NVIDIA_GPU_Computing_SDK/C/common'

make[1]: Entering directory `/home/me/NVIDIA_GPU_Computing_SDK/C/common'

make[1]: Leaving directory `/home/me/NVIDIA_GPU_Computing_SDK/C/common'

make[1]: Entering directory `/home/me/NVIDIA_GPU_Computing_SDK/C/common'

make[1]: Leaving directory `/home/me/NVIDIA_GPU_Computing_SDK/C/common'

make[1]: Entering directory `/home/me/NVIDIA_GPU_Computing_SDK/shared'

make[1]: Leaving directory `/home/me/NVIDIA_GPU_Computing_SDK/shared'

make[1]: Entering directory `/home/me/NVIDIA_GPU_Computing_SDK/C/src/asyncAPI'

make[1]: Leaving directory `/home/me/NVIDIA_GPU_Computing_SDK/C/src/asyncAPI'

make[1]: Entering directory `/home/me/NVIDIA_GPU_Computing_SDK/C/src/simpleMPI'

MPI not found, not building simpleMPI.

make[1]: Leaving directory `/home/me/NVIDIA_GPU_Computing_SDK/C/src/simpleMPI'

make[1]: Entering directory `/home/me/NVIDIA_GPU_Computing_SDK/C/src/simpleStreams'

make[1]: Leaving directory `/home/me/NVIDIA_GPU_Computing_SDK/C/src/simpleStreams'

make[1]: Entering directory `/home/me/NVIDIA_GPU_Computing_SDK/C/src/MonteCarloMultiGPU'

In file included from ../../common/inc/cutil_inline.h:20,

                 from MonteCarlo_gold.cpp:18:

../../common/inc/cutil_inline_runtime.h: In function ‘void __curandSafeCall(curandStatus_t, const char*, int)’:

../../common/inc/cutil_inline_runtime.h:328: error: ‘CURAND_STATUS_DOUBLE_PRECISION_REQUIRED’ was not declared in this scope

make[1]: *** [obj/x86_64/release/MonteCarlo_gold.cpp.o] Error 1

make[1]: Leaving directory `/home/me/NVIDIA_GPU_Computing_SDK/C/src/MonteCarloMultiGPU'

make: *** [src/MonteCarloMultiGPU/Makefile.ph_build] Error 2
#2

Are you sure you are trying to compile CUDA SDK 4.0? I dont’ see these errors on my system.

Also, since you mentioned you are doing CUDA for your master thesis. Have you looked at ArrayFire( Free GPU library with a huge set of functions {FFT’s, convolutions, image processing, grapphics support}). You can integrate your custom CUDA code with ArrayFire and save lot of your development time. It has both a C++ and device(pointer) interface. Take a look here