convolutionSeparable compilation question

Accelerated Computing CUDA CUDA Programming and Performance
#1

So I am trying to compile and run the convolutionSeparable program. The provided executable file will run, but I was trying to compile it myself since we are doing a project at work that is basically the same thing. I am trying to compile it using nvcc, but it seems it can’t find any of the needed. The error I get is: convolutionSeparable.cu:46:26: error: cutil_inline.h: No such file or directory. Am I just totally missing where this file is located? The command I ran was:

nvcc convolutionSeparable.cu -o convolutionSeparable.o

Maybe I was missing a flag I need? I am new to CUDA and to C…so trying to learn as go.

Any help would be greatly appreciated.

#2

Use the NVIDIA SDK Makefile to build the project. You can also do fast separable convolutions on GPU with ArrayFire(free GPU library) Documented here

#3

Okay. I think I am just missing a step or something. But when I use the make file I do this:

~/NVIDIA_CUDA_SDK/projects/convolutionSeparable> make ‘convolutionSeparable’

make: *** No rule to make target `convolutionSeparable'.  Stop.

I found this error usually means that the target is already made…so I figured it might be what is in the “obj” directory in that folder…but when I tried to run that I got:

~/NVIDIA_CUDA_SDK/projects/convolutionSeparable/obj/release> ./convolutionSeparable.cu.o

./convolutionSeparable.cu.o: Exec format error. Binary file not executable.

Am I just missing something in the build process to make it executable?

Eris