Hello, I am just beginner in CUDA.
Thank you for reading this post.
Let me introduce my system environment.
x86_64
Fedora release 14 (Laughlin)
Kernel ver 2.6.32.27
# lspci | grep -i nVidia
03:00.0 VGA compatible controller: nVidia Corporation Device 1080 (rev a1)
03:00.1 Audio device: nVidia Corporation Device 0e09 (rev a1)
87:00.0 VGA compatible controller: nVidia Corporation Device 1080 (rev a1)
87:00.1 Audio device: nVidia Corporation Device 0e09 (rev a1)
# gcc --version
gcc (GCC) 4.5.1 20100924 (Red Hat 4.5.1-4)
First, I download below 3 files in NVIDIA Developer Zone.
[b]Developer Drivers for Linux (256.40)
CUDA Toolkit for Fedora 12
GPU Computing SDK code samples[/b]
I installed Developer Driver.
# sh devdriver_3.1_linux_64_256.40.run
And I installed CUDA Toolkit.
# sh cudatoolkit_3.1_linux_64_fedora12.run
And I installed SDK code samples, and try to make them.
# sh gpucomputingsdk_3.1_linux.run
Enter install path (default ~/NVIDIA_GPU_Computing_SDK): ~/CUDA/SDK
...
# cd ~/CUDA/SDK/C
# make
I solved errors with missing Library.
But after that problem, some codes are succeeded in ‘make’ but the others are failed.
The first hitter is ‘SobelFilter’.
SobelFilter_kernels.cu(53): error: calling a host function from a __device__/__global__ function is not allowed
I am try and search, but fail to solve this problem.
For temporary solution, I moved codes that failed in ‘make’ to other directory.
And try make, and another code fail, move that code, and try make again… Finally…
Below are succeed in ‘make’.
BlackScholes deviceQueryDrv matrixMulDrv reduction smokeParticles
FDTD3d dxtc matrixMulDynlinkJIT simpleCUBLAS threadMigration
cppIntegration fluidsGL mergeSort simpleTextureDrv vectorAddDrv
deviceQuery lineOfSight ptxjit simpleVoteIntrinsics
Below are Failed in ‘make’.
FunctionPointers binomialOptions histogram scalarProd simpleTexture3D
Mandelbrot boxFilter imageDenoising scan simpleZeroCopy
MersenneTwister clock marchingCubes simpleAtomicIntrinsics sortingNetworks
MonteCarlo concurrentKernels matrixMul simpleCUFFT template
MonteCarloMultiGPU convolutionFFT2D nbody simpleGL threadFenceReduction
SobelFilter convolutionSeparable oceanFFT simpleMultiCopy transpose
SobolQRNG convolutionTexture particles simpleMultiGPU transposeNew
alignedTypes dct8x8 postProcessGL simplePitchLinearTexture vectorAdd
asyncAPI dwtHaar1D quasirandomGenerator simpleStreams volumeRender
bandwidthTest eigenvalues radixSort simpleTemplates
bicubicTexture fastWalshTransform recursiveGaussian simpleTexture
Is any wrong in my install process? Or do you know how to compile those sample codes?
How can I finish CUDA install and execute SDK sample codes?
I will do polling this post. Thank to read.