CUDA examples don't work

[ramarromarrone]

Hi all,

i compiled sdk examples but some of them doesn’t work.

i receive this errors:

./fluidsGL 

[fluidsGL] - [OpenGL/CUDA simulation]

CUDA device [GeForce GTS 250] has 16 Multi-Processors

fluidsGL_kernels.cu(29) : cutilCheckMsg() CUTIL CUDA error : cudaMalloc failed : invalid argument.

./smokeParticles 

[ CUDA Smoke Particles ]

Loaded './../../../src/smokeParticles/data/floortile.ppm', 256 x 256 pixels

ParticleSystem.cu(84) : cudaSafeCall() Runtime API error : invalid argument.

./postProcessGL 

main.cpp(269) : cudaSafeCall() Runtime API error : invalid argument

./imageDenoising 

[CUDA ImageDenoising] 

Allocating host and CUDA memory and loading image file...

Loading ./../../../src/imageDenoising/data/portrait_noise.bmp...

BMP width: 320

BMP height: 408

BMP file loaded successfully!

Data init done.

Initializing GLUT...

OpenGL window created.

Loading extensions: No error

imageDenoisingGL.cpp(580) : cudaSafeCall() Runtime API error : invalid resource handle.

my system is:

GeForce GTS 250 1GB

Ubuntu 10.04 64bit, cuda 3.1.1

thanks a lot

[ramarromarrone]

Hi all,

i compiled sdk examples but some of them doesn’t work.

i receive this errors:

./fluidsGL 

[fluidsGL] - [OpenGL/CUDA simulation]

CUDA device [GeForce GTS 250] has 16 Multi-Processors

fluidsGL_kernels.cu(29) : cutilCheckMsg() CUTIL CUDA error : cudaMalloc failed : invalid argument.

./smokeParticles 

[ CUDA Smoke Particles ]

Loaded './../../../src/smokeParticles/data/floortile.ppm', 256 x 256 pixels

ParticleSystem.cu(84) : cudaSafeCall() Runtime API error : invalid argument.

./postProcessGL 

main.cpp(269) : cudaSafeCall() Runtime API error : invalid argument

./imageDenoising 

[CUDA ImageDenoising] 

Allocating host and CUDA memory and loading image file...

Loading ./../../../src/imageDenoising/data/portrait_noise.bmp...

BMP width: 320

BMP height: 408

BMP file loaded successfully!

Data init done.

Initializing GLUT...

OpenGL window created.

Loading extensions: No error

imageDenoisingGL.cpp(580) : cudaSafeCall() Runtime API error : invalid resource handle.

my system is:

GeForce GTS 250 1GB

Ubuntu 10.04 64bit, cuda 3.1.1

thanks a lot

[sparvu]

Hi all,

i compiled sdk examples but some of them doesn’t work.

i receive this errors:

[code]./fluidsGL

[code]./smokeParticles

[code]./postProcessGL

[code] ./imageDenoising

my system is:

GeForce GTS 250 1GB

Ubuntu 10.04 64bit, cuda 3.1.1

thanks a lot

what deviceQuery says when you try to run it ? Im on Ubuntu 10.04 64bit and works fine for me.

I do a have a Fermi GPU versus your GPU but not sure if this is a OpenGL issue or a CUDA.

Have you tried to run:

http://developer.nvidia.com/object/cuda_3_…t_rc.html#Linux ?

stefan

[sparvu]

Hi all,

i compiled sdk examples but some of them doesn’t work.

i receive this errors:

[code]./fluidsGL

[code]./smokeParticles

[code]./postProcessGL

[code] ./imageDenoising

my system is:

GeForce GTS 250 1GB

Ubuntu 10.04 64bit, cuda 3.1.1

thanks a lot

what deviceQuery says when you try to run it ? Im on Ubuntu 10.04 64bit and works fine for me.

I do a have a Fermi GPU versus your GPU but not sure if this is a OpenGL issue or a CUDA.

Have you tried to run:

http://developer.nvidia.com/object/cuda_3_…t_rc.html#Linux ?

stefan

[ramarromarrone]

what deviceQuery says when you try to run it ? Im on Ubuntu 10.04 64bit and works fine for me.

I do a have a Fermi GPU versus your GPU but not sure if this is a OpenGL issue or a CUDA.

Have you tried to run:

http://developer.nvidia.com/object/cuda_3_…t_rc.html#Linux ?

stefan

uhmm i thought it was an opengl issue but actually deviceQuery doesn’t work too.

i will re-install driver, toolkit and sdk i think it’s the only way

[ramarromarrone]

what deviceQuery says when you try to run it ? Im on Ubuntu 10.04 64bit and works fine for me.

I do a have a Fermi GPU versus your GPU but not sure if this is a OpenGL issue or a CUDA.

Have you tried to run:

http://developer.nvidia.com/object/cuda_3_…t_rc.html#Linux ?

stefan

uhmm i thought it was an opengl issue but actually deviceQuery doesn’t work too.

i will re-install driver, toolkit and sdk i think it’s the only way

[nikhil.stephen]

i have same probem.using GeForce9300M GS with 8 cuda cores 512 mb memory
On ubuntu 10.04.
Is their no way other than reinstalling driver ,toolkit and sdk???

[nikhil.stephen]

i have same probem.using GeForce9300M GS with 8 cuda cores 512 mb memory
On ubuntu 10.04.
Is their no way other than reinstalling driver ,toolkit and sdk???

[Tomasz_Rybak]

i have same probem.using GeForce9300M GS with 8 cuda cores 512 mb memory

On ubuntu 10.04.

Is their no way other than reinstalling driver ,toolkit and sdk???

If you have mixed versions of drivers and CUDA library, your code will not work.

You need to be sure that you CUDA version matches drivers version (3.1 for 256.x, 3.2 for 260.x, etc.) and that you do not have any residual files from old drivers and/or libraries. One of problems might be caused by ldconfig - tool that manages loading libraries for programs. It has cache of all installed libraries and uses it to decide which file to link with which program. If this cache contains information about old libraries, you program will not work or will fail with strange effects.

So I am affraid that you need to remove old drivers and libraries and install new ones - else you might still have problems with running CUDA code.

[Tomasz_Rybak]

i have same probem.using GeForce9300M GS with 8 cuda cores 512 mb memory

On ubuntu 10.04.

Is their no way other than reinstalling driver ,toolkit and sdk???

If you have mixed versions of drivers and CUDA library, your code will not work.

You need to be sure that you CUDA version matches drivers version (3.1 for 256.x, 3.2 for 260.x, etc.) and that you do not have any residual files from old drivers and/or libraries. One of problems might be caused by ldconfig - tool that manages loading libraries for programs. It has cache of all installed libraries and uses it to decide which file to link with which program. If this cache contains information about old libraries, you program will not work or will fail with strange effects.

So I am affraid that you need to remove old drivers and libraries and install new ones - else you might still have problems with running CUDA code.