Problem with examples most of them crash

Accelerated Computing CUDA CUDA Programming and Performance
#1

Hello!

I just installed CUDA and everything with the latest drivers… but when i run some of the example they crash… I have a geforce 8600

let’s start with the first:

mascarpone@mascarpone-desktop:~/NVIDIA_CUDA_SDK/bin/linux/release$ ./alignedTypes 

Using device 0: GeForce 8400 GS

Allocating memory...

Generating host input data array...

Uploading input data to GPU memory...

Testing misaligned types...

uint8...

Avg. time: 466.200165 ms / Copy throughput: 0.099884 GB/s.

TEST FAILED

uint16...

Avg. time: 0.008313 ms / Copy throughput: 5601.940243 GB/s.

TEST FAILED

RGBA8_misaligned...

Avg. time: 0.008156 ms / Copy throughput: 5709.257001 GB/s.

TEST FAILED

LA32_misaligned...

Avg. time: 0.008156 ms / Copy throughput: 5709.257001 GB/s.

TEST FAILED

RGB32_misaligned...

Avg. time: 0.008125 ms / Copy throughput: 5731.215137 GB/s.

TEST FAILED

RGBA32_misaligned...

Avg. time: 0.008219 ms / Copy throughput: 5665.840512 GB/s.

TEST FAILED

Testing aligned types...

RGBA8...

Avg. time: 0.008219 ms / Copy throughput: 5665.840512 GB/s.

TEST FAILED

I32...

Avg. time: 0.008187 ms / Copy throughput: 5687.466224 GB/s.

TEST FAILED

LA32...

Avg. time: 0.008219 ms / Copy throughput: 5665.840512 GB/s.

TEST FAILED

RGB32...

Avg. time: 0.008250 ms / Copy throughput: 5644.379271 GB/s.

TEST FAILED

RGBA32...

Avg. time: 0.008250 ms / Copy throughput: 5644.379271 GB/s.

TEST FAILED

RGBA32_2...

Avg. time: 0.008250 ms / Copy throughput: 5644.379271 GB/s.

TEST FAILED

Shutting down...

Press ENTER to exit...
#2

hey guys,

i just wrote a script so if you have to launch all examples you won’t loose too much time… here it is

echo "Starting commands..."

./alignedTypes

./asyncAPI

./bandwidthTest

./bicubicTexture

./binomialOptions

./bitonic

./BlackScholes

./boxFilter

./clock

./convolutionFFT2D

./convolutionSeparable

./convolutionTexture

./cppIntegration

./dct8x8

./deviceQuery

./dwtHaar1D

./dxtc

./eigenvalues

./fastWalshTransform

./fluidsGL

./histogram64

./histogram256

./imageDenoising

./lineOfSight

./Mandelbrot

./marchingCubes

./matrixMul

./matrixMulDrv

./MersenneTwister

./MonteCarlo

./MonteCarloMultiGPU

./nbody

./oceanFFT

./particles

./postProcessGL

./quasirandomGenerator

./recursiveGaussian

./reduction

./scalarProd

./scan

./scanLargeArray

./simpleAtomicIntrinsics

./simpleCUBLAS

./simpleCUFFT

./simpleGL

./simpleMultiGPU

./simpleStreams

./simpleTemplates

./simpleTexture

./simpleTexture3D

./simpleTextureDrv

./simpleVoteIntrinsics

./SobelFilter

./template

./threadMigration

./transpose

./volumeRender

echo "Done!"

copy it in your example folder, save it as myscript, and then write in your console:

sh myscript

here you go!

#3

Or just enter example directory and type:

for one in `ls`; do ./$one; done

Isn’t it simplier? :)

#4

The program isn’t actually crashing. On the other hand, it isn’t exactly behaving as expected either.

I can confirm that I see something similar here. It fails when I run from a terminal within gnome but works with a minimal setup as in: startx xterm – :1

From the console it fails consistently for ordinary users but sometimes works for root. When it fails, I get an NVRM Xid: (0003:00): 13

Premature termination with Ctrl-C makes the X server start behaving weirdly, dropping pixel noise all over the place and rendering OpenGL in noisy 16x16 chunks.

This is with driver 177.73 on a g98

#5

ehhehe thank you! It was my first SH program… eehheheh… still much to learn!

#6

Did you tweak driver? I experienced similar problem when I add parameter NVreg_UsePageAttributeTable=0 to nvidia kernel module. Graphics run a bit faster with MTRR instead of with PAT, but CUDA becomes completely unusable.

#7

hi!

no at all! As I said i used the default options since I’m a noob… If no input will be given in the next few days I’ll try to make a new install and see how it goes…

thank you anyhow!