Floating Point Exception

Accelerated Computing CUDA CUDA Programming and Performance
1

Hey all,

I have 2 8800GTX cards in 2 different machines, with identical hardware and running slightly different OS.

When trying to run the SDK examples on one machine, it runs without problems.

On the other machine, it only gives Floating Point exceptions:

$ ./matrixMul

Floating point exception

$

further information:

On both machines, i succesfully installed the latest nvidia drivers (NVIDIA-Linux-x86-180.22-pkg1.run), Cuda 2.1(cudatoolkit_2.1_linux32_rhel5.2.run) and all required packages for installing

Both machines succesfully compiled the SDK examples.

The working machine has a custom 2.6.18 i686 debian Sarge kernel

The faulty machine has a standard debian kernel, 2.6.24-etchnhalf.1-686.

Can anyone give me pointers in what is going wrong here?

thnx,

Tijs

2

I have seen this exception with CUDA binaries I have compiled and moved between Linux distributions that differ too much in their system libraries. I’ve never seen it though with a binary compiled on the same computer it was being executed on, though.