How can I debug the error "The CUDA error code was: CUDA_ERROR_UNKNOWN."

I'm using a new Geforce GTX 1080 Ti, replacing a Quadro 5000 which previously worked with the same code.

The problem occurs when I try to conduct math on a vector of gpuArray data. For example X = rand(1000); G = gpuArray(X); G + 2

This will return the following error message:

Error using gpuArray/subsref An error occurred during PTX compilation of image. The information log was:

The error log was:

The CUDA error code was: CUDA_ERROR_UNKNOWN.

Error in dispInternal>iTransferPortionDense (line 36) data = gather( subsref( x, s ) );

Error in parallel.internal.shared.buildDisplayHelper>iFirstNNumericDisplayHelper (line 71) maybeTruncatedValue = transferDenseFcn( x, rangeStruct );

Error in parallel.internal.shared.buildDisplayHelper>iBuildDisplayHelper (line 33) dh = iFirstNNumericDisplayHelper( ...

Error in parallel.internal.shared.buildDisplayHelper (line 24) dh = iBuildDisplayHelper( x, transferDenseFcn, transferSparseFcn, xClassName, xName, N );

Error in dispInternal (line 13) dh = parallel.internal.shared.buildDisplayHelper( ...

Error in gpuArray/display (line 21) dh = dispInternal( obj, thisClassName, objName );

Many thanks for any help you can provide!

Answer 1

[Jason Cassibry]

I solved my own problem. I am using Ubuntu linux, and if you run matlab from a terminal as a super user

sudo matlab

it fixes the problem. Everything is up and running fine!

[Joss Knight]

Thanks for that; it's interesting that permissions should have such an effect, and certainly seems like a bug with the GPU driver. Unfortunately there seem to be a lot of problems with the 1080 Ti; you should remember that these GeForce cards are designed for high performance graphics and their usefulness for compute has its limitations; the main one seeming to be their tendency to just crash without explanation, particularly when memory usage is being stretched.