interp1 gpuArray bug

There is a bug in interp1 for gpuArrays that happens with the 2 input option. It works fine with 3 inputs.

The below works fine yielding the 2x2 array below

interp1((1:80)', randn(80, 2), gpuArray(((1:2))'))
ans =
   -0.2063    0.2255
    0.2836   -0.7136

Now if calling the same code but with only two inputs, where the first input should be implicitly (1:80)', we get the below. I have tried this call on 2 different computers.

interp1(randn(80, 2), gpuArray(((1:2))'))
Error using gpuArray/gather
An unexpected error occurred during CUDA execution. The CUDA error was:
CUDA_ERROR_ILLEGAL_ADDRESS
Error in parallel.internal.shared.buildDisplayHelper>iFirstNNumericDisplayHelper (line 73)
        maybeTruncatedValue = gather( x );
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 ); 

The gpu I'm using is:

                      Name: 'GeForce GTX 1060'
                     Index: 1
         ComputeCapability: '6.1'
            SupportsDouble: 1
             DriverVersion: 10
            ToolkitVersion: 9.1000
            

Does anyone have this same problem and/or advice on how to fix it?

[Walter Roberson]

I confirm on my GTX750M

[Edric Ellis]

Thanks for reporting this. I can confirm that I can reproduce the problem here, I'll pass this on to the relevant development group.

[Kyle]

Is there any update on this?

Answer 1

[Matt J]

The bug appears to have been fixed in R2019b:

>> A=rand(80,2);
>> interp1((1:80)', A, gpuArray(((1:2))'))

ans =

    0.6073    0.3774
    0.4501    0.2160

>> interp1(A, gpuArray(((1:2))'))

ans =

    0.6073    0.3774
    0.4501    0.2160

[Kyle]

ok great, thank you

[Matt J]

You're welcome, but please Accept-click the answer to indicate that you are satisfied with it.