Error using gpuArray/arrayfun ... Use of functional workspace is not supported.

Question

Soohun 2013-12-26

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/110787-error-using-gpuarray-arrayfun-use-of-functional-workspace-is-not-supported

评论： LASIN T 2022-8-10

I define a simple function of summation.m as follows:

--------------------------------------------

function out = summation(a, b, c, d)
out = a+b+c+d;

--------------------------------------------

Then, I run the following code:

--------------------------------------------

R1 = gpuArray.rand(10,10);
R2 = gpuArray.rand(10,10);
R3 = gpuArray.rand(10,10);
f = @(a,b,c)summation( a, b, c, 5);
R = arrayfun(f,R1,R2,R3);

--------------------------------------------

This works fine. However, if I change the function handle as

--------------------------------------------

R1 = gpuArray.rand(10,10);
R2 = gpuArray.rand(10,10);
R3 = gpuArray.rand(10,10);
*_d = 5;_*
f = @(a,b,c)summation( a, b, c, *_d_* );
R = arrayfun(f,R1,R2,R3);

--------------------------------------------

I get the error message "Error using gpuArray/arrayfun Use of functional workspace is not supported." Is there a way to resolve this issue? I would like to resolve this so that I can use GPU in the case that d is either a vector or a matrix.

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Jill Reese 2013-12-30

1
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/110787-error-using-gpuarray-arrayfun-use-of-functional-workspace-is-not-supported#answer_119674

As of the latest MATLAB release (R2013b) arrayfun on the GPU does not support anonymous functions that have been created using variables from the existing MATLAB workspace. This is the meaning of the error message you have encountered.

Unless I am missing something, you should not need to create your function handle using your variable, d. Arrayfun on the GPU supports scalar expansion so d can be a scalar, a vector, or a matrix as long as its dimensions are either the same as your R1, R2, and R3 variables or the dimension is 1. So for the dimensions you mention above, d can be passed into the arrayfun call in the same way as R1, R2, and R3 and the call will succeed if d has size 1 x 1, 10 x 1, 1 x 10, or 10 x 10. Does this not work for you?