Illegal address error

The following code gives an IllegalAddressError when executing on a GPU. The commented code is my workaround. I’d love to know what the specific reason behind this error is as the description doesn’t give much explanation.

max_GPU = 2048
blk_size = 128

function [] = __device__ fsub_col_unit(A: mat'const'unchecked, b: vec'unchecked, i_start:int, j_start: int, n: int)
    for j = 0..n-2
        for i = j+1..n-1
            b[i_start+i] -= A[i_start+i,j_start+j]*b[i_start+j]
        end    
    end
end

function [] = __kernel__ trsv(A: mat'unchecked, i_start:int, j_start: int, k: int, jb: int, pos: ivec1)
    temp = A[:, j_start+k+jb+pos]
    fsub_col_unit(A, temp, i_start, j_start, jb)
    A[:, j_start+k+jb+pos] = temp
end

function [] = f(A: mat'unchecked, i_start:int, j_start: int, cols: int)
%    if cols-blk_size > max_GPU
%        for k = 0..max_GPU..cols-blk_size-1
%            jb2 = min(max_GPU, cols-blk_size-k)
%            parallel_do(jb2, A, i_start, j_start,k, blk_size, trsv)
%        end
%    else
        parallel_do(cols-blk_size, A, 0, 0, 0, blk_size, trsv)
%    endif      
end

dim = 2200
A = rand(dim,dim)

f(A, 0,0,dim)