Which is faster, the sparse or spdiag function?

In each iteration of a for loop I'm creating a big (approximatlly 10k*10k) five diagonal sparse matrix. Does anybody knows which function work faster? The sparse or spdiags. Is there a faster way to it if I only update the values of an existing sparse matrix in each iteration?

[Walter Roberson]

Be careful, updating an existing sparse matrix is slow if you are adding non-zero elements beyond what was allocated for the matrix with spalloc()

[James Tursa]

Even if you are not exceeding the currently allocated memory for the matrix, it can be slow if the operation causes a large number of existing elements to be moved around in this memory.

Answer 1

[Walter Roberson]

spdiags() invokes sparse() after packaging together the appropriate content and indices.

[d]

So, it means that sparse() is better in terms of time consuming. Thanks.