if keep_it_real:
R = np.zeros_like(T)
else:
R = np.zeros((n, n), dtype=complex)
T_diag = T_diag.astype(complex)
// Take square roots of the diagonal of the triangular matrix.
R[np.diag_indices(n)] = np.sqrt(T_diag)
// Compute the number of blocks to use; use at least one block.
nblocks = max(n // blocksize, 1)
After Change
keep_it_real = (not _has_complex_dtype_char(T)) and (np.min(T_diag) >= 0)
if not keep_it_real:
T_diag = T_diag.astype(complex)
R = np.diag(np.sqrt(T_diag))
// Compute the number of blocks to use; use at least one block.
n, n = T.shape
nblocks = max(n // blocksize, 1)
