def idd_lssolve(queue, m, n, a, krank):
for j in range(n - krank):
blas.trsv(queue, a[:krank,:krank], a[:krank,krank+j],lower=False)
ctx = queue.get_info(cl.command_queue_info.CONTEXT)
prg = cl.Program(ctx, util.get_source('id_kerns.cl')).build()
prg.moveup(queue, [krank, n-krank], None, a.data, np.int32(krank), np.int32(n))
A = np.tril(A)
# allocate OpenCL memory on the device
clA = Array(queue, A.shape, A.dtype)
clA_upper = Array(queue, A.shape, A.dtype)
clx = Array(queue, x.shape, x.dtype)
clx1 = Array(queue, x1.shape, x1.dtype)
clx2 = Array(queue, x2.shape, x2.dtype)
# copy data to device
clA.set(A)
clA_upper.set(A_upper)
clx.set(x)
# compute a triangular solve (trsv)
blas.trsv(queue, clA, clx)
# check the result
print("Expected: ", np.linalg.solve(A, x))
print("Actual: ", clx.get())
print()
# try a triangular solve with the transpose
clx1.set(x1)
blas.trsv(queue, clA, clx1, transA=True)
print("Expected: ", np.linalg.solve(A.T, x1))
print("Actual: ", clx1.get())
print()
# trye an upper triangular solve
clx2.set(x2)
