def test_get_cntl(self):
ma57 = MA57Interface()
self.assertEqual(ma57.get_icntl(1), 6)
self.assertEqual(ma57.get_icntl(7), 1)
self.assertAlmostEqual(ma57.get_cntl(1),
1e-2) # Numerical pivot threshold
self.assertAlmostEqual(ma57.get_cntl(2), 1e-20) # Null pivot threshold
def test_set_icntl(self):
ma57 = MA57Interface()
ma57.set_icntl(5, 4) # Set output printing to max verbosity
ma57.set_icntl(8, 1) # Keep factors when we run out of space
# (so MA57ED can be used)
icntl5 = ma57.get_icntl(5)
icntl8 = ma57.get_icntl(8)
self.assertEqual(icntl5, 4)
self.assertEqual(icntl8, 1)
with self.assertRaisesRegex(TypeError, 'must be an integer'):
ma57.set_icntl(1.0, 0)
with self.assertRaisesRegex(IndexError, 'is out of range'):
ma57.set_icntl(100, 0)
with self.assertRaises(ctypes.ArgumentError):
ma57.set_icntl(1, 0.0)
def test_do_numeric_factorization(self):
ma57 = MA57Interface()
n = 5
ne = 7
irn = np.array([1, 1, 2, 2, 3, 3, 5], dtype=np.intc)
jcn = np.array([1, 2, 3, 5, 3, 4, 5], dtype=np.intc)
irn = irn - 1
jcn = jcn - 1
ent = np.array([2., 3., 4., 6., 1., 5., 1.], dtype=np.double)
ma57.do_symbolic_factorization(n, irn, jcn)
ma57.fact_factor = 1.5
ma57.ifact_factor = 1.5
# ^ No way to check whether these are handled properly... Would have to
# access the struct to get LFACT, LIFACT
status = ma57.do_numeric_factorization(n, ent)
self.assertEqual(status, 0)
self.assertEqual(ma57.get_info(14), 12) # 12 entries in factors
self.assertEqual(ma57.get_info(24), 2) # 2 negative eigenvalues
self.assertEqual(ma57.get_info(22), 1) # 1 2x2 pivot
self.assertEqual(ma57.get_info(23), 0) # 0 delayed pivots
ent2 = np.array([1., 5., 1., 6., 4., 3., 2.], dtype=np.double)
ma57.do_numeric_factorization(n, ent2)
self.assertEqual(status, 0)
bad_ent = np.array([2., 3., 4., 6., 1., 5.], dtype=np.double)
with self.assertRaisesRegex(AssertionError, 'Wrong number of entries'):
ma57.do_numeric_factorization(n, bad_ent)
with self.assertRaisesRegex(AssertionError, 'Dimension mismatch'):
ma57.do_numeric_factorization(n + 1, ent)
n = 5
ne = 8
irn = np.array([1, 1, 2, 2, 3, 3, 5, 5], dtype=np.intc)
jcn = np.array([1, 2, 3, 5, 3, 4, 5, 1], dtype=np.intc)
irn = irn - 1
jcn = jcn - 1
ent = np.array([2., 3., 4., 6., 1., 5., 1., -1.3], dtype=np.double)
status = ma57.do_symbolic_factorization(n, irn, jcn)
self.assertEqual(status, 0)
status = ma57.do_numeric_factorization(n, ent)
self.assertEqual(status, 0)
self.assertEqual(ma57.get_info(24), 2)
self.assertEqual(ma57.get_info(23), 0)
def test_do_backsolve(self):
ma57 = MA57Interface()
n = 5
ne = 7
irn = np.array([1, 1, 2, 2, 3, 3, 5], dtype=np.intc)
jcn = np.array([1, 2, 3, 5, 3, 4, 5], dtype=np.intc)
irn = irn - 1
jcn = jcn - 1
ent = np.array([2., 3., 4., 6., 1., 5., 1.], dtype=np.double)
rhs = np.array([8., 45., 31., 15., 17.], dtype=np.double)
status = ma57.do_symbolic_factorization(n, irn, jcn)
status = ma57.do_numeric_factorization(n, ent)
sol = ma57.do_backsolve(rhs)
expected_sol = [1, 2, 3, 4, 5]
old_rhs = np.array([8., 45., 31., 15., 17.])
for i in range(n):
self.assertAlmostEqual(sol[i], expected_sol[i])
self.assertEqual(old_rhs[i], rhs[i])
def test_do_symbolic_factorization(self):
ma57 = MA57Interface()
n = 5
ne = 7
irn = np.array([1, 1, 2, 2, 3, 3, 5], dtype=np.intc)
jcn = np.array([1, 2, 3, 5, 3, 4, 5], dtype=np.intc)
# Copied these Fortran-style indices from HSL docs.
# Interface expects C-style indices, as is typical in Python.
irn = irn - 1
jcn = jcn - 1
bad_jcn = np.array([1, 2, 3, 5, 3, 4], dtype=np.intc)
ma57.do_symbolic_factorization(n, irn, jcn)
self.assertEqual(ma57.get_info(1), 0)
self.assertEqual(ma57.get_info(4), 0)
self.assertEqual(ma57.get_info(9), 48) # Min required length of FACT
self.assertEqual(ma57.get_info(10), 53) # Min required length of IFACT
self.assertEqual(ma57.get_info(14), 0) # Should not yet be set
with self.assertRaisesRegex(AssertionError, 'Dimension mismatch'):
ma57.do_symbolic_factorization(n, irn, bad_jcn)
class TestLinearSolvers(unittest.TestCase):
def create_blocks(self, m: np.ndarray, x: np.ndarray):
m = coo_matrix(m)
r = m * x
bm = BlockMatrix(2, 2)
bm.set_block(0, 0, m.copy())
bm.set_block(1, 1, m.copy())
br = BlockVector(2)
br.set_block(0, r.copy())
br.set_block(1, r.copy())
bx = BlockVector(2)
bx.set_block(0, x.copy())
bx.set_block(1, x.copy())
return bm, bx, br
def solve_helper(self, m: np.ndarray, x: np.ndarray, solver: LinearSolverInterface):
bm, bx, br = self.create_blocks(m, x)
bx2, res = solver.solve(bm, br)
self.assertEqual(res.status, LinearSolverStatus.successful)
err = np.max(np.abs(bx - bx2))
self.assertAlmostEqual(err, 0)
def symmetric_helper(self, solver: LinearSolverInterface):
m = np.array([[1, 2], [2, -1]], dtype=np.double)
x = np.array([4, 7], dtype=np.double)
self.solve_helper(m, x, solver)
def unsymmetric_helper(self, solver: LinearSolverInterface):
m = np.array([[1, 2], [0, -1]], dtype=np.double)
x = np.array([4, 7], dtype=np.double)
self.solve_helper(m, x, solver)
def singular_helper(self, solver: LinearSolverInterface):
m = np.array([[1, 1], [1, 1]], dtype=np.double)
x = np.array([4, 7], dtype=np.double)
bm, bx, br = self.create_blocks(m, x)
br.get_block(0)[1] += 1
br.get_block(1)[1] += 1
bx2, res = solver.solve(bm, br, raise_on_error=False)
self.assertNotEqual(res.status, LinearSolverStatus.successful)
@unittest.skipIf(not MA27Interface.available(), reason="MA27 not available")
def test_ma27(self):
solver = MA27()
self.symmetric_helper(solver)
self.singular_helper(solver)
@unittest.skipIf(not MA57Interface.available(), reason="MA57 not available")
def test_ma57(self):
solver = MA57()
self.symmetric_helper(solver)
self.singular_helper(solver)
def test_scipy_direct(self):
solver = ScipyLU()
self.symmetric_helper(solver)
self.unsymmetric_helper(solver)
self.singular_helper(solver)
def test_scipy_iterative(self):
solver = ScipyIterative(gmres)
solver.options["atol"] = 1e-8
self.symmetric_helper(solver)
self.unsymmetric_helper(solver)
self.singular_helper(solver)
@unittest.skipIf(not mumps_available, reason="mumps not available")
def test_mumps(self):
solver = MumpsCentralizedAssembledLinearSolver(sym=2)
self.symmetric_helper(solver)
self.singular_helper(solver)
solver = MumpsCentralizedAssembledLinearSolver(sym=0)
self.unsymmetric_helper(solver)
def test_set_cntl(self):
ma57 = MA57Interface()
ma57.set_cntl(1, 1e-8)
ma57.set_cntl(2, 1e-12)
self.assertAlmostEqual(ma57.get_cntl(1), 1e-8)
self.assertAlmostEqual(ma57.get_cntl(2), 1e-12)
# Copyright (c) 2008-2022
# National Technology and Engineering Solutions of Sandia, LLC
# Under the terms of Contract DE-NA0003525 with National Technology and
# Engineering Solutions of Sandia, LLC, the U.S. Government retains certain
# rights in this software.
# This software is distributed under the 3-clause BSD License.
# ___________________________________________________________________________
import ctypes
import pyomo.common.unittest as unittest
from pyomo.contrib.pynumero.dependencies import numpy as np, numpy_available
if not numpy_available:
raise unittest.SkipTest('pynumero MA27 tests require numpy')
from pyomo.contrib.pynumero.linalg.ma57 import MA57Interface
@unittest.skipIf(not MA57Interface.available(), reason='MA57 not available')
class TestMA57Interface(unittest.TestCase):
def test_get_cntl(self):
ma57 = MA57Interface()
self.assertEqual(ma57.get_icntl(1), 6)
self.assertEqual(ma57.get_icntl(7), 1)
self.assertAlmostEqual(ma57.get_cntl(1),
1e-2) # Numerical pivot threshold
self.assertAlmostEqual(ma57.get_cntl(2), 1e-20) # Null pivot threshold
def test_set_icntl(self):
ma57 = MA57Interface()
ma57.set_icntl(5, 4) # Set output printing to max verbosity
ma57.set_icntl(8, 1) # Keep factors when we run out of space
# (so MA57ED can be used)