Exemple #1
0
 def test_scalez_complex_vector(self):
     A = cn.array(self.complex_veca, dtype=cn.complex64)
     B = A.multiply(math.pi+1.37j)
     # numpy
     a = np.array(self.complex_veca, dtype=np.complex64)
     b = np.multiply(a, math.pi+1.37j)
     self.assert_(test.arrays_equal(B.toarray(), b)) 
Exemple #2
0
 def test_scale_real_matrix(self):
     A = cn.array(self.real_mata, dtype=cn.float32)
     B = A.multiply(math.pi)
     # numpy
     a = np.array(self.real_mata, dtype=np.float32)
     b = np.multiply(a, math.pi)
     self.assert_(test.arrays_equal(B.toarray(), b)) 
Exemple #3
0
 def test_complex_single_matrix_svd(self):
     A_ = cn.array(self.complex_mata, dtype=cn.complex64)
     A = np.array(self.complex_mata, dtype=np.complex64)
     # gpu
     U, S, VT = [x.toarray() for x in A_.svd()]
     R = np.dot(U, np.dot(np.diag(S), VT))
     self.assert_(test.arrays_equal(R, A, 1e-03))
Exemple #4
0
 def test_scale_complex_matrix(self):
     A = cn.array(self.complex_mata, dtype=cn.complex64)
     B = A.multiply(math.pi)
     # numpy
     a = np.array(self.complex_mata, dtype=np.complex64)
     b = np.multiply(a, math.pi)
     self.assert_(test.arrays_equal(B.toarray(), b)) 
Exemple #5
0
 def test_real_double_matrix_svd(self):
     A_ = cn.array(self.real_mata, dtype=cn.float64)
     A = np.array(self.real_mata, dtype=np.float64)
     # gpu
     U, S, VT = [x.toarray() for x in A_.svd()]
     R = np.dot(U, np.dot(np.diag(S), VT))
     #print S
     self.assert_(test.arrays_equal(R, A, 1e-03))
 def test_complex_double_matrix_eigensystem(self):
     A_ = cn.array(self.complex_mata, dtype=cn.complex128)
     A = np.array(self.complex_mata, dtype=np.complex128)
     # gpu
     LV, E, RV = [x.toarray() for x in A_.eigensystem(left_vectors=True, right_vectors=True)]
     # cpu
     NE, NLV, NRV = la.eig(A, left=True, right=True)
     self.assert_(test.arrays_equal(E, NE, 1e-03))
Exemple #7
0
 def test_real_single_matrix_mul_matrix(self):
     A_ = cn.array(self.real_mata, dtype=cn.float32)
     A = np.array(self.real_mata, dtype=np.float32)
     # gpu
     R_ = A_.mul(A_)
     # cpu
     R = A * A
     #
     self.assert_(test.arrays_equal(R, R_.toarray(), 1E-03))
Exemple #8
0
 def test_real_double_vector_vector_dot2(self):
     va = cn.array(self.real_veca, dtype=cn.float64)
     vb = cn.array(self.real_vecb, dtype=cn.float64)
     vc = va.dot2(vb)
     # check we get the same answer as numpy
     na = np.array(self.real_veca, dtype=np.float64)
     nb = np.array(self.real_vecb, dtype=np.float64)
     nc = np.dot(na,nb)
     self.assert_(test.arrays_equal(vc.toarray(), nc, epsilon=0.0001))
Exemple #9
0
 def test_real_single_vector_matrix_dot2(self):
     va = cn.array(self.real_veca, dtype=cn.float32)
     vb = cn.array(self.real_matb, dtype=cn.float32)
     vc = va.dot2(vb)
     # check we get the same answer as numpy
     na = np.array(self.real_veca, dtype=np.float32)
     nb = np.array(self.real_matb, dtype=np.float32)
     nc = np.dot(na,nb)
     self.assert_(test.arrays_equal(vc.toarray(), nc, epsilon=0.05))
 def test_real_single_matrix_eigensystem(self):
     A_ = cn.array(self.real_mata, dtype=cn.float32)
     A = np.array(self.real_mata, dtype=np.float32)
     # gpu
     LV, RU, IU, RV = [x.toarray() for x in A_.eigensystem(left_vectors=True, right_vectors=True)]
     E = RU + 1j * IU # glue egienvalues into complex array
     #
     NE, NLV, NRV = la.eig(A, left=True, right=True)
     self.assert_(test.arrays_equal(E, NE, 1e-03))
Exemple #11
0
 def test_complex_double_vector_matrix_dot2(self):
     va = cn.array(self.complex_veca, dtype=cn.complex128)
     vb = cn.array(self.complex_matb, dtype=cn.complex128)
     vc = va.dot2(vb)
     # check we get the same answer as numpy
     na = np.array(self.complex_veca, dtype=np.complex128)
     nb = np.array(self.complex_matb, dtype=np.complex128)
     nc = np.dot(na,nb)
     self.assert_(test.arrays_equal(vc.toarray(), nc, epsilon=0.0001))
Exemple #12
0
 def test_complex_single_vector_vector_dot2(self):
     va = cn.array(self.complex_veca, dtype=cn.complex64)
     vb = cn.array(self.complex_vecb, dtype=cn.complex64)
     vc = va.dot2(vb)
     # check we get the same answer as numpy
     na = np.array(self.complex_veca, dtype=np.complex64)
     nb = np.array(self.complex_vecb, dtype=np.complex64)
     nc = np.dot(na,nb)
     self.assert_(test.arrays_equal(vc.toarray(), nc, epsilon=0.05))
Exemple #13
0
 def test_real_single_mul_vector(self):
     V = np.array(self.real_vector, dtype=np.float32)
     V_ = cn.array(self.real_vector, dtype=cn.float32)
     # gpu
     R_ = V_.mul(V_)
     # cpu
     R = V * V
     #
     self.assert_(test.arrays_equal(R, R_.toarray(), 1E-03))
Exemple #14
0
 def test_real_single_matrix_mul_column_vector(self):
     A_ = cn.array(self.real_mata, dtype=cn.float32)
     A = np.array(self.real_mata, dtype=np.float32)
     V = np.array(self.real_colvec, dtype=np.float32)
     V_ = cn.array(self.real_colvec, dtype=cn.float32)
     # gpu
     R_ = A_.mul(V_)
     # cpu
     R = A * V
     #
     self.assert_(test.arrays_equal(R, R_.toarray(), 1E-03))
Exemple #15
0
 def test_real_single_matrix_cmin(self):
     A_ = cn.array(self.real_mata, dtype=cn.float32)
     A = np.array(self.real_mata, dtype=np.float32)
     # gpu
     s = A_.cmin().toarray()
     #print s
     # cpu
     sn = np.min(A, axis=0)
     #print sn
     # XXX this precision doesn't always return true
     self.assert_(test.arrays_equal(sn, s, 1E-03))
Exemple #16
0
 def test_real_single_matrix_mul_scalar(self):
     A_ = cn.array(self.real_mata, dtype=cn.float32)
     A = np.array(self.real_mata, dtype=np.float32)
     S = 27.345
     # gpu
     s = A_.mul(S)
     #print s
     # cpu
     sn = A * S
     #print sn
     # XXX this precision doesn't always return true
     self.assert_(test.arrays_equal(sn, s.toarray(), 1E-03))
 def test_fn_purity(self):
     A_ = cn.array(self.real_mata, dtype=cn.float32)
     A = np.array(self.real_mata, dtype=np.float32)
     # gpu
     RU, IU = [x.toarray() for x in A_.eigensystem()]
     self.assert_(test.arrays_equal(A, A_.toarray()))
Exemple #18
0
 def test_linalg_transpose_idempotency(self):
     a = cn.array(self.real_mata, dtype=cn.float32)
     b = cn.array(self.real_matb, dtype=cn.float32)
     c = a.dot(b)
     d = b.T.dot(a.T).T
     self.assert_(test.arrays_equal(c.toarray(), d.toarray(), epsilon=0.05))
Exemple #19
0
 def test_matrix_transpose_idempotency(self):
     a = cn.array(self.real_mata, dtype=cn.float32)
     self.assert_(test.arrays_equal(a.toarray(), a.T.T.toarray()))
Exemple #20
0
 def test_vector_transpose_idempotency(self):
     # transpose is a noop for vectors - they are always columns
     v = cn.array(self.real_veca, dtype=cn.float32)
     self.assert_(test.arrays_equal(v.T.toarray(), v.toarray()))
Exemple #21
0
 def test_real_single_matrix_matrix_dot(self):
     na = np.array(self.real_mata, dtype=np.float32)
     nb = np.array(self.real_matb, dtype=np.float32)
     nc = np.dot(na, nb)
     nd = cn.pdot(na, nb)
     self.assert_(test.arrays_equal(nd, nc, epsilon=0.05))
Exemple #22
0
 def test_fn_purity(self):
     A_ = cn.array(self.real_mata, dtype=cn.float32)
     A = np.array(self.real_mata, dtype=np.float32)
     # gpu
     U, S, VT = [x.toarray() for x in A_.svd()]
     self.assert_(test.arrays_equal(A, A_.toarray()))