niklas/pwr_project
1
0
Fork 0
Code Issues Activity

More work on parallel stuff

Browse Source
This commit is contained in:
Niklas Birk 2024-04-23 17:55:18 +02:00
parent 6848bc7b26
commit 23df9f37f2
5 changed files with 85 additions and 104 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

111
src/cg.py
View File

@ -1,91 +1,58 @@
import matplotlib.pyplot as plt
import numpy as np
from mpi4py import MPI
from matrix_mpi import MatrixMPI as Matrix
from vector_mpi import VectorMPI as Vector
comm = MPI.COMM_WORLD
size = comm.Get_size()
rank = comm.Get_rank()
# Funktion zur Berechnung der 2-Norm
def norm(vec):
sum_of_squares = 0
for i in range(0, len(vec)):
sum_of_squares = sum_of_squares + vec[i] ** 2
return np.sqrt(sum_of_squares)
# Test, ob die Matrix M positiv definit ist, mittels Cholesky-Zerlegung
def ist_positiv_definit(m):
try:
np.linalg.cholesky(m)
return True
except np.linalg.LinAlgError:
return False
n = 1000
h = 1 / (n - 1)
# Initialisierung der Matrix A und des Vektor f für LGS Au = f
A = np.diag(-1 * np.ones(n - 1), k=1) + np.diag(2 * np.ones(n), k=0) + np.diag(-1 * np.ones(n - 1), k=-1)
f = h ** 2 * 2 * np.ones(n)
# Teste, ob A positiv definitv ist mittels Eigenwerten
if np.all(np.linalg.eigvals(A) <= 0): # Prüfen, ob alle Eigenwerte positiv sind
raise ValueError("A ist nicht positiv definit.")
# Teste, ob A positiv definit ist mittels Cholesky-Zerlegung
if not (ist_positiv_definit(A)):
raise ValueError("A ist nicht positiv definit.")
# Teste, ob A symmetrisch ist
if (A != A.T).all():
raise ValueError("A ist nicht symmetrisch.")
def cg(n: int, A: Matrix, f: Vector, tol: float):
# Intialisierung des Startvektors x
x = np.ones(n)
x = Vector([1] * n)
# Toleranz epsilon
eps = 0.001
count = 1
# Anzahl der Schritte
count = 0
# Anfangswerte berechnen
r = f - A @ x # Anfangsresiduum
r = f - A * x # Anfangsresiduum
p = r # Anfangsabstiegsrichtung
print("0. Iterationsschritt: \n")
print("Startvektor:", x)
print("Norm des Anfangsresiduums: ", norm(r))
while r.norm() > tol and count < 1000:
print(f"{count}. Iterationsschritt:\n")
# print("Iterierte:", x)
# print("Residuumsnorm: ", r.norm())
while eps < norm(r) < 1000:
z = A @ p # Matrix-Vektorprodukt berechnen und speichern
z = A * p # Matrix-Vektorprodukt berechnen und speichern
alpha = (np.dot(r, p)) / (np.dot(p, z))
print(alpha)
# Minimiere phi in Richung p um neue Iterierte x zu finden
alpha = (r.T() * p) / (p.T() * z) # (np.dot(r , p)) / (np.dot(p , z))
# print(alpha)
x = x + alpha * p # neue Itterierte x
r = r - alpha * z # neues Residuum
# Bestimmung der neuen Suchrichtung
beta = - (np.dot(r, z)) / (np.dot(p, z))
beta = - (r.T() * z) / (p.T() * z) # (np.dot(r , z)) / (np.dot(p , z))
p = r + beta * p # neue konjugierte Abstiegsrichtung
print(count, ". Iterationsschritt: \n")
print("Aktuelle Iterierte:", x)
print("Norm des Residuums: ", norm(r))
count = count + 1
# Vergleich mit numpy-interner Lsg
u = np.linalg.solve(A, f)
print("Lösung mit CG-Verfahren:", x)
print("Numpy interne Lösung:", u)
if norm(u - x) > eps:
print("Der CG-Algorithmus hat nicht richtig funktioniert!")
else:
print("Der CG-Algorithmus war erfolgreich.")
plt.plot(x, linewidth=2)
plt.plot(u, linewidth=2)
plt.show()
print(f"{rank} APFELSTRUDEL")
# if rank == 0:
# # Vergleich mit numpy-interner Lsg
# u = np.linalg.solve(np.array(A.get_data()), np.array(f.get_data()))
#
# print("Lösung mit CG-Verfahren:", x)
# print("Numpy interne Lösung:", u)
#
# if (Vector(u) - x).norm() > eps:
# print("Der CG-Algorithmus hat nicht richtig funktioniert!")
# else:
# print("Der CG-Algorithmus war erfolgreich.")
#
# plt.plot(x.get_data(), linewidth=2)
# plt.plot(u, linewidth=2)
#
# plt.show()

32
src/main.py
View File

@ -1,26 +1,18 @@
# m1 = MatrixMPI(numpy.random.uniform(0, 1, 1_000_000), (1000, 1000))
import numpy as np
from mpi4py import MPI
from matrix_mpi import MatrixMPI
from vector_mpi import VectorMPI
import cg
comm = MPI.COMM_WORLD
rank = comm.Get_rank()
size = comm.Get_size()
from matrix_mpi import MatrixMPI as Matrix
from vector_mpi import VectorMPI as Vector
m1 = MatrixMPI(list(range(1, 21)), (4, 5))
m2 = MatrixMPI(list(range(1, 16)), (5, 3))
n = 1_00
h = 1 / (n - 1)
m_mul = m1 * m2
# Initialisierung der Matrix A und des Vektor f für LGS Au = f
A = Matrix(np.diag(-1 * np.ones(n - 1), k=1) + np.diag(2 * np.ones(n), k=0) + np.diag(-1 * np.ones(n - 1), k=-1))
f = Vector([h ** 2 * 2] * n)
v1 = VectorMPI(list(range(1, 21)))
v2 = VectorMPI(list(reversed(list(range(1, 21)))))
# Toleranz epsilon
tol = 0.001
v_add = v1 + v2
v_mul = v1.T() * v2
if rank == 0:
print(m_mul)
print("---")
print(v_add)
print(v_mul)
cg.cg(n, A, f, tol)

2
src/matrix.py
View File

@ -98,7 +98,7 @@ class Matrix:
flattened_data = []
rows = 0
for matrix in matrices:
flattened_data.extend(matrix.get_data())
flattened_data.extend(matrix.get_matrix())
rows += matrix.__shape__[0]
cols = matrices[0].__shape__[1]
return Matrix(flattened_data, (rows, cols))

15
src/matrix_mpi.py
View File

@ -34,9 +34,20 @@ class MatrixMPI:
cols = self.__data__.shape()[1]
return Matrix(self.__data__[self.__chunk__], (rows, cols))
def get_data(self):
def get_matrix(self):
"""
Returns the ``Matrix`` that is used internally
:return: The ``Matrix`` that is used internally
"""
return self.__data__
def get_data(self):
"""
Returns the raw data of the internal data structure
:return: The raw data of the internal data structure
"""
return self.__data__.get_data()
def transpose(self):
"""
:return: the transpose of the matrix
@ -91,7 +102,7 @@ class MatrixMPI:
def __mul__(self, other):
if isinstance(other, MatrixMPI):
other = other.get_data()
other = other.get_matrix()
gathered_data = self.__mpi_comm__.gather(self.get_rank_submatrix() * other)
data = self.__mpi_comm__.bcast(gathered_data)
return MatrixMPI.of(Matrix.flatten(data))

13
src/vector_mpi.py
View File

@ -13,8 +13,19 @@ class VectorMPI(MatrixMPI):
return VectorMPI(vector.get_data(), vector.shape())
def get_vector(self):
"""
Returns the ``Vector`` that is used internally
:return: The ``Vector`` that is used internally
"""
return self.__data__
def get_data(self):
"""
Returns the raw data of the internal data structure
:return: The raw data of the internal data structure
"""
return self.__data__.get_data()
def shape(self):
return self.__data__.shape()
@ -55,7 +66,7 @@ class VectorMPI(MatrixMPI):
def __rmul__(self, other):
if isinstance(other, MatrixMPI):
return VectorMPI.of(other.get_data() * self.get_vector())
return VectorMPI.of(other.get_matrix() * self.get_vector())
return self * other
def __truediv__(self, other):