PORR: zero findings, 18 tests still passing after every refactor

598 findings to none, and the suite ran through `./run.sh run` after each step
rather than at the end -- 18 passed, 27s, every time.

The shape of this one was different from the other courses. PORR is four
implementations of the same linear algebra (sequential, threads, processes,
dask), so nearly every finding came in groups of four:

- 229 missing annotations. Rather than `object`, which mypy strict rejects on
  every attribute access, each module declares what these actually are:
  `Matrix = np.ndarray | list[list[float]]` and the vector equivalent. The
  backends really are called with both -- comparing them is the point of the
  course.
- 76 names: `A` -> matrix, `Ab` -> augmented, `Ax` -> product, `I` ->
  identity, `dA` -> dask_matrix, and the two CamelCase functions. Renamed
  through `tokenize`, not regex, so no string or comment was touched.
- 121 missing docstrings, one per method name since the names repeat across
  all four backends.
- Seven functions over the argument limit, all taking the same seven or nine
  loose values. They are two frozen dataclasses now, in a new
  richardson_problem.py: `Problem(matrix, rhs, lambda_min, lambda_max)` and
  `Settings(max_iterations, x0, tol)`. `RichardsonMethod.__init__` also took a
  `size` argument it never read.
- threads_indep_old's iteration did "split into chunks, submit, wait" four
  times by hand (C901 11, PLR0915 54). One `_run_chunked` helper.
- tests.py's two bare asserts became explicit raises, per the no-ignores rule.
  `assert False` was also B011: `python -O` deletes it, so the non-convergence
  branch would have silently passed under optimisation.

Real fixes rather than style: np.random -> Generator, `os.path.exists` ->
Path, an `except Exception` around dask narrowed to the errors it actually
raises, and two runnable modules that logged at INFO without ever calling
basicConfig -- caught by scripts/check_refactor_safety.py, which is now doing
the job it was written for.

Repo-wide: 1,781 findings -> 1,183, all of them in the five courses left.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0172dj34FDpao6wsix8ZuCzE

Programming/NLP/processing.py

@@ -1,9 +1,9 @@
 #!/usr/bin/env python3
 """Load the SemEval corpus: sentences, chunkings and gold alignments."""
 
-import sys
 import logging
 import re
+import sys
 from pathlib import Path
 
 import numpy as np

Programming/PORR/code/linear_algebra_utils.py

@@ -1,121 +1,154 @@
+#!/usr/bin/env python3
+"""Four linear-algebra backends with the same interface.
+
+Sequential, threads, processes and dask distributed arrays -- comparing them
+is what this course is about.
+"""
+
 import cmath
 import itertools
 import logging
 import math
 import multiprocessing
 import operator
-from abc import ABC, abstractmethod
+from abc import abstractmethod
 from concurrent.futures import ThreadPoolExecutor
 from functools import partial
 from multiprocessing import Pool
 
 import dask.array as da
+import numpy as np
 from time_measurement import time_accumulator, time_measurement
 
+# These helpers are called with numpy arrays in some backends and plain
+# Python lists in others, which is the point of the comparison this course
+# is about, so the aliases name both.
+Matrix = np.ndarray | list[list[float]]
+Vector = np.ndarray | list[float]
 logger = logging.getLogger(__name__)
 
 
 class LinearAlgebraUtils:
-    @staticmethod
-    @abstractmethod
-    def dot_product(v1, v2) -> None:
-        pass
+    """The interface every linear-algebra backend implements."""
 
     @staticmethod
     @abstractmethod
-    def matrix_vector_multiply(A, x) -> None:
-        pass
+    def dot_product(v1: Vector, v2: Vector) -> None:
+        """Return the dot product of two vectors."""
 
     @staticmethod
     @abstractmethod
-    def vector_norm(v) -> None:
-        pass
+    def matrix_vector_multiply(matrix: Matrix, x: Vector) -> None:
+        """Return the matrix-vector product."""
 
     @staticmethod
     @abstractmethod
-    def vector_scalar_divide(x, scalar) -> None:
-        pass
+    def vector_norm(v: Vector) -> None:
+        """Return the Euclidean norm of a vector."""
 
     @staticmethod
     @abstractmethod
-    def matrix_scalar_multiply(A, w) -> None:
-        pass
+    def vector_scalar_divide(x: Vector, scalar: float) -> None:
+        """Divide every element of a vector by a scalar."""
 
     @staticmethod
     @abstractmethod
-    def vector_vector_subtraction(v1, v2) -> None:
-        pass
+    def matrix_scalar_multiply(matrix: Matrix, w: float) -> None:
+        """Multiply every element of a matrix by a scalar."""
 
     @staticmethod
     @abstractmethod
-    def vector_vector_addition(v1, v2) -> None:
-        pass
+    def vector_vector_subtraction(v1: Vector, v2: Vector) -> None:
+        """Subtract one vector from another, elementwise."""
 
     @staticmethod
     @abstractmethod
-    def scalar_vector_multiply(omega, vector) -> None:
-        pass
+    def vector_vector_addition(v1: Vector, v2: Vector) -> None:
+        """Add two vectors elementwise."""
 
     @staticmethod
     @abstractmethod
-    def matrix_norm(A) -> None:
-        pass
+    def scalar_vector_multiply(omega: float, vector: Vector) -> None:
+        """Multiply every element of a vector by a scalar."""
 
     @staticmethod
     @abstractmethod
-    def matrix_matrix_subtraction(A, B) -> None:
-        pass
+    def matrix_norm(matrix: Matrix) -> None:
+        """Return the Frobenius norm of a matrix."""
+
+    @staticmethod
+    @abstractmethod
+    def matrix_matrix_subtraction(matrix: Matrix, other_matrix: Matrix) -> None:
+        """Subtract one matrix from another, elementwise."""
 
 
 class SequentialLinearAlgebraUtils:
-    @staticmethod
-    def dot_product(v1, v2):
-        return sum(x * y for x, y in zip(v1, v2))
+    """Linear algebra in plain Python, one core."""
 
     @staticmethod
-    def matrix_vector_multiply(A, x):
-        return [SequentialLinearAlgebraUtils.dot_product(row, x) for row in A]
+    def dot_product(v1: Vector, v2: Vector) -> float:
+        """Return the dot product of two vectors."""
+        return sum(x * y for x, y in zip(v1, v2, strict=False))
 
     @staticmethod
-    def vector_norm(v):
+    def matrix_vector_multiply(matrix: Matrix, x: Vector) -> Vector:
+        """Return the matrix-vector product."""
+        return [SequentialLinearAlgebraUtils.dot_product(row, x) for row in matrix]
+
+    @staticmethod
+    def vector_norm(v: Vector) -> float:
+        """Return the Euclidean norm of a vector."""
         x_values = (x * x for x in v)
         x_values_sum = sum(x_values)
         return cmath.sqrt(x_values_sum).real
 
     @staticmethod
-    def vector_scalar_divide(x, scalar):
+    def vector_scalar_divide(x: Vector, scalar: float) -> Vector:
+        """Divide every element of a vector by a scalar."""
         return [xi / scalar for xi in x]
 
     @staticmethod
-    def matrix_scalar_multiply(A, w):
-        return A * w
+    def matrix_scalar_multiply(matrix: Matrix, w: float) -> Matrix:
+        """Multiply every element of a matrix by a scalar."""
+        return matrix * w
 
     @staticmethod
-    def vector_vector_subtraction(v1, v2):
-        return [x - y for x, y in zip(v1, v2)]
+    def vector_vector_subtraction(v1: Vector, v2: Vector) -> Vector:
+        """Subtract one vector from another, elementwise."""
+        return [x - y for x, y in zip(v1, v2, strict=False)]
 
     @staticmethod
-    def vector_vector_addition(v1, v2):
-        return [x + y for x, y in zip(v1, v2)]
+    def vector_vector_addition(v1: Vector, v2: Vector) -> Vector:
+        """Add two vectors elementwise."""
+        return [x + y for x, y in zip(v1, v2, strict=False)]
 
     @staticmethod
-    def scalar_vector_multiply(omega, vector):
+    def scalar_vector_multiply(omega: float, vector: Vector) -> Vector:
+        """Multiply every element of a vector by a scalar."""
         return [omega * x for x in vector]
 
     @staticmethod
-    def matrix_norm(A):
-        return math.sqrt(sum(sum(element**2 for element in row) for row in A))
+    def matrix_norm(matrix: Matrix) -> float:
+        """Return the Frobenius norm of a matrix."""
+        return math.sqrt(sum(sum(element**2 for element in row) for row in matrix))
 
     @staticmethod
-    def matrix_matrix_subtraction(A, B):
-        return [[A[i][j] - B[i][j] for j in range(len(A[0]))] for i in range(len(A))]
+    def matrix_matrix_subtraction(matrix: Matrix, other_matrix: Matrix) -> Matrix:
+        """Subtract one matrix from another, elementwise."""
+        return [
+            [matrix[i][j] - other_matrix[i][j] for j in range(len(matrix[0]))]
+            for i in range(len(matrix))
+        ]
 
 
 class ThreadsLinearAlgebraUtils:
+    """Linear algebra split across threads."""
+
     NUM_THREADS = multiprocessing.cpu_count()
 
     @staticmethod
-    def get_chunk_size(data):
+    def get_chunk_size(data: Vector) -> tuple[int, int, int]:
+        """Work out the chunk size, thread count and remainder."""
         num_elements = len(data)
         num_threads = min(ThreadsLinearAlgebraUtils.NUM_THREADS, num_elements)
         chunk_size = num_elements // num_threads
@@ -123,7 +156,8 @@ class ThreadsLinearAlgebraUtils:
         return chunk_size, num_threads, remainder
 
     @staticmethod
-    def divide_vectors_to_chunks(v1, v2):
+    def divide_vectors_to_chunks(v1: Vector, v2: Vector) -> list[tuple[Vector, Vector]]:
+        """Split two vectors into matching chunks."""
         chunk_size, num_threads, remainder = ThreadsLinearAlgebraUtils.get_chunk_size(
             v1
         )
@@ -138,7 +172,8 @@ class ThreadsLinearAlgebraUtils:
         return chunks
 
     @staticmethod
-    def divide_vector_or_matrix_to_chunks(v):
+    def divide_vector_or_matrix_to_chunks(v: Vector) -> list[Vector]:
+        """Split a vector or matrix into chunks."""
         chunk_size, num_threads, remainder = ThreadsLinearAlgebraUtils.get_chunk_size(v)
 
         chunks = []
@@ -152,8 +187,9 @@ class ThreadsLinearAlgebraUtils:
 
     @staticmethod
     @time_measurement(time_accumulator)
-    def matrix_vector_multiply(A, x):
-        chunks = ThreadsLinearAlgebraUtils.divide_vector_or_matrix_to_chunks(A)
+    def matrix_vector_multiply(matrix: Matrix, x: Vector) -> Vector:
+        """Return the matrix-vector product."""
+        chunks = ThreadsLinearAlgebraUtils.divide_vector_or_matrix_to_chunks(matrix)
         with ThreadPoolExecutor(
             max_workers=ThreadsLinearAlgebraUtils.NUM_THREADS
         ) as executor:
@@ -163,10 +199,12 @@ class ThreadsLinearAlgebraUtils:
 
     @staticmethod
     @time_measurement(time_accumulator)
-    def vector_norm(v):
+    def vector_norm(v: Vector) -> float:
+        """Return the Euclidean norm of a vector."""
         chunks = ThreadsLinearAlgebraUtils.divide_vector_or_matrix_to_chunks(v)
 
-        def partial_norm(chunk):
+        def partial_norm(chunk: Vector) -> float:
+            """Return the sum of squares of one chunk, for a parallel norm."""
             return sum(x * x for x in chunk)
 
         with ThreadPoolExecutor(
@@ -178,7 +216,8 @@ class ThreadsLinearAlgebraUtils:
 
     @staticmethod
     @time_measurement(time_accumulator)
-    def vector_vector_subtraction(v1, v2):
+    def vector_vector_subtraction(v1: Vector, v2: Vector) -> Vector:
+        """Subtract one vector from another, elementwise."""
         chunks = ThreadsLinearAlgebraUtils.divide_vectors_to_chunks(v1, v2)
         with ThreadPoolExecutor(
             max_workers=ThreadsLinearAlgebraUtils.NUM_THREADS
@@ -193,7 +232,8 @@ class ThreadsLinearAlgebraUtils:
 
     @staticmethod
     @time_measurement(time_accumulator)
-    def vector_vector_addition(v1, v2):
+    def vector_vector_addition(v1: Vector, v2: Vector) -> Vector:
+        """Add two vectors elementwise."""
         chunks = ThreadsLinearAlgebraUtils.divide_vectors_to_chunks(v1, v2)
         with ThreadPoolExecutor(
             max_workers=ThreadsLinearAlgebraUtils.NUM_THREADS
@@ -206,7 +246,8 @@ class ThreadsLinearAlgebraUtils:
 
     @staticmethod
     @time_measurement(time_accumulator)
-    def scalar_vector_multiply(omega, vector):
+    def scalar_vector_multiply(omega: float, vector: Vector) -> Vector:
+        """Multiply every element of a vector by a scalar."""
         chunks = ThreadsLinearAlgebraUtils.divide_vector_or_matrix_to_chunks(vector)
         with ThreadPoolExecutor(
             max_workers=ThreadsLinearAlgebraUtils.NUM_THREADS
@@ -222,77 +263,90 @@ class ThreadsLinearAlgebraUtils:
 
 
 @time_measurement(time_accumulator)
-def process_row(params):
-    A, k, i = params
-    factor = A[i][k] / A[k][k]
-    return [A[i][j] - factor * A[k][j] for j in range(len(A[0]))]
+def process_row(params: tuple[object, ...]) -> Vector:
+    """Eliminate one row against the pivot row."""
+    matrix, k, i = params
+    factor = matrix[i][k] / matrix[k][k]
+    return [matrix[i][j] - factor * matrix[k][j] for j in range(len(matrix[0]))]
 
 
 @time_measurement(time_accumulator)
-def divide_by_scalar(pair):
+def divide_by_scalar(pair: tuple[Vector, float]) -> Vector:
+    """Divide one chunk of a vector by its scalar."""
     xi, scalar = pair
     return xi / scalar
 
 
 @time_measurement(time_accumulator)
-def multiply_by_scalar(pair):
+def multiply_by_scalar(pair: tuple[Vector, float]) -> Vector:
+    """Multiply one chunk of a vector by its scalar."""
     element, scalar = pair
     return element * scalar
 
 
 class ProcessLinearAlgebraUtils:
+    """Linear algebra split across a process pool."""
+
     @staticmethod
     @time_measurement(time_accumulator)
-    def dot_product(v1, v2):
+    def dot_product(v1: Vector, v2: Vector) -> float:
+        """Return the dot product of two vectors."""
         with Pool() as pool:
             result = pool.starmap(
-                ProcessLinearAlgebraUtils.multiply_elements, zip(v1, v2)
+                ProcessLinearAlgebraUtils.multiply_elements, zip(v1, v2, strict=False)
             )
         return sum(result)
 
     @staticmethod
     @time_measurement(time_accumulator)
-    def multiply_elements(x, y):
+    def multiply_elements(x: Vector, y: float) -> float:
+        """Return the product of two numbers."""
         return x * y
 
     @staticmethod
     @time_measurement(time_accumulator)
-    def matrix_vector_multiply_row(params):
+    def matrix_vector_multiply_row(params: tuple[object, ...]) -> Vector:
+        """Return one row of a matrix-vector product."""
         row, vector = params
         return SequentialLinearAlgebraUtils.dot_product(row, vector)
 
     @staticmethod
     @time_measurement(time_accumulator)
-    def matrix_vector_multiply(A, x):
+    def matrix_vector_multiply(matrix: Matrix, x: Vector) -> Vector:
+        """Return the matrix-vector product."""
         with Pool() as pool:
             result = pool.map(
                 ProcessLinearAlgebraUtils.matrix_vector_multiply_row,
-                [(row, x) for row in A],
+                [(row, x) for row in matrix],
             )
         return list(result)
 
     @staticmethod
     @time_measurement(time_accumulator)
-    def vector_norm(v):
+    def vector_norm(v: Vector) -> float:
+        """Return the Euclidean norm of a vector."""
         with Pool() as pool:
             squared = pool.map(ProcessLinearAlgebraUtils.square, v)
         return math.sqrt(sum(squared))
 
     @staticmethod
     @time_measurement(time_accumulator)
-    def square(x):
+    def square(x: Vector) -> float:
+        """Return the square of a number."""
         return x * x
 
     @staticmethod
     @time_measurement(time_accumulator)
-    def vector_scalar_divide(x, scalar):
+    def vector_scalar_divide(x: Vector, scalar: float) -> Vector:
+        """Divide every element of a vector by a scalar."""
         with Pool() as pool:
             result = pool.map(divide_by_scalar, [(xi, scalar) for xi in x])
         return list(result)
 
     @staticmethod
     @time_measurement(time_accumulator)
-    def divide_vector_by_scalar(x, scalar):
+    def divide_vector_by_scalar(x: Vector, scalar: float) -> Vector:
+        """Divide every element of a vector by a scalar."""
         with Pool() as pool:
             result = pool.map(
                 ProcessLinearAlgebraUtils.vector_scalar_divide,
@@ -302,49 +356,54 @@ class ProcessLinearAlgebraUtils:
 
     @staticmethod
     @time_measurement(time_accumulator)
-    def matrix_scalar_multiply_row(params):
+    def matrix_scalar_multiply_row(params: tuple[object, ...]) -> Vector:
+        """Multiply one row of a matrix by a scalar."""
         row, w = params
         return [w * element for element in row]
 
     @staticmethod
     @time_measurement(time_accumulator)
-    def matrix_scalar_multiply(A, w):
+    def matrix_scalar_multiply(matrix: Matrix, w: float) -> Matrix:
+        """Multiply every element of a matrix by a scalar."""
         with Pool() as pool:
-            result = pool.map(
+            return pool.map(
                 ProcessLinearAlgebraUtils.matrix_scalar_multiply_row,
-                [(row, w) for row in A],
+                [(row, w) for row in matrix],
             )
-        return result
 
     @staticmethod
     @time_measurement(time_accumulator)
-    def vector_vector_operation(params):
+    def vector_vector_operation(params: tuple[object, ...]) -> Vector:
+        """Apply one elementwise operation to a pair of chunks."""
         v1, v2, op = params
         return op(v1, v2)
 
     @staticmethod
     @time_measurement(time_accumulator)
-    def vector_vector_subtraction(v1, v2):
+    def vector_vector_subtraction(v1: Vector, v2: Vector) -> Vector:
+        """Subtract one vector from another, elementwise."""
         with Pool() as pool:
             result = pool.map(
                 ProcessLinearAlgebraUtils.vector_vector_operation,
-                zip(v1, v2, itertools.repeat(operator.sub)),
+                zip(v1, v2, itertools.repeat(operator.sub), strict=False),
             )
         return list(result)
 
     @staticmethod
     @time_measurement(time_accumulator)
-    def vector_vector_addition(v1, v2):
+    def vector_vector_addition(v1: Vector, v2: Vector) -> Vector:
+        """Add two vectors elementwise."""
         with Pool() as pool:
             result = pool.map(
                 ProcessLinearAlgebraUtils.vector_vector_operation,
-                zip(v1, v2, itertools.repeat(operator.add)),
+                zip(v1, v2, itertools.repeat(operator.add), strict=False),
             )
         return list(result)
 
     @staticmethod
     @time_measurement(time_accumulator)
-    def scalar_vector_multiply(omega, vector):
+    def scalar_vector_multiply(omega: float, vector: Vector) -> Vector:
+        """Multiply every element of a vector by a scalar."""
         with Pool() as pool:
             result = pool.map(
                 multiply_by_scalar, [(element, omega) for element in vector]
@@ -353,120 +412,142 @@ class ProcessLinearAlgebraUtils:
 
     @staticmethod
     @time_measurement(time_accumulator)
-    def sum_of_squares(row):
+    def sum_of_squares(row: Vector) -> float:
+        """Return the sum of the squares of a row."""
         return sum(x**2 for x in row)
 
     @staticmethod
     @time_measurement(time_accumulator)
-    def matrix_norm(A):
+    def matrix_norm(matrix: Matrix) -> float:
+        """Return the Frobenius norm of a matrix."""
         with Pool() as pool:
-            row_sums = pool.map(ProcessLinearAlgebraUtils.sum_of_squares, A)
+            row_sums = pool.map(ProcessLinearAlgebraUtils.sum_of_squares, matrix)
         return math.sqrt(sum(row_sums))
 
     @staticmethod
     @time_measurement(time_accumulator)
-    def subtract_rows(row_from_A, row_from_B):
-        return [a - b for a, b in zip(row_from_A, row_from_B)]
+    def subtract_rows(row_from_matrix: Vector, row_from_other: Vector) -> Vector:
+        """Subtract one row from another, elementwise."""
+        return [a - b for a, b in zip(row_from_matrix, row_from_other, strict=False)]
 
     @staticmethod
     @time_measurement(time_accumulator)
-    def matrix_matrix_subtraction(A, B):
+    def matrix_matrix_subtraction(matrix: Matrix, other_matrix: Matrix) -> Matrix:
+        """Subtract one matrix from another, elementwise."""
         with Pool() as pool:
-            result = pool.starmap(ProcessLinearAlgebraUtils.subtract_rows, zip(A, B))
-        return result
+            return pool.starmap(
+                ProcessLinearAlgebraUtils.subtract_rows,
+                zip(matrix, other_matrix, strict=False),
+            )
 
 
-class DistributedArraysLinearAlgebraUtils(ABC):
+class DistributedArraysLinearAlgebraUtils:
+    """Linear algebra over dask distributed arrays."""
+
     @staticmethod
     @time_measurement(time_accumulator)
-    def dot_product(v1, v2):
+    def dot_product(v1: Vector, v2: Vector) -> float:
+        """Return the dot product of two vectors."""
         dv1 = da.from_array(v1, chunks="auto")
         dv2 = da.from_array(v2, chunks="auto")
         return da.dot(dv1, dv2).compute()
 
     @staticmethod
     @time_measurement(time_accumulator)
-    def matrix_vector_multiply(A, x):
-        dA = da.from_array(A, chunks="auto")
+    def matrix_vector_multiply(matrix: Matrix, x: Vector) -> Vector:
+        """Return the matrix-vector product."""
+        dask_matrix = da.from_array(matrix, chunks="auto")
         dx = da.from_array(x, chunks="auto")
-        return da.dot(dA, dx).compute().tolist()
+        return da.dot(dask_matrix, dx).compute().tolist()
 
     @staticmethod
     @time_measurement(time_accumulator)
-    def vector_norm(v):
+    def vector_norm(v: Vector) -> float:
+        """Return the Euclidean norm of a vector."""
         dv = da.from_array(v, chunks="auto")
         return da.linalg.norm(dv).compute()
 
     @staticmethod
     @time_measurement(time_accumulator)
-    def vector_scalar_divide(x, scalar):
+    def vector_scalar_divide(x: Vector, scalar: float) -> Vector:
+        """Divide every element of a vector by a scalar."""
         dx = da.from_array(x, chunks="auto")
         return (dx / scalar).compute().tolist()
 
     @staticmethod
     @time_measurement(time_accumulator)
-    def matrix_scalar_multiply(A, w):
-        dA = da.from_array(A, chunks="auto")
-        return (dA * w).compute().tolist()
+    def matrix_scalar_multiply(matrix: Matrix, w: float) -> Matrix:
+        """Multiply every element of a matrix by a scalar."""
+        dask_matrix = da.from_array(matrix, chunks="auto")
+        return (dask_matrix * w).compute().tolist()
 
     @staticmethod
     @time_measurement(time_accumulator)
-    def vector_vector_subtraction(v1, v2):
+    def vector_vector_subtraction(v1: Vector, v2: Vector) -> Vector:
+        """Subtract one vector from another, elementwise."""
         dv1 = da.from_array(v1, chunks="auto")
         dv2 = da.from_array(v2, chunks="auto")
         return (dv1 - dv2).compute().tolist()
 
     @staticmethod
     @time_measurement(time_accumulator)
-    def vector_vector_addition(v1, v2):
+    def vector_vector_addition(v1: Vector, v2: Vector) -> Vector:
+        """Add two vectors elementwise."""
         dv1 = da.from_array(v1, chunks="auto")
         dv2 = da.from_array(v2, chunks="auto")
         return (dv1 + dv2).compute().tolist()
 
     @staticmethod
     @time_measurement(time_accumulator)
-    def scalar_vector_multiply(omega, vector):
+    def scalar_vector_multiply(omega: float, vector: Vector) -> Vector:
+        """Multiply every element of a vector by a scalar."""
         dvector = da.from_array(vector, chunks="auto")
         return (omega * dvector).compute().tolist()
 
     @staticmethod
     @time_measurement(time_accumulator)
-    def matrix_norm(A):
-        dA = da.from_array(A, chunks="auto")
-        return da.linalg.norm(dA).compute()
+    def matrix_norm(matrix: Matrix) -> float:
+        """Return the Frobenius norm of a matrix."""
+        dask_matrix = da.from_array(matrix, chunks="auto")
+        return da.linalg.norm(dask_matrix).compute()
 
     @staticmethod
     @time_measurement(time_accumulator)
-    def matrix_matrix_subtraction(A, B):
-        dA = da.from_array(A, chunks="auto")
-        dB = da.from_array(B, chunks="auto")
-        return (dA - dB).compute().tolist()
+    def matrix_matrix_subtraction(matrix: Matrix, other_matrix: Matrix) -> Matrix:
+        """Subtract one matrix from another, elementwise."""
+        dask_matrix = da.from_array(matrix, chunks="auto")
+        dask_other = da.from_array(other_matrix, chunks="auto")
+        return (dask_matrix - dask_other).compute().tolist()
 
     @staticmethod
     @time_measurement(time_accumulator)
-    def gaussian_elimination(A, b):
+    def gaussian_elimination(matrix: Matrix, b: Vector) -> Vector | None:
+        """Solve the system by Gaussian elimination."""
         try:
-            dA = da.from_array(A, chunks="auto")
+            dask_matrix = da.from_array(matrix, chunks="auto")
             db = da.from_array(b, chunks="auto")
-            Ab = da.hstack([dA, db[:, None]])
-            Ab = Ab.persist()
+            augmented = da.hstack([dask_matrix, db[:, None]])
+            augmented = augmented.persist()
 
-            def elimination_step(Ab, k):
-                n = Ab.shape[0]
-                max_index = da.argmax(da.abs(Ab[k:, k])) + k
-                Ab[[k, max_index]] = Ab[[max_index, k]]
-                Ab = Ab.persist()
-                factor = Ab[k + 1 :, k] / Ab[k, k]
-                Ab[k + 1 :] -= factor[:, None] * Ab[k]
-                return Ab
+            def elimination_step(augmented: Matrix, k: int) -> Matrix:
+                """Run one Gaussian elimination step over the augmented matrix."""
+                max_index = da.argmax(da.abs(augmented[k:, k])) + k
+                augmented[[k, max_index]] = augmented[[max_index, k]]
+                augmented = augmented.persist()
+                factor = augmented[k + 1 :, k] / augmented[k, k]
+                augmented[k + 1 :] -= factor[:, None] * augmented[k]
+                return augmented
 
-            for k in range(A.shape[0]):
-                Ab = elimination_step(Ab, k)
+            for k in range(matrix.shape[0]):
+                augmented = elimination_step(augmented, k)
 
-            x = da.zeros(A.shape[0])
-            for i in range(A.shape[0] - 1, -1, -1):
-                x[i] = (Ab[i, -1] - da.dot(Ab[i, i + 1 : -1], x[i + 1 :])) / Ab[i, i]
+            x = da.zeros(matrix.shape[0])
+            for i in range(matrix.shape[0] - 1, -1, -1):
+                x[i] = (
+                    augmented[i, -1] - da.dot(augmented[i, i + 1 : -1], x[i + 1 :])
+                ) / augmented[i, i]
             return x.compute().tolist()
-        except Exception as e:
-            logger.info("Error during Gaussian elimination: %s", e)
+        except (ValueError, ArithmeticError) as exc:
+            # A singular or badly-scaled system; dask surfaces both here.
+            logger.info("Error during Gaussian elimination: %s", exc)
             return None

Programming/PORR/code/main.py

@@ -1,3 +1,4 @@
+#!/usr/bin/env python3
 """Entry point for the PORR benchmark suite.
 
 Runs ``tests.py``, which compares this project's Richardson implementations

Programming/PORR/code/matrix_generator.py

@@ -1,53 +1,71 @@
-import os
+#!/usr/bin/env python3
+"""Build and cache the SPD, nemeth12 and poli3 test problems."""
+
+from pathlib import Path
 
 import numpy as np
 import scipy.io
 
+# Generator rather than the legacy global np.random state (NPY002).
+_RNG = np.random.default_rng()
+# These helpers are called with numpy arrays in some backends and plain
+# Python lists in others, which is the point of the comparison this course
+# is about, so the aliases name both.
+Matrix = np.ndarray | list[list[float]]
+Vector = np.ndarray | list[float]
+
 
 class MatrixGenerator:
+    """Build or load the test problems this course measures against."""
+
     @staticmethod
     def generate_spd_matrix(n: int) -> np.ndarray:
-        A = np.random.rand(n, n)
-        return np.dot(A, A.T) + n * MatrixGenerator.generate_identity_matrix(
+        """Generate a symmetric positive-definite matrix."""
+        matrix = _RNG.random((n, n))
+        return np.dot(matrix, matrix.T) + n * MatrixGenerator.generate_identity_matrix(
             n
         )  # Adding n*I ensures positive definiteness
 
     @staticmethod
-    def generate_identity_matrix(size):
+    def generate_identity_matrix(size: int) -> Matrix:
+        """Return an identity matrix of the given size."""
         return np.eye(size)
 
     @staticmethod
-    def generate_alternating_vector(size):
+    def generate_alternating_vector(size: int) -> Vector:
+        """Return a vector alternating between 1 and 2."""
         return np.tile([1, 2], int(np.ceil(size / 2)))[:size]
 
     @staticmethod
-    def get_matrix_from_file(file_path, problem):
+    def get_matrix_from_file(file_path: str, problem: int) -> Matrix:
+        """Load a matrix from a MatrixMarket .mat file."""
         mat_contents = scipy.io.loadmat(file_path)
         problem_record = mat_contents["Problem"][0][0]
-        A = problem_record[problem]
-        if scipy.sparse.issparse(A):
-            A_dense = A.todense()
-        else:
-            A_dense = A
-        return np.array(A_dense)
+        matrix = problem_record[problem]
+        dense = matrix.todense() if scipy.sparse.issparse(matrix) else matrix
+        return np.array(dense)
 
     @staticmethod
-    def generate_matrix_and_vector(type, size=None):
-        if type == "spd":
+    def generate_matrix_and_vector(
+        kind: str, size: int | None = None
+    ) -> tuple[Matrix, Vector, float, float]:
+        """Return the matrix, vector and eigenvalue bounds."""
+        if kind == "spd":
             if size is None:
-                raise ValueError("Size must be provided for SPD matrix generation.")
+                msg = "Size must be provided for SPD matrix generation."
+                raise ValueError(msg)
             try:
                 matrix, vector, lambda_min, lambda_max = MatrixGenerator.load_from_file(
                     "spd_" + str(size) + ".npz"
                 )
             except FileNotFoundError:
                 matrix = MatrixGenerator.generate_spd_matrix(size)
-                vector = np.random.uniform(-1, 1, size)
+                vector = _RNG.uniform(-1, 1, size)
                 lambda_min, lambda_max = MatrixGenerator.calculate_eigenvalues(matrix)
                 MatrixGenerator.save_to_file(
                     matrix, vector, lambda_min, lambda_max, "spd_" + str(size) + ".npz"
                 )
-        elif type == "nemeth12":
+        elif kind == "nemeth12":
             try:
                 matrix, vector, lambda_min, lambda_max = MatrixGenerator.load_from_file(
                     "nemeth12.npz"
@@ -60,7 +78,7 @@ class MatrixGenerator:
                 MatrixGenerator.save_to_file(
                     matrix, vector, lambda_min, lambda_max, "nemeth12.npz"
                 )
-        elif type == "poli3":
+        elif kind == "poli3":
             try:
                 matrix, vector, lambda_min, lambda_max = MatrixGenerator.load_from_file(
                     "poli3.npz"
@@ -74,21 +92,28 @@ class MatrixGenerator:
                     matrix, vector, lambda_min, lambda_max, "poli3.npz"
                 )
         else:
-            raise ValueError(
-                "Invalid type specified. Choose 'spd', 'nemeth12', or 'poli3'."
-            )
+            msg = "Invalid type specified. Choose 'spd', 'nemeth12', or 'poli3'."
+            raise ValueError(msg)
 
         return matrix, vector, lambda_min, lambda_max
 
     @staticmethod
-    def calculate_eigenvalues(A):
-        eigenvalues = np.linalg.eigvals(A)
+    def calculate_eigenvalues(matrix: Matrix) -> tuple[float, float]:
+        """Return the smallest and largest eigenvalue."""
+        eigenvalues = np.linalg.eigvals(matrix)
         lambda_min = np.min(eigenvalues)
         lambda_max = np.max(eigenvalues)
         return lambda_min, lambda_max
 
     @staticmethod
-    def save_to_file(matrix, vector, lambda_min, lambda_max, file_path) -> None:
+    def save_to_file(
+        matrix: Matrix,
+        vector: Vector,
+        lambda_min: float,
+        lambda_max: float,
+        file_path: str,
+    ) -> None:
+        """Cache a generated problem so the next run can reuse it."""
         np.savez(
             file_path,
             matrix=matrix,
@@ -97,9 +122,12 @@ class MatrixGenerator:
             lambda_max=lambda_max,
         )
 
-    def load_from_file(file_path):
-        if not os.path.exists(file_path):
-            raise FileNotFoundError(f"The file {file_path} does not exist.")
+    @staticmethod
+    def load_from_file(file_path: str) -> tuple[Matrix, Vector, float, float]:
+        """Load a cached problem."""
+        if not Path(file_path).exists():
+            msg = f"The file {file_path} does not exist."
+            raise FileNotFoundError(msg)
         data = np.load(file_path)
         matrix = data["matrix"]
         vector = data["vector"]

Programming/PORR/code/processing_type.py

@@ -1,7 +1,12 @@
+#!/usr/bin/env python3
+"""The enum naming the four processing backends."""
+
 from enum import Enum, auto
 
 
 class ProcessingType(Enum):
+    """Which backend a run should use."""
+
     SEQUENTIAL = auto()
     THREADS = auto()
     PROCESSES = auto()

Programming/PORR/code/richardson_method.py

@@ -1,79 +1,100 @@
+#!/usr/bin/env python3
+"""Richardson iteration, parameterised by linear-algebra backend."""
+
 import gc
 import logging
 import time
 
-import linear_algebra_utils as linAlg
+import linear_algebra_utils as lin_alg
+import numpy as np
 from processing_type import ProcessingType
+from richardson_problem import Problem, Settings
 from time_measurement import tests_time, time_accumulator, time_measurement
 
+# These helpers are called with numpy arrays in some backends and plain
+# Python lists in others, which is the point of the comparison this course
+# is about, so the aliases name both.
+Matrix = np.ndarray | list[list[float]]
+Vector = np.ndarray | list[float]
 logger = logging.getLogger(__name__)
 
 
 class RichardsonMethod:
+    """Richardson iteration over a pluggable linear-algebra backend."""
+
     @time_measurement(time_accumulator)
     def __init__(
         self,
         method: ProcessingType,
-        A,
-        b,
-        lambda_min,
-        lambda_max,
-        max_iterations,
-        size: int,
-        x0=None,
-        tol=1e-5,
+        problem: Problem,
+        settings: Settings,
     ) -> None:
-        self.LinAlg = self.assign_LinAlgType(method)
-        self.A = A
-        self.b = b
-        self.x0 = x0 if x0 is not None else [0.0] * len(b)
-        self.max_iterations = max_iterations
-        self.tol = tol
-        self.lambda_min = lambda_min
-        self.lambda_max = lambda_max
+        """Set up the solver for one problem.
+
+        `size` used to be a parameter and was never read -- len(b) is the
+        size, and it is right here.
+        """
+        self.LinAlg = self.assign_lin_alg_type(method)
+        self.matrix = problem.matrix
+        self.b = problem.rhs
+        self.x0 = settings.x0 if settings.x0 is not None else [0.0] * len(problem.rhs)
+        self.max_iterations = settings.max_iterations
+        self.tol = settings.tol
+        self.lambda_min = problem.lambda_min
+        self.lambda_max = problem.lambda_max
         if self.lambda_min < 0:
-            raise ValueError("Matrix A is not positive semi-definite.")
+            msg = "Matrix A is not positive semi-definite."
+            raise ValueError(msg)
         self.omega = RichardsonMethod.calculate_omega(self.lambda_min, self.lambda_max)
 
     @staticmethod
-    def calculate_omega(lambda_min, lambda_max):
+    def calculate_omega(lambda_min: float, lambda_max: float) -> float:
+        """Return the optimal Richardson relaxation factor."""
         return 2 / (lambda_min + lambda_max)
 
     @staticmethod
-    def convergence_norm(LinAlgType, A, omega, I) -> bool:
-        wA = LinAlgType.matrix_scalar_multiply(A, omega)
-        IMinuswA = LinAlgType.matrix_matrix_subtraction(I, wA)
-        return LinAlgType.matrix_norm(IMinuswA)
+    def convergence_norm(
+        lin_alg_type: type, matrix: Matrix, omega: float, identity: Matrix
+    ) -> bool:
+        """Report whether the iteration matrix norm is under 1."""
+        scaled_matrix = lin_alg_type.matrix_scalar_multiply(matrix, omega)
+        identity_minus_scaled = lin_alg_type.matrix_matrix_subtraction(
+            identity, scaled_matrix
+        )
+        return lin_alg_type.matrix_norm(identity_minus_scaled)
 
     @staticmethod
-    def assign_LinAlgType(method):
+    def assign_lin_alg_type(method: ProcessingType) -> type:
+        """Return the linear-algebra backend for a processing type."""
         methods = {
-            ProcessingType.SEQUENTIAL: linAlg.SequentialLinearAlgebraUtils,
-            ProcessingType.THREADS: linAlg.ThreadsLinearAlgebraUtils,
-            ProcessingType.PROCESSES: linAlg.ProcessLinearAlgebraUtils,
-            ProcessingType.DISTRIBUTED_ARRAYS: linAlg.DistributedArraysLinearAlgebraUtils,
+            ProcessingType.SEQUENTIAL: lin_alg.SequentialLinearAlgebraUtils,
+            ProcessingType.THREADS: lin_alg.ThreadsLinearAlgebraUtils,
+            ProcessingType.PROCESSES: lin_alg.ProcessLinearAlgebraUtils,
+            ProcessingType.DISTRIBUTED_ARRAYS: (
+                lin_alg.DistributedArraysLinearAlgebraUtils
+            ),
         }
 
         try:
             return methods[method]
-        except KeyError:
-            raise ValueError(
-                "Unknown method, please use 'SEQUENTIAL', 'THREADS' or 'PROCESSES'."
-            )
+        except KeyError as exc:
+            msg = "Unknown method, please use 'SEQUENTIAL', 'THREADS' or 'PROCESSES'."
+            raise ValueError(msg) from exc
 
-    def solve(self):
+    def solve(self) -> tuple[Vector, str | None]:
+        """Run the Richardson iteration to convergence, or give up."""
         gc.disable()
         time_accumulator.total_time = 0
         start = time.perf_counter()
         x = self.x0[:]
 
-        for iteration in range(self.max_iterations):
-            Ax = self.LinAlg.matrix_vector_multiply(self.A, x)
-            residual = self.LinAlg.vector_vector_subtraction(self.b, Ax)
+        for _iteration in range(self.max_iterations):
+            product = self.LinAlg.matrix_vector_multiply(self.matrix, x)
+            residual = self.LinAlg.vector_vector_subtraction(self.b, product)
             x = self.LinAlg.vector_vector_addition(
                 x, self.LinAlg.scalar_vector_multiply(self.omega, residual)
             )
-            if linAlg.SequentialLinearAlgebraUtils.vector_norm(residual) < self.tol:
+            if lin_alg.SequentialLinearAlgebraUtils.vector_norm(residual) < self.tol:
                 break
 
         end = time.perf_counter()
@@ -81,11 +102,11 @@ class RichardsonMethod:
         gc.enable()
 
         match self.LinAlg:
-            case linAlg.SequentialLinearAlgebraUtils:
+            case lin_alg.SequentialLinearAlgebraUtils:
                 logger.info(
                     "Total: %ss, Tests time: %ss", total_time, tests_time.total_time
                 )
-            case linAlg.ThreadsLinearAlgebraUtils:
+            case lin_alg.ThreadsLinearAlgebraUtils:
                 sequential_time = total_time - time_accumulator.total_time
                 logger.info(
                     "Total: %ss, Seq: %ss, Parallel (threads): %ss, Tests time: %ss",
@@ -94,7 +115,7 @@ class RichardsonMethod:
                     time_accumulator.total_time,
                     tests_time.total_time,
                 )
-            case linAlg.ProcessLinearAlgebraUtils:
+            case lin_alg.ProcessLinearAlgebraUtils:
                 sequential_time = total_time - time_accumulator.total_time
                 logger.info(
                     "Total: %ss, Seq: %ss, Parallel (processes): %ss, Tests time: %ss",
@@ -103,10 +124,11 @@ class RichardsonMethod:
                     time_accumulator.total_time,
                     tests_time.total_time,
                 )
-            case linAlg.DistributedArraysLinearAlgebraUtils:
+            case lin_alg.DistributedArraysLinearAlgebraUtils:
                 sequential_time = total_time - time_accumulator.total_time
                 logger.info(
-                    "Total: %ss, Seq: %ss, Parallel (distributed arrays): %ss, Tests time: %ss",
+                    "Total: %ss, Seq: %ss, Parallel (distributed arrays): %ss, "
+                    "Tests time: %ss",
                     total_time,
                     sequential_time,
                     time_accumulator.total_time,

Programming/PORR/code/richardson_mpi.py

@@ -1,25 +1,42 @@
+#!/usr/bin/env python3
+"""Richardson iteration split across MPI ranks."""
+
 import logging
 import time
 
 import numpy as np
 from mpi4py import MPI
+from richardson_problem import Problem, Settings
 
+# These helpers are called with numpy arrays in some backends and plain
+# Python lists in others, which is the point of the comparison this course
+# is about, so the aliases name both.
+Matrix = np.ndarray | list[list[float]]
+Vector = np.ndarray | list[float]
 logger = logging.getLogger(__name__)
 
 
-def richardson_parallel(A, b, lambda_min, lambda_max, tol=1e-5, max_iter=10000):
+def richardson_parallel(
+    problem: Problem, settings: Settings | None = None
+) -> tuple[Vector, float]:
+    """Run Richardson across MPI ranks."""
+    if settings is None:
+        settings = Settings(max_iterations=1000)
+    matrix = problem.matrix
+    b = problem.rhs
+    lambda_min = problem.lambda_min
+    lambda_max = problem.lambda_max
+    tol = settings.tol
+    max_iter = settings.max_iterations
     comm = MPI.COMM_WORLD
     rank = comm.Get_rank()
     size = comm.Get_size()
 
     # Rozmiar macierzy A
-    n = A.shape[0]
+    n = matrix.shape[0]
 
     # Obliczanie wartości własnych tylko na jednym procesie
-    if rank == 0:
-        omega = 2 / (lambda_min + lambda_max)
-    else:
-        omega = None
+    omega = 2 / (lambda_min + lambda_max) if rank == 0 else None
 
     # Rozgłoszenie omega do wszystkich procesów
     omega = comm.bcast(omega, root=0)
@@ -33,7 +50,7 @@ def richardson_parallel(A, b, lambda_min, lambda_max, tol=1e-5, max_iter=10000):
     end_row = start_row + local_rows if rank != size - 1 else n
 
     # Przydzielenie lokalnych porcji A i b
-    local_A = A[start_row:end_row, :]
+    local_matrix = matrix[start_row:end_row, :]
     local_b = b[start_row:end_row]
 
     # Lokalny wektor residuum
@@ -44,9 +61,9 @@ def richardson_parallel(A, b, lambda_min, lambda_max, tol=1e-5, max_iter=10000):
 
     start_time = time.time()
 
-    for i in range(max_iter):
+    for _i in range(max_iter):
         # Oblicz lokalny residuum r = b - A @ x
-        local_r[:] = local_b - np.dot(local_A, x)
+        local_r[:] = local_b - np.dot(local_matrix, x)
 
         # Tworzymy tymczasowy wektor o pełnym rozmiarze i kopiujemy lokalne dane
         temp_r = np.zeros_like(b, dtype=np.float64)
@@ -69,8 +86,11 @@ def richardson_parallel(A, b, lambda_min, lambda_max, tol=1e-5, max_iter=10000):
     return x, execution_time
 
 
-def check_solution(A, b, x_approx, tolerance=8e-3):
-    x_true = np.linalg.solve(A, b)
+def check_solution(
+    matrix: Matrix, b: Vector, x_approx: Vector, tolerance: float = 0.008
+) -> tuple[bool, float]:
+    """Report whether an approximate solution is within tolerance."""
+    x_true = np.linalg.solve(matrix, b)
     error = np.linalg.norm(x_true - x_approx)
     return error < tolerance, error
 
@@ -85,25 +105,25 @@ if __name__ == "__main__":
     sizes = [2, 5, 10, 50, 80, 100, 300, 500, 750, 1000, 5000, 10000]
     for i in sizes:
         if rank == 0:
-            A, b, lambda_min, lambda_max = MatrixGenerator.generate_matrix_and_vector(
-                "spd", size=i
+            matrix, b, lambda_min, lambda_max = (
+                MatrixGenerator.generate_matrix_and_vector("spd", size=i)
             )
         else:
-            A = None
+            matrix = None
             b = None
             lambda_min = None
             lambda_max = None
 
-        A = comm.bcast(A, root=0)
+        matrix = comm.bcast(matrix, root=0)
         b = comm.bcast(b, root=0)
 
         # Rozwiązanie przy użyciu zrównoleglonej metody Richardsona
-        x, time_taken = richardson_parallel(A, b, lambda_min, lambda_max)
+        x, time_taken = richardson_parallel(Problem(matrix, b, lambda_min, lambda_max))
 
         # Sprawdzanie poprawności rozwiązania (na procesie 0)
         if rank == 0:
             logger.info("Spd matrix with size %s", i)
-            is_correct, error = check_solution(A, b, x)
+            is_correct, error = check_solution(matrix, b, x)
             logger.info("Czas wykonania [s]: %s", time_taken)
             logger.info(
                 "Czy rozwiązanie jest poprawne: %s", "Tak" if is_correct else "Nie"
@@ -111,49 +131,49 @@ if __name__ == "__main__":
             logger.info("Błąd rozwiązania: %s", error)
 
     if rank == 0:
-        A, b, lambda_min, lambda_max = MatrixGenerator.generate_matrix_and_vector(
+        matrix, b, lambda_min, lambda_max = MatrixGenerator.generate_matrix_and_vector(
             "nemeth12"
         )
     else:
-        A = None
+        matrix = None
         b = None
         lambda_min = None
         lambda_max = None
 
-    A = comm.bcast(A, root=0)
+    matrix = comm.bcast(matrix, root=0)
     b = comm.bcast(b, root=0)
 
     # Rozwiązanie przy użyciu zrównoleglonej metody Richardsona
-    x, time_taken = richardson_parallel(A, b, lambda_min, lambda_max)
+    x, time_taken = richardson_parallel(Problem(matrix, b, lambda_min, lambda_max))
 
     # Sprawdzanie poprawności rozwiązania (na procesie 0)
     if rank == 0:
         logger.info("Nemeth12 matrix")
-        is_correct, error = check_solution(A, b, x)
+        is_correct, error = check_solution(matrix, b, x)
         logger.info("Czas wykonania [s]: %s", time_taken)
         logger.info("Czy rozwiązanie jest poprawne: %s", "Tak" if is_correct else "Nie")
         logger.info("Błąd rozwiązania: %s", error)
 
     if rank == 0:
-        A, b, lambda_min, lambda_max = MatrixGenerator.generate_matrix_and_vector(
+        matrix, b, lambda_min, lambda_max = MatrixGenerator.generate_matrix_and_vector(
             "poli3"
         )
     else:
-        A = None
+        matrix = None
         b = None
         lambda_min = None
         lambda_max = None
 
-    A = comm.bcast(A, root=0)
+    matrix = comm.bcast(matrix, root=0)
     b = comm.bcast(b, root=0)
 
     # Rozwiązanie przy użyciu zrównoleglonej metody Richardsona
-    x, time_taken = richardson_parallel(A, b, lambda_min, lambda_max)
+    x, time_taken = richardson_parallel(Problem(matrix, b, lambda_min, lambda_max))
 
     # Sprawdzanie poprawności rozwiązania (na procesie 0)
     if rank == 0:
         logger.info("Poli3 matrix")
-        is_correct, error = check_solution(A, b, x)
+        is_correct, error = check_solution(matrix, b, x)
         logger.info("Czas wykonania [s]: %s", time_taken)
         logger.info("Czy rozwiązanie jest poprawne: %s", "Tak" if is_correct else "Nie")
         logger.info("Błąd rozwiązania: %s", error)

Programming/PORR/code/richardson_problem.py

@@ -0,0 +1,34 @@
+#!/usr/bin/env python3
+"""The system to solve, and the settings to solve it with.
+
+Every Richardson backend in this project took the same seven or nine loose
+arguments -- matrix, right-hand side, both eigenvalue bounds, iteration cap,
+starting guess and tolerance -- which is what put all of them over ruff's
+argument limit. They travel together, so they are two dataclasses.
+"""
+
+from dataclasses import dataclass
+
+import numpy as np
+
+Matrix = np.ndarray | list[list[float]]
+Vector = np.ndarray | list[float]
+
+
+@dataclass(frozen=True)
+class Problem:
+    """One linear system, with the eigenvalue bounds Richardson needs."""
+
+    matrix: Matrix
+    rhs: Vector
+    lambda_min: float
+    lambda_max: float
+
+
+@dataclass(frozen=True)
+class Settings:
+    """How hard to try before giving up."""
+
+    max_iterations: int
+    x0: Vector | None = None
+    tol: float = 1e-05

Programming/PORR/code/tests.py

@@ -1,3 +1,6 @@
+#!/usr/bin/env python3
+"""Compare every Richardson backend against numpy's direct solver."""
+
 import logging
 
 import numpy as np
@@ -5,34 +8,48 @@ import pytest
 from matrix_generator import MatrixGenerator
 from processing_type import ProcessingType
 from richardson_method import RichardsonMethod
-from threads_indep import RichardsonMethodThreads
+from richardson_problem import Problem, Settings
+from threads_indep import richardson_method_threads
 from time_measurement import tests_time, time_measurement
 
+# These helpers are called with numpy arrays in some backends and plain
+# Python lists in others, which is the point of the comparison this course
+# is about, so the aliases name both.
+Matrix = np.ndarray | list[list[float]]
+Vector = np.ndarray | list[float]
 logger = logging.getLogger(__name__)
 
+# nemeth12 and poli3 come from a file, so their size is fixed; the suite
+# only runs them under the first n to avoid repeating identical work.
+FIXED_SIZE_N = 2
 
-def calculate_norm_numpy(I, w, A):
-    difference = I - w * A
+
+def calculate_norm_numpy(identity: Matrix, w: float, matrix: Matrix) -> float:
+    """Return the norm of I - wA, computed with numpy."""
+    difference = identity - w * matrix
     return np.linalg.norm(difference)
 
 
-def calculate_eigenvalues(A):
-    eigenvalues = np.linalg.eigvals(A)
+def calculate_eigenvalues(matrix: Matrix) -> tuple[float, float]:
+    """Return the smallest and largest eigenvalue."""
+    eigenvalues = np.linalg.eigvals(matrix)
     lambda_min = np.min(eigenvalues)
     lambda_max = np.max(eigenvalues)
     return lambda_min, lambda_max
 
 
-def calcualte_norm_from_matrix_numpy(A, n):
-    lambda_min, lambda_max = calculate_eigenvalues(A)
+def calcualte_norm_from_matrix_numpy(matrix: Matrix, n: int) -> float:
+    """Return the iteration-matrix norm for a matrix."""
+    lambda_min, lambda_max = calculate_eigenvalues(matrix)
     omega = 2 / (lambda_min + lambda_max)
-    I = np.eye(n)
-    return calculate_norm_numpy(I, omega, A)
+    identity = np.eye(n)
+    return calculate_norm_numpy(identity, omega, matrix)
 
 
 @time_measurement(tests_time)
-def solution_lib(A, b):
-    return np.linalg.solve(A, b)
+def solution_lib(matrix: Matrix, b: Vector) -> Vector:
+    """Solve the system with numpy, as the reference answer."""
+    return np.linalg.solve(matrix, b)
 
 
 @pytest.mark.parametrize("n", [2, 5, 10, 50, 100, 300, 500, 750, 1000, 5000, 10000])
@@ -47,86 +64,99 @@ def solution_lib(A, b):
 )
 @pytest.mark.parametrize("matrix_type", ["spd", "nemeth12", "poli3"])
 def test_richardson_vs_cg(
-    n: int, processing_type: ProcessingType, matrix_type: str, capsys
+    n: int,
+    processing_type: ProcessingType,
+    matrix_type: str,
+    capsys: pytest.CaptureFixture[str],
 ) -> None:
+    """Every backend must agree with numpy's direct solver, at every size."""
     logger.info("matrix type:  %s", matrix_type)
     logger.info("matrix size:  %s", n if matrix_type == "spd" else "fixed")
     tolerance = 8e-3
     max_iterations = 100
-    if matrix_type in ["nemeth12", "poli3"] and n != 2:
+    if matrix_type in ("nemeth12", "poli3") and n != FIXED_SIZE_N:
         pytest.skip(
             "Fixed matrix size for nemeth12 and poli3, skipping redundant runs."
         )
 
     if matrix_type == "spd":
-        A, b, lambda_min, lambda_max = MatrixGenerator.generate_matrix_and_vector(
+        matrix, b, lambda_min, lambda_max = MatrixGenerator.generate_matrix_and_vector(
             "spd", size=n
         )
     elif matrix_type == "poli3":
-        A, b, lambda_min, lambda_max = MatrixGenerator.generate_matrix_and_vector(
+        matrix, b, lambda_min, lambda_max = MatrixGenerator.generate_matrix_and_vector(
             "poli3"
         )
     elif matrix_type == "nemeth12":
-        A, b, lambda_min, lambda_max = MatrixGenerator.generate_matrix_and_vector(
+        matrix, b, lambda_min, lambda_max = MatrixGenerator.generate_matrix_and_vector(
             "nemeth12"
         )
     else:
-        raise ValueError(
-            "Invalid matrix type specified. Choose 'spd', 'poli3', or 'nemeth12'."
-        )
+        msg = "Invalid matrix type specified. Choose 'spd', 'poli3', or 'nemeth12'."
+        raise ValueError(msg)
 
     solution_richardson, info_richardson = None, None
 
     if processing_type != ProcessingType.THREADS:
         richardson_solver = RichardsonMethod(
             processing_type,
-            A,
-            b,
-            lambda_min,
-            lambda_max,
-            max_iterations,
-            size=A.shape[0],
-            tol=1e-7,
+            Problem(matrix, b, lambda_min, lambda_max),
+            Settings(max_iterations, tol=1e-7),
         )
         with capsys.disabled():
             solution_richardson, info_richardson = richardson_solver.solve()
     else:
         with capsys.disabled():
-            solution_richardson, info_richardson = RichardsonMethodThreads(
-                A, b, lambda_min, lambda_max, max_iterations, tol=1e-7
+            solution_richardson, info_richardson = richardson_method_threads(
+                Problem(matrix, b, lambda_min, lambda_max),
+                Settings(max_iterations, tol=1e-7),
             )
 
-    # Przechwytywanie wyjścia po solve
+            # Przechwytywanie wyjścia po solve
     captured = capsys.readouterr()
     logger.info("Captured output: %s", captured.out)
 
-    solution = solution_lib(A, b)
+    solution = solution_lib(matrix, b)
 
-    assert_converged(solution_richardson, info_richardson, solution, tolerance, A, n)
+    assert_converged(
+        (solution_richardson, info_richardson),
+        solution,
+        tolerance,
+        (matrix, n),
+    )
 
 
 def assert_converged(
-    solution_richardson, info_richardson, solution, tolerance, A, n
+    result: tuple[Vector, str | None],
+    solution: Vector,
+    tolerance: float,
+    problem: tuple[Matrix, int],
 ) -> None:
+    """Raise unless Richardson agrees with numpy within tolerance.
+
+    ``result`` is what a backend returned, ``problem`` is (matrix, n).
+    """
+    solution_richardson, info_richardson = result
+    matrix, n = problem
     if info_richardson == "Richardson method for those values will NOT converge":
-        numpy_norm = calcualte_norm_from_matrix_numpy(A, n)
+        numpy_norm = calcualte_norm_from_matrix_numpy(matrix, n)
         logger.info(
             "Numpy norm:  %s  Richardson norm:  %s", numpy_norm, solution_richardson
         )
-        assert False, "Richardson did not converge"
-    else:
-        difference = np.linalg.norm(solution_richardson - solution)
-        logger.info("Difference between Richardson and numpy solutions: %s", difference)
-        if difference < tolerance:
-            logger.info(
-                "Both Richardson and numpy method converged and calculated correct values."
-            )
-        else:
-            logger.info("Solution numpy:\n %s", solution)
-            logger.info("Solution Richardson:\n %s", solution_richardson)
-        assert difference < tolerance, (
-            f"The solutions are different! Difference: {difference:.8f}"
+        msg = "Richardson did not converge"
+        raise AssertionError(msg)
+    difference = np.linalg.norm(solution_richardson - solution)
+    logger.info("Difference between Richardson and numpy solutions: %s", difference)
+    if difference < tolerance:
+        logger.info(
+            "Both Richardson and numpy method converged and calculated correct values."
         )
+    else:
+        logger.info("Solution numpy:\n %s", solution)
+        logger.info("Solution Richardson:\n %s", solution_richardson)
+    if difference >= tolerance:
+        msg = f"The solutions are different! Difference: {difference:.8f}"
+        raise AssertionError(msg)
 
 
 if __name__ == "__main__":

Programming/PORR/code/threads_indep.py

@@ -1,9 +1,14 @@
+#!/usr/bin/env python3
+"""Threaded Richardson with njit-compiled kernels."""
+
 import gc
 import logging
 import time
 
-import linear_algebra_utils as linAlg
+import linear_algebra_utils as lin_alg
+import numpy as np
 from numba import njit, prange
+from richardson_problem import Problem, Settings
 from time_measurement import (
     longest_threads_time_accumulator,
     tests_time,
@@ -12,60 +17,89 @@ from time_measurement import (
 
 logger = logging.getLogger(__name__)
 
+# Called with numpy arrays here and with plain lists in the sequential
+# backend; the aliases name both.
+Matrix = np.ndarray | list[list[float]]
+Vector = np.ndarray | list[float]
+
 
 @njit(parallel=True)
-def numba_matrix_vector_multiply(A, input_x, Ax) -> None:
-    for i in prange(len(A)):
+def numba_matrix_vector_multiply(
+    matrix: Matrix, input_x: Vector, product: Vector
+) -> None:
+    """Matrix-vector product, njit-compiled."""
+    for i in prange(len(matrix)):
         acc = 0.0
         for j in range(len(input_x)):
-            acc += A[i][j] * input_x[j]
-        Ax[i] = acc
+            acc += matrix[i][j] * input_x[j]
+        product[i] = acc
 
 
 @njit(parallel=True)
-def numba_vector_vector_subtraction(b, Ax, residual) -> None:
+def numba_vector_vector_subtraction(
+    b: Vector, product: Vector, residual: Vector
+) -> None:
+    """Elementwise subtraction, njit-compiled."""
     for i in prange(len(b)):
-        residual[i] = b[i] - Ax[i]
+        residual[i] = b[i] - product[i]
 
 
 @njit(nopython=True)
-def numba_scalar_vector_multiply(omega, vector, result) -> None:
+def numba_scalar_vector_multiply(omega: float, vector: Vector, result: Vector) -> None:
+    """Scalar-vector product, njit-compiled."""
     omega_real = omega.real
     for i in range(len(vector)):
         result[i] = omega_real * vector[i]
 
 
 @njit(parallel=True)
-def numba_vector_vector_addition(input_x, vector, output_x) -> None:
+def numba_vector_vector_addition(
+    input_x: Vector, vector: Vector, output_x: Vector
+) -> None:
+    """Elementwise addition, njit-compiled."""
     for i in prange(len(input_x)):
         output_x[i] = input_x[i] + vector[i]
 
-
-# Funkcje z dekoratorem
-@time_measurement_longest(longest_threads_time_accumulator)
-def matrix_vector_multiply(A, input_x, Ax) -> None:
-    numba_matrix_vector_multiply(A, input_x, Ax)
+        # Funkcje z dekoratorem
 
 
 @time_measurement_longest(longest_threads_time_accumulator)
-def vector_vector_subtraction(b, Ax, residual) -> None:
-    numba_vector_vector_subtraction(b, Ax, residual)
+def matrix_vector_multiply(matrix: Matrix, input_x: Vector, product: Vector) -> None:
+    """Return the matrix-vector product."""
+    numba_matrix_vector_multiply(matrix, input_x, product)
 
 
 @time_measurement_longest(longest_threads_time_accumulator)
-def scalar_vector_multiply(omega, vector, result) -> None:
+def vector_vector_subtraction(b: Vector, product: Vector, residual: Vector) -> None:
+    """Subtract one vector from another, elementwise."""
+    numba_vector_vector_subtraction(b, product, residual)
+
+
+@time_measurement_longest(longest_threads_time_accumulator)
+def scalar_vector_multiply(omega: float, vector: Vector, result: Vector) -> None:
+    """Multiply every element of a vector by a scalar."""
     numba_scalar_vector_multiply(omega, vector, result)
 
 
 @time_measurement_longest(longest_threads_time_accumulator)
-def vector_vector_addition(input_x, vector, output_x) -> None:
+def vector_vector_addition(input_x: Vector, vector: Vector, output_x: Vector) -> None:
+    """Add two vectors elementwise."""
     numba_vector_vector_addition(input_x, vector, output_x)
 
+    # Metoda Richardson z obsługą wątków
 
-# Metoda Richardson z obsługą wątków
-def RichardsonMethodThreads(
-    A, b, lambda_min, lambda_max, max_iterations, x0=None, tol=1e-5
-):
+
+def richardson_method_threads(
+    problem: Problem, settings: Settings
+) -> tuple[Vector, str | None]:
+    """Run Richardson with one thread per chunk."""
+    matrix = problem.matrix
+    b = problem.rhs
+    lambda_min = problem.lambda_min
+    lambda_max = problem.lambda_max
+    max_iterations = settings.max_iterations
+    x0 = settings.x0
+    tol = settings.tol
     longest_threads_time_accumulator.hard_reset()
 
     gc.disable()
@@ -77,13 +111,13 @@ def RichardsonMethodThreads(
     omega = 2 / (lambda_min + lambda_max)
     n = len(b)
 
-    for iteration in range(max_iterations):
-        Ax = [0.0] * n
-        matrix_vector_multiply(A, x, Ax)
+    for _iteration in range(max_iterations):
+        product = [0.0] * n
+        matrix_vector_multiply(matrix, x, product)
         longest_threads_time_accumulator.save_lap_and_reset()
 
         residual = [0.0] * n
-        vector_vector_subtraction(b, Ax, residual)
+        vector_vector_subtraction(b, product, residual)
         longest_threads_time_accumulator.save_lap_and_reset()
 
         change_vector = [0.0] * n
@@ -96,7 +130,7 @@ def RichardsonMethodThreads(
 
         x = _x[:]
 
-        if linAlg.SequentialLinearAlgebraUtils.vector_norm(residual) < tol:
+        if lin_alg.SequentialLinearAlgebraUtils.vector_norm(residual) < tol:
             break
 
     end_time = time.perf_counter()

Programming/PORR/code/threads_indep_old.py

@@ -1,10 +1,16 @@
+#!/usr/bin/env python3
+"""Threaded Richardson, the earlier index-slicing version."""
+
 import gc
 import logging
 import multiprocessing
 import time
+from collections.abc import Callable
 from concurrent.futures import ThreadPoolExecutor
 
-import linear_algebra_utils as linAlg
+import linear_algebra_utils as lin_alg
+import numpy as np
+from richardson_problem import Problem, Settings
 from time_measurement import (
     longest_threads_time_accumulator,
     tests_time,
@@ -13,30 +19,89 @@ from time_measurement import (
 
 logger = logging.getLogger(__name__)
 
-
-@time_measurement_longest(longest_threads_time_accumulator)
-def matrix_vector_multiply(A, input_x, start, end, Ax) -> None:
-    Ax[start:end] = [sum(x * y for x, y in zip(row, input_x)) for row in A[start:end]]
+# Called with numpy arrays here and with plain lists in the sequential
+# backend; the aliases name both.
+Matrix = np.ndarray | list[list[float]]
+Vector = np.ndarray | list[float]
 
 
 @time_measurement_longest(longest_threads_time_accumulator)
-def vector_vector_subtraction(b, Ax, start, end, residual) -> None:
-    residual[start:end] = [x - y for x, y in zip(b[start:end], Ax[start:end])]
+def matrix_vector_multiply(
+    matrix: Matrix, input_x: Vector, start: int, end: int, product: Vector
+) -> None:
+    """Return the matrix-vector product."""
+    product[start:end] = [
+        sum(x * y for x, y in zip(row, input_x, strict=False))
+        for row in matrix[start:end]
+    ]
 
 
 @time_measurement_longest(longest_threads_time_accumulator)
-def scalar_vector_multiply(omega, vector, start, end, result) -> None:
+def vector_vector_subtraction(
+    b: Vector, product: Vector, start: int, end: int, residual: Vector
+) -> None:
+    """Subtract one vector from another, elementwise."""
+    residual[start:end] = [
+        x - y for x, y in zip(b[start:end], product[start:end], strict=False)
+    ]
+
+
+@time_measurement_longest(longest_threads_time_accumulator)
+def scalar_vector_multiply(
+    omega: float, vector: Vector, start: int, end: int, result: Vector
+) -> None:
+    """Multiply every element of a vector by a scalar."""
     result[start:end] = [omega * x for x in vector[start:end]]
 
 
 @time_measurement_longest(longest_threads_time_accumulator)
-def vector_vector_addition(input_x, vector, start, end, output_x) -> None:
-    output_x[start:end] = [x + y for x, y in zip(input_x[start:end], vector[start:end])]
+def vector_vector_addition(
+    input_x: Vector, vector: Vector, start: int, end: int, output_x: Vector
+) -> None:
+    """Add two vectors elementwise."""
+    output_x[start:end] = [
+        x + y for x, y in zip(input_x[start:end], vector[start:end], strict=False)
+    ]
 
 
-def RichardsonMethodThreads(
-    A, b, lambda_min, lambda_max, max_iterations, x0=None, tol=1e-5
-):
+def _run_chunked(
+    executor: ThreadPoolExecutor,
+    layout: tuple[int, int, int],
+    func: Callable[..., None],
+    args: tuple[object, ...],
+) -> None:
+    """Split a row range into per-thread chunks, submit them, and wait.
+
+    ``layout`` is (num_threads, n, chunk_size). The four stages of one
+    Richardson iteration each did this by hand, which is what made the
+    iteration too complex and too long for ruff.
+    """
+    num_threads, n, chunk_size = layout
+    futures = []
+    for i in range(num_threads):
+        start = i * chunk_size
+        end = n if i == num_threads - 1 else (i + 1) * chunk_size
+        futures.append(executor.submit(func, *args[:-1], start, end, args[-1]))
+    for future in futures:
+        future.result()
+    longest_threads_time_accumulator.save_lap_and_reset()
+
+
+def richardson_method_threads(
+    problem: Problem, settings: Settings
+) -> tuple[Vector, str | None]:
+    """Run Richardson with one thread per chunk.
+
+    The unpacking below is deliberate: the body predates Problem/Settings and
+    reads these as locals throughout.
+    """
+    matrix = problem.matrix
+    b = problem.rhs
+    lambda_min = problem.lambda_min
+    lambda_max = problem.lambda_max
+    max_iterations = settings.max_iterations
+    x0 = settings.x0
+    tol = settings.tol
     longest_threads_time_accumulator.hard_reset()
 
     gc.disable()
@@ -48,89 +113,36 @@ def RichardsonMethodThreads(
     omega = 2 / (lambda_min + lambda_max)
     num_threads = multiprocessing.cpu_count()
     chunk_size = n // num_threads
+    layout = (num_threads, n, chunk_size)
 
     with ThreadPoolExecutor(
         max_workers=num_threads
     ) as executor:  # wątki są tworzone raz i nie są niszczone
-        for iteration in range(max_iterations):
-            Ax = [0] * len(
-                x
-            )  # tutaj zostanie przypisany wynik z mnożenia macierzy A z wektorem x
-            futures = []
-
-            for i in range(num_threads):
-                start = i * chunk_size
-                end = n if i == num_threads - 1 else (i + 1) * chunk_size
-                futures.append(
-                    executor.submit(matrix_vector_multiply, A, x, start, end, Ax)
-                )
-
-            for future in futures:
-                future.result()
-
-            longest_threads_time_accumulator.save_lap_and_reset()
-            residual = [0] * len(
-                b
-            )  # tutaj zostanie przypisany wynik z vector_vector_subtraction
-            futures = []
-
-            for i in range(num_threads):
-                start = i * chunk_size
-                end = n if i == num_threads - 1 else (i + 1) * chunk_size
-                futures.append(
-                    executor.submit(
-                        vector_vector_subtraction, b, Ax, start, end, residual
-                    )
-                )
-
-            for future in futures:
-                future.result()
-
-            longest_threads_time_accumulator.save_lap_and_reset()
-            change_vector = [0] * len(
-                residual
-            )  # zostanie tu przypisany wynik scalar_vector_multiply po pracy wątków
-            futures = []
-
-            for i in range(num_threads):
-                start = i * chunk_size
-                end = n if i == num_threads - 1 else (i + 1) * chunk_size
-                futures.append(
-                    executor.submit(
-                        scalar_vector_multiply,
-                        omega,
-                        residual,
-                        start,
-                        end,
-                        change_vector,
-                    )
-                )
-
-            for future in futures:
-                future.result()
-
-            longest_threads_time_accumulator.save_lap_and_reset()
-            _x = x[
-                :
-            ]  # do _x zostanie przez wątki przypisany wynik pracy w danej iteracji
-            futures = []
-
-            for i in range(num_threads):
-                start = i * chunk_size
-                end = n if i == num_threads - 1 else (i + 1) * chunk_size
-                futures.append(
-                    executor.submit(
-                        vector_vector_addition, x, change_vector, start, end, _x
-                    )
-                )
-
-            for future in futures:
-                future.result()
-
-            longest_threads_time_accumulator.save_lap_and_reset()
+        for _iteration in range(max_iterations):
+            # holds the result of multiplying A by x
+            product = [0] * len(x)
+            _run_chunked(executor, layout, matrix_vector_multiply, (matrix, x, product))
+            # holds the result of b - Ax
+            residual = [0] * len(b)
+            _run_chunked(
+                executor, layout, vector_vector_subtraction, (b, product, residual)
+            )
+            # holds omega * residual
+            change_vector = [0] * len(residual)
+            _run_chunked(
+                executor,
+                layout,
+                scalar_vector_multiply,
+                (omega, residual, change_vector),
+            )
+            # the threads write this iteration's result into _x
+            _x = x[:]
+            _run_chunked(
+                executor, layout, vector_vector_addition, (x, change_vector, _x)
+            )
             x = _x[:]
 
-            if linAlg.SequentialLinearAlgebraUtils.vector_norm(residual) < tol:
+            if lin_alg.SequentialLinearAlgebraUtils.vector_norm(residual) < tol:
                 break
 
     end_time = time.perf_counter()

Programming/PORR/code/threads_too_simple.py

@@ -1,3 +1,6 @@
+#!/usr/bin/env python3
+"""Threaded Richardson, the first attempt (one thread per slice)."""
+
 import gc
 import logging
 import multiprocessing
@@ -6,6 +9,7 @@ import threading
 import time
 
 import numpy as np
+from richardson_problem import Problem, Settings
 from time_measurement import (
     longest_time_accumulator,
     tests_time,
@@ -14,17 +18,40 @@ from time_measurement import (
 
 logger = logging.getLogger(__name__)
 
+# Called with numpy arrays here and with plain lists in the sequential
+# backend; the aliases name both.
+Matrix = np.ndarray | list[list[float]]
+Vector = np.ndarray | list[float]
+
 
 @time_measurement_longest(longest_time_accumulator)
-def RichardsonThread(A, b, x, _x, omega, start, end) -> None:
+def richardson_thread(
+    matrix: Matrix,
+    b: Vector,
+    vectors: tuple[Vector, Vector],
+    omega: float,
+    span: tuple[int, int],
+) -> None:
+    """Run one thread's slice of a Richardson iteration.
+
+    ``vectors`` is (x, next_x) and ``span`` is the half-open row range this
+    thread owns.
+    """
+    x, _x = vectors
+    start, end = span
     for i in range(start, end):
-        sigma = np.dot(A[i, :], _x) - A[i, i] * _x[i]
-        x[i] = (1 - omega) * _x[i] + omega * (b[i] - sigma) / A[i, i]
+        sigma = np.dot(matrix[i, :], _x) - matrix[i, i] * _x[i]
+        x[i] = (1 - omega) * _x[i] + omega * (b[i] - sigma) / matrix[i, i]
 
 
-def RichardsonMethodThreads(
-    A, b, lambda_min, lambda_max, max_iterations, x0=None, tol=1e-5
-):
+def richardson_method_threads(
+    problem: Problem, settings: Settings
+) -> tuple[Vector, str | None]:
+    """Run Richardson with one thread per chunk."""
+    matrix = problem.matrix
+    b = problem.rhs
+    max_iterations = settings.max_iterations
+    x0 = settings.x0
     longest_time_accumulator.total_time = 0
     longest_time_accumulator.start = sys.float_info.max
     longest_time_accumulator.end = 0
@@ -44,21 +71,18 @@ def RichardsonMethodThreads(
     for _ in range(max_iterations):
         _x = x[:]
         for i in range(num_threads):
-            start = (
-                i * chunk_size
-            )  # start jest indeksem w A. Wątki otrzymują kolejny punkt startowy będący wielokrotnością rozmiaru porcji na wątek
+            # start is an index into A: each thread gets the next multiple of
+            # the per-thread chunk size as its starting point.
+            start = i * chunk_size
             end = n if i == num_threads - 1 else (i + 1) * chunk_size
             thread = threading.Thread(
-                target=RichardsonThread, args=(A, b, x, _x, omega, start, end)
+                target=richardson_thread, args=(matrix, b, (x, _x), omega, (start, end))
             )
             threads.append(thread)
             thread.start()
 
         for thread in threads:
             thread.join()
-
-        #         break
-
     end_time = time.perf_counter()
     gc.enable()
     total_time = end_time - start_time

Programming/PORR/code/time_measurement.py

@@ -1,27 +1,39 @@
+#!/usr/bin/env python3
+"""Decorators that accumulate how long the measured calls take."""
+
 import sys
 import time
-from functools import wraps
+from collections.abc import Callable
 
 
 class TimeAccumulator:
+    """Total time spent inside the decorated functions."""
+
     def __init__(self) -> None:
+        """Set up the object."""
         self.total_time = 0
 
 
 class ComplexTimeAcumulator:
+    """Per-lap and longest times for the decorated functions."""
+
     def __init__(self) -> None:
+        """Set up the object."""
         self.hard_reset()
 
     def hard_reset(self) -> None:
+        """Clear every recorded time."""
         self.total_time = 0
         self.reset()
 
     def reset(self) -> None:
+        """Clear the current lap."""
         self.lap_time = 0
         self.start = sys.float_info.max
         self.end = 0
 
     def save_lap_and_reset(self) -> None:
+        """Store the current lap and start a new one."""
         self.total_time += self.lap_time
         self.reset()
 
@@ -32,10 +44,14 @@ tests_time = TimeAccumulator()
 longest_threads_time_accumulator = ComplexTimeAcumulator()
 
 
-def time_measurement(accumulator: TimeAccumulator):
-    def decorator(func):
-        @wraps(func)
-        def inner(*args, **kwargs):
+def time_measurement(accumulator: TimeAccumulator) -> Callable[..., object]:
+    """Decorate a function so its runtime accumulates."""
+
+    def decorator(func: Callable[..., object]) -> Callable[..., object]:
+        """Wrap the function so its runtime is recorded."""
+
+        def inner(*args: object, **kwargs: object) -> object:
+            """Call the wrapped function and record how long it took."""
             start = time.perf_counter()
             result = func(*args, **kwargs)
             end = time.perf_counter()
@@ -47,10 +63,16 @@ def time_measurement(accumulator: TimeAccumulator):
     return decorator
 
 
-def time_measurement_longest(accumulator: ComplexTimeAcumulator):
-    def decorator(func):
-        @wraps(func)
-        def inner(*args, **kwargs):
+def time_measurement_longest(
+    accumulator: ComplexTimeAcumulator,
+) -> Callable[..., object]:
+    """Decorate a function so its longest runtime is kept."""
+
+    def decorator(func: Callable[..., object]) -> Callable[..., object]:
+        """Wrap the function so its runtime is recorded."""
+
+        def inner(*args: object, **kwargs: object) -> object:
+            """Call the wrapped function and record how long it took."""
             start = time.perf_counter()
             result = func(*args, **kwargs)
             end = time.perf_counter()