1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
use crate::{
RLNCError,
common::{
gf256::Gf256,
simd::{gf256_inplace_mul_vec_by_scalar, gf256_mul_vec_by_scalar_then_add_into_vec},
},
};
use std::ops::{Index, IndexMut};
#[derive(Clone, Debug, PartialEq)]
pub struct DecoderMatrix {
num_pieces_coded_together: usize,
rows: usize,
cols: usize,
elements: Vec<u8>,
}
impl DecoderMatrix {
/// Given RLNC encoding configuration, it sets up a decoder matrix.
///
/// This decoder matrix can be used to add incoming erasure-coded pieces,
/// and incrementally decode them using Gaussian Elimination, if it's a
/// useful (i.e. linearly independent) piece.
///
/// # Arguments
/// * `num_pieces_coded_together` - The minimum number of useful coded pieces needed for decoding.
/// * `piece_byte_length` - The byte length of each original data piece.
///
/// # Returns
/// An instance of decoder matrix - ready to use for decoding.
pub fn new(num_pieces_coded_together: usize, piece_byte_length: usize) -> Self {
let full_coded_piece_byte_len = num_pieces_coded_together + piece_byte_length;
let total_byte_len = num_pieces_coded_together * full_coded_piece_byte_len;
let elements = Vec::with_capacity(total_byte_len);
Self {
num_pieces_coded_together,
rows: 0,
cols: full_coded_piece_byte_len,
elements,
}
}
/// Adds a new row to the decoder matrix.
///
/// # Arguments
/// `row` - A byte slice, representing a full erasure-coded piece i.e. containing the coefficients followed by
/// the coded data for one piece. Its length must be `num_pieces_coded_together + piece_byte_length`.
///
/// # Returns
/// * Ok(&mut Self) - If full erasure-coded piece is of valid length.
/// * Err(RLNCError::InvalidPieceLength) - If full erasure-coded piece length doesn't match expected value.
pub fn add_row(&mut self, row: &[u8]) -> Result<&mut Self, RLNCError> {
if row.len() != self.cols {
return Err(RLNCError::InvalidPieceLength);
}
self.elements.extend_from_slice(row);
self.rows += 1;
Ok(self)
}
/// Swaps two rows in the decoder's matrix.
///
/// # Arguments
/// * `row1_idx` - The index of the first row.
/// * `row2_idx` - The index of the second row.
pub fn swap_rows(&mut self, row1_idx: usize, row2_idx: usize) -> &mut Self {
if row1_idx == row2_idx {
return self;
}
let (r1, r2) = if row1_idx < row2_idx { (row1_idx, row2_idx) } else { (row2_idx, row1_idx) };
let start1 = r1 * self.cols;
let end1 = start1 + self.cols;
let start2 = r2 * self.cols;
let (left, right) = self.elements.split_at_mut(start2);
// row1 is in the first part
let row1 = &mut left[start1..end1];
// row2 is the beginning of the second part
let row2 = &mut right[..self.cols];
row1.swap_with_slice(row2);
self
}
/// Computes the Reduced Row Echelon Form (RREF) of the matrix.
///
/// This involves forward elimination (`Self::clean_forward`), backward elimination
/// (`Self::clean_backward`), and removing any resulting zero rows (`Self::remove_zero_rows`).
///
/// This function updates the number of rows to reflect the current rank of the matrix.
/// It is safe to call `Self::rank` after calling this function.
pub fn rref(&mut self) -> &mut Self {
self.clean_forward().clean_backward().remove_zero_rows()
}
/// Returns the current rank of the matrix, which is same as the number
/// of rows, after calling `Self::rref`.
pub fn rank(&self) -> usize {
self.rows
}
/// Returns underlying data i.e. `self.rows` many full erasure-coded pieces.
/// Calling this function, consumes the decoder matrix instance.
pub fn extract_data(self) -> Vec<u8> {
self.elements
}
/// Performs the forward phase of Gaussian elimination (to row echelon form).
///
/// Pivots are selected, rows are swapped if necessary to get a non-zero
/// pivot, and rows below the pivot are cleared by subtracting a multiple
/// of the pivot row.
fn clean_forward(&mut self) -> &mut Self {
let boundary = self.rows.min(self.cols);
for i in 0..boundary {
if self[(i, i)] == Gf256::zero() {
let mut is_non_zero_col = false;
let mut pivot_row_idx = i + 1;
while pivot_row_idx < self.rows {
if self[(pivot_row_idx, i)] != Gf256::zero() {
is_non_zero_col = true;
break;
}
pivot_row_idx += 1;
}
if !is_non_zero_col {
continue;
}
self.swap_rows(i, pivot_row_idx);
}
for j in (i + 1)..self.rows {
if self[(j, i)] == Gf256::zero() {
continue;
}
let quotient = unsafe { (self[(j, i)] / self[(i, i)]).unwrap_unchecked().get() };
let i_th_row_starts_at = i * self.cols;
let i_th_row_ends_at = i_th_row_starts_at + self.cols;
let j_th_row_starts_at = j * self.cols;
let j_th_row_ends_at = j_th_row_starts_at + self.cols;
let (left, right) = self.elements.split_at_mut(i_th_row_ends_at);
let i_th_row = &left[(i_th_row_starts_at + i)..];
let j_th_row = &mut right[(j_th_row_starts_at - i_th_row_ends_at + i)..(j_th_row_ends_at - i_th_row_ends_at)];
gf256_mul_vec_by_scalar_then_add_into_vec(j_th_row, i_th_row, quotient);
}
}
self
}
/// Performs the backward phase of Gaussian elimination (to reduced row echelon form).
///
/// Clears entries above the pivots and normalizes pivots to 1.
fn clean_backward(&mut self) -> &mut Self {
let boundary = self.rows.min(self.cols);
for i in (0..boundary).rev() {
if self[(i, i)] == Gf256::zero() {
continue;
}
for j in 0..i {
if self[(j, i)] == Gf256::zero() {
continue;
}
let quotient = unsafe { (self[(j, i)] / self[(i, i)]).unwrap_unchecked().get() };
let j_th_row_starts_at = j * self.cols;
let j_th_row_ends_at = j_th_row_starts_at + self.cols;
let i_th_row_starts_at = i * self.cols;
let i_th_row_ends_at = i_th_row_starts_at + self.cols;
let (left, right) = self.elements.split_at_mut(j_th_row_ends_at);
let j_th_row = &mut left[(j_th_row_starts_at + i)..];
let i_th_row = &right[(i_th_row_starts_at - j_th_row_ends_at + i)..(i_th_row_ends_at - j_th_row_ends_at)];
gf256_mul_vec_by_scalar_then_add_into_vec(j_th_row, i_th_row, quotient);
}
if self[(i, i)] == Gf256::one() {
continue;
}
let inv = unsafe { self[(i, i)].inv().unwrap_unchecked().get() };
self[(i, i)] = Gf256::one();
let i_th_row_starts_at = i * self.cols;
let i_th_row_ends_at = i_th_row_starts_at + self.cols;
let i_th_row = &mut self.elements[(i_th_row_starts_at + (i + 1))..i_th_row_ends_at];
gf256_inplace_mul_vec_by_scalar(i_th_row, inv);
}
self
}
/// Removes zero rows from the matrix and updates `useful_piece_count`.
///
/// A row is considered a zero row if all its coefficient columns are zero.
/// This step is crucial after RREF to determine the true rank and compact
/// the matrix to only the useful rows.
fn remove_zero_rows(&mut self) -> &mut Self {
let mut i = 0;
while i < self.rows {
let is_nonzero_row = (0..self.num_pieces_coded_together).any(|cidx| self[(i, cidx)] != Gf256::zero());
if is_nonzero_row {
i += 1;
continue;
}
let start_idx_of_row_to_remove = i * self.cols;
let start_idx_of_next_row = (i + 1) * self.cols;
if start_idx_of_next_row < self.elements.len() {
self.elements.copy_within(start_idx_of_next_row.., start_idx_of_row_to_remove);
}
self.rows -= 1;
}
let updated_num_elements = self.rows * self.cols;
self.elements.truncate(updated_num_elements);
self
}
}
impl Index<(usize, usize)> for DecoderMatrix {
type Output = Gf256;
/// Returns an immutable reference to an element of matrix at the specified row and column,
/// converting it to a `Gf256` element.
///
/// # Arguments
/// * `index` - A tuple `(row_index, col_index)` specifying the position.
///
/// # Returns
/// Returns the element as a `Gf256`.
///
/// # Panics
/// Panics if the index is out of bounds.
fn index(&self, index: (usize, usize)) -> &Self::Output {
let (row_idx, col_idx) = index;
let lin_idx = row_idx * self.cols + col_idx;
unsafe { std::mem::transmute(self.elements.get_unchecked(lin_idx)) }
}
}
impl IndexMut<(usize, usize)> for DecoderMatrix {
/// Returns a mutable reference to an element of matrix at the specified row and column,
/// converting it to a `Gf256` element.
///
/// # Arguments
/// * `index` - A tuple `(row_index, col_index)` specifying the position.
/// * `val` - The `Gf256` value to set.
///
/// # Panics
/// Panics if the index is out of bounds.
fn index_mut(&mut self, index: (usize, usize)) -> &mut Self::Output {
let (row_idx, col_idx) = index;
let lin_idx = row_idx * self.cols + col_idx;
unsafe { std::mem::transmute(self.elements.get_unchecked_mut(lin_idx)) }
}
}
#[cfg(test)]
mod test {
use crate::full::decoder_matrix::DecoderMatrix;
use rand::Rng;
fn make_random_matrix<R: Rng + ?Sized>(num_rows: usize, num_cols: usize, rng: &mut R) -> DecoderMatrix {
let mut matrix = DecoderMatrix::new(num_cols, 0);
(0..num_rows).for_each(|_| {
let random_row = (0..num_cols).map(|_| rng.random()).collect::<Vec<u8>>();
matrix.add_row(&random_row).expect("adding new must not fail");
});
matrix
}
#[test]
fn prop_test_rref_is_idempotent() {
const NUM_TEST_ITERATIONS: usize = 1000;
const MIN_NUM_ROWS: usize = 1;
const MAX_NUM_ROWS: usize = 1000;
const MIN_NUM_COLS: usize = 1;
const MAX_NUM_COLS: usize = 1000;
let mut rng = rand::rng();
(0..NUM_TEST_ITERATIONS).for_each(|_| {
let num_rows = rng.random_range(MIN_NUM_ROWS..=MAX_NUM_ROWS);
let num_cols = rng.random_range(MIN_NUM_COLS..=MAX_NUM_COLS);
let mut matrix = make_random_matrix(num_rows, num_cols, &mut rng);
let rrefed = matrix.rref().clone().rref().to_owned();
assert_eq!(matrix, rrefed);
});
}
#[test]
fn test_swap_rows() {
// Setup a deterministic 4x5 matrix for testing
let num_pieces = 3;
let piece_len = 2;
let mut matrix = DecoderMatrix::new(num_pieces, piece_len);
// Initial matrix content:
// Row 0: [1, 1, 1, 10, 10]
// Row 1: [2, 2, 2, 20, 20]
// Row 2: [3, 3, 3, 30, 30]
// Row 3: [4, 4, 4, 40, 40]
matrix.add_row(&[1, 1, 1, 10, 10]).unwrap();
matrix.add_row(&[2, 2, 2, 20, 20]).unwrap();
matrix.add_row(&[3, 3, 3, 30, 30]).unwrap();
matrix.add_row(&[4, 4, 4, 40, 40]).unwrap();
// Test Case 1: Standard swap (row1 < row2)
let mut matrix_case_1 = matrix.clone();
matrix_case_1.swap_rows(0, 2);
let mut expected_1 = DecoderMatrix::new(num_pieces, piece_len);
expected_1.add_row(&[3, 3, 3, 30, 30]).unwrap(); // Swapped from row 2
expected_1.add_row(&[2, 2, 2, 20, 20]).unwrap();
expected_1.add_row(&[1, 1, 1, 10, 10]).unwrap(); // Swapped from row 0
expected_1.add_row(&[4, 4, 4, 40, 40]).unwrap();
assert_eq!(matrix_case_1, expected_1, "Failed standard swap (0, 2)");
// Test Case 2: Reverse order swap (row1 > row2)
let mut matrix_case_2 = matrix.clone();
matrix_case_2.swap_rows(3, 1);
let mut expected_2 = DecoderMatrix::new(num_pieces, piece_len);
expected_2.add_row(&[1, 1, 1, 10, 10]).unwrap();
expected_2.add_row(&[4, 4, 4, 40, 40]).unwrap(); // Swapped from row 3
expected_2.add_row(&[3, 3, 3, 30, 30]).unwrap();
expected_2.add_row(&[2, 2, 2, 20, 20]).unwrap(); // Swapped from row 1
assert_eq!(matrix_case_2, expected_2, "Failed reverse order swap (3, 1)");
// Test Case 3: Identity swap (row1 == row2)
let mut matrix_case_3 = matrix.clone();
matrix_case_3.swap_rows(1, 1);
// The matrix should be unchanged
assert_eq!(matrix_case_3, matrix, "Failed identity swap (1, 1)");
// Test Case 4: Swap first and last rows
let mut matrix_case_4 = matrix.clone();
matrix_case_4.swap_rows(0, 3);
let mut expected_4 = DecoderMatrix::new(num_pieces, piece_len);
expected_4.add_row(&[4, 4, 4, 40, 40]).unwrap(); // Swapped from row 3
expected_4.add_row(&[2, 2, 2, 20, 20]).unwrap();
expected_4.add_row(&[3, 3, 3, 30, 30]).unwrap();
expected_4.add_row(&[1, 1, 1, 10, 10]).unwrap(); // Swapped from row 0
assert_eq!(matrix_case_4, expected_4, "Failed swap of first and last rows (0, 3)");
}
}