nmod_mat continued

fredrik-johansson · Jan 14, 2025 · 1afaa38 · 1afaa38
1 parent a8a16ae
commit 1afaa38
Show file tree

Hide file tree

Showing 10 changed files with 89 additions and 72 deletions.
diff --git a/src/nmod_mat/charpoly.c b/src/nmod_mat/charpoly.c
@@ -181,15 +181,10 @@ void nmod_mat_charpoly_danilevsky(nmod_poly_t p, const nmod_mat_t M)
             }
          } else
          {
-            ulong * ptr;
-            ulong t;
-
             nmod_mat_swap_rows(M2, NULL, n - i - k - 1, n - i - 1);
 
             for (j = 1; j <= n - i + 1; j++)
-            {
                 FLINT_SWAP(ulong, A(j - 1, n - i - k - 1), A(j - 1, n - i - 1));
-            }
          }
 
          h = A(n - i, n - i - 1);

diff --git a/src/nmod_mat/det_howell.c b/src/nmod_mat/det_howell.c
@@ -55,7 +55,6 @@ static inline int
 _nmod_mat_pivot(nmod_mat_t A, slong start_row, slong col)
 {
     slong j;
-    nn_ptr u;
 
     if (nmod_mat_entry(A, start_row, col) != 0)
         return 1;

diff --git a/src/nmod_mat/lu_classical.c b/src/nmod_mat/lu_classical.c
@@ -16,8 +16,7 @@
 static inline int
 nmod_mat_pivot(nmod_mat_t A, slong * P, slong start_row, slong col)
 {
-    slong j, t;
-    nn_ptr u;
+    slong j;
 
     if (nmod_mat_entry(A, start_row, col) != 0)
         return 1;

diff --git a/src/nmod_mat/lu_classical_delayed.c b/src/nmod_mat/lu_classical_delayed.c
@@ -26,9 +26,8 @@ nmod_mat_lu_classical_delayed_1(slong * P, nmod_mat_t A, int rank_check)
 {
     ulong d, e, f, *aa;
     nmod_t mod;
-    slong i, j, nrows, ncols, rank, row, col, pivot_row, tmp_index;
+    slong i, j, nrows, ncols, rank, row, col, pivot_row;
     slong stride = A->stride;
-    nn_ptr tnn_ptr;
 
     nrows = A->r;
     ncols = A->c;
@@ -121,9 +120,8 @@ nmod_mat_lu_classical_delayed_2(slong * P, nmod_mat_t A, int rank_check)
 {
     ulong d, e, f, *aa;
     nmod_t mod;
-    slong i, j, nrows, ncols, rank, row, col, pivot_row, tmp_index;
+    slong i, j, nrows, ncols, rank, row, col, pivot_row;
     slong stride = A->stride;
-    nn_ptr tnn_ptr;
     nn_ptr b;
     TMP_INIT;
 
@@ -290,9 +288,8 @@ nmod_mat_lu_classical_delayed_3(slong * P, nmod_mat_t A, int rank_check)
 {
     ulong d, e, f, *aa;
     nmod_t mod;
-    slong i, j, nrows, ncols, rank, row, col, pivot_row, tmp_index;
+    slong i, j, nrows, ncols, rank, row, col, pivot_row;
     slong stride = A->stride;
-    nn_ptr tnn_ptr;
     nn_ptr b;
     TMP_INIT;
 

diff --git a/src/nmod_mat/lu_recursive.c b/src/nmod_mat/lu_recursive.c
@@ -1,6 +1,6 @@
 /*
     Copyright (C) 2011 Fredrik Johansson
-
+g
     This file is part of FLINT.
 
     FLINT is free software: you can redistribute it and/or modify it under
@@ -13,8 +13,7 @@
 #include "nmod_mat.h"
 
 static void
-_apply_permutation_P(slong * AP, nmod_mat_t A, slong * P,
-    slong n, slong offset)
+_apply_permutation_P(slong * AP, slong * P, slong n, slong offset)
 {
     if (n != 0)
     {
@@ -31,19 +30,22 @@ _apply_permutation_P(slong * AP, nmod_mat_t A, slong * P,
 }
 
 static void
-_apply_permutation_A(slong * AP, nmod_mat_t A, slong * P,
-    slong n, slong col_offset)
+_apply_permutation_A(nmod_mat_t A, slong * P,
+    slong num_rows, slong row_offset, slong num_cols, slong col_offset)
 {
-    if (n != 0)
+    if (num_rows != 0)
     {
         ulong * Atmp;
         slong i;
-        slong c = A->c - col_offset;
 
-        Atmp = flint_malloc(sizeof(ulong) * n * c);
+        /* todo: reduce memory allocation */
+        Atmp = flint_malloc(sizeof(ulong) * num_rows * num_cols);
+
+        for (i = 0; i < num_rows; i++)
+            _nmod_vec_set(Atmp + i * num_cols, nmod_mat_entry_ptr(A, P[i] + row_offset, col_offset), num_cols);
 
-        for (i = 0; i < n; i++) _nmod_vec_set(Atmp + i * c, nmod_mat_entry_ptr(A, P[i], col_offset), c);
-        for (i = 0; i < n; i++) _nmod_vec_set(nmod_mat_entry_ptr(A, i, col_offset), Atmp + i * c, c);
+        for (i = 0; i < num_rows; i++)
+            _nmod_vec_set(nmod_mat_entry_ptr(A, i + row_offset, col_offset), Atmp + i * num_cols, num_cols);
 
         flint_free(Atmp);
     }
@@ -85,8 +87,8 @@ nmod_mat_lu_recursive(slong * P, nmod_mat_t A, int rank_check)
 
     if (r1 != 0)
     {
-        _apply_permutation_A(P, A, P1, m, 0);
-        _apply_permutation_P(P, A, P1, m, 0);
+        _apply_permutation_A(A, P1, m, 0, n - n1, n1);
+        _apply_permutation_P(P, P1, m, 0);
     }
 
     nmod_mat_window_init(A00, A, 0, 0, r1, r1);
@@ -108,7 +110,8 @@ nmod_mat_lu_recursive(slong * P, nmod_mat_t A, int rank_check)
     }
     else
     {
-        _apply_permutation_P(P, A, P1, m - r1, r1);
+        _apply_permutation_A(A, P1, m - r1, r1, n1, 0);
+        _apply_permutation_P(P, P1, m - r1, r1);
 
         /* Compress L */
         if (r1 != n1)

diff --git a/src/nmod_mat/mul.c b/src/nmod_mat/mul.c
@@ -88,7 +88,7 @@ nmod_mat_mul(nmod_mat_t C, const nmod_mat_t A, const nmod_mat_t B)
 
     if (flint_num_threads > 1)
 	    nmod_mat_mul_classical_threaded(C, A, B);
-    else if (min_dim < cutoff || 1)
+    else if (min_dim < cutoff)
         nmod_mat_mul_classical(C, A, B);
     else
         nmod_mat_mul_strassen(C, A, B);