gf2matrix *invert_matrix(gf2matrix *dest, const gf2matrix *m)
{
if (!m)
return NULL;
mzd_t *temp = mzd_inv_m4ri(NULL, (mzd_t*) m, 0);
if(!temp) {
dest = NULL;
}
else {
mzd_t *I = make_identity_matrix(m->nrows);
mzd_t *prod = mul_matrices(NULL, m, temp);
if(comp_matrices(prod, I) == 0) {
// inversion was successful
if (dest) {
mzd_copy(dest, temp);
mzd_free(temp);
}
else {
dest = temp;
}
}
else {
dest = NULL;
}
mzd_free(prod);
mzd_free(I);
}
return dest;
}
int test_pluq_structured(rci_t m, rci_t n) {
printf("pluq: testing structured m: %5d, n: %5d", m, n);
mzd_t* A = mzd_init(m, n);
mzd_t* L = mzd_init(m, m);
mzd_t* U = mzd_init(m, n);
for(rci_t i = 0; i < m; i += 2)
for (rci_t j = i; j < n; ++j)
mzd_write_bit(A, i, j, 1);
mzd_t* Acopy = mzd_copy (NULL,A);
mzp_t* P = mzp_init(m);
mzp_t* Q = mzp_init(n);
rci_t r = mzd_pluq(A, P, Q, 0);
printf(", rank: %5d ",r);
for (rci_t i = 0; i < r; ++i){
for (rci_t j = 0; j < i; ++j)
mzd_write_bit(L, i, j, mzd_read_bit(A,i,j));
for (rci_t j = i + 1; j < n; ++j)
mzd_write_bit(U, i, j, mzd_read_bit(A,i,j));
}
for (rci_t i = r; i < m; ++i)
for (rci_t j = 0; j < r; ++j)
mzd_write_bit(L, i, j, mzd_read_bit(A,i,j));
for (rci_t i = 0; i < r; ++i){
mzd_write_bit(L,i,i, 1);
mzd_write_bit(U,i,i, 1);
}
mzd_apply_p_left(Acopy, P);
mzd_apply_p_right_trans(Acopy, Q);
mzd_addmul(Acopy, L, U, 0);
int status = 0;
for (rci_t i = 0; i < m; ++i)
for (rci_t j = 0; j < n; ++j){
if (mzd_read_bit (Acopy,i,j)){
status = 1;
break;
}
}
if (status) {
printf("\n");
printf(" ... FAILED\n");
} else
printf (" ... passed\n");
mzd_free(U);
mzd_free(L);
mzd_free(A);
mzd_free(Acopy);
mzp_free(P);
mzp_free(Q);
return status;
}
int test_trsm_upper_right (int m, int n, int offset, const char* description){
printf("upper_right: %s m: %4d n: %4d offset: %4d ... ",description, m, n, offset);
mzd_t* Ubase = mzd_init (2048,2048);
mzd_t* Bbase = mzd_init (2048,2048);
mzd_randomize (Ubase);
mzd_randomize (Bbase);
mzd_t* Bbasecopy = mzd_copy (NULL, Bbase);
mzd_t* U = mzd_init_window (Ubase, 0, offset, n, offset + n);
mzd_t* B = mzd_init_window (Bbase, 0, offset, m, offset + n);
mzd_t* W = mzd_copy (NULL, B);
size_t i,j;
for (i=0; i<n; ++i){
for (j=0; j<i;++j)
mzd_write_bit(U,i,j, 0);
mzd_write_bit(U,i,i, 1);
}
mzd_trsm_upper_right (U, B, 2048);
mzd_addmul(W, B, U, 2048);
int status = 0;
for ( i=0; i<m; ++i)
for ( j=0; j<n; ++j){
if (mzd_read_bit (W,i,j)){
status = 1;
}
}
// Verifiying that nothing has been changed around the submatrices
mzd_addmul(W, B, U, 2048);
mzd_copy (B, W);
for ( i=0; i<2048; ++i)
for ( j=0; j<2048/RADIX; ++j){
if (Bbase->rows[i][j] != Bbasecopy->rows[i][j]){
status = 1;
}
}
mzd_free_window (U);
mzd_free_window (B);
mzd_free (W);
mzd_free(Ubase);
mzd_free(Bbase);
mzd_free(Bbasecopy);
if (!status)
printf("passed\n");
else
printf("FAILED\n");
return status;
}
int test_trsm_lower_left (int m, int n, int offsetL, int offsetB){
mzd_t* Lbase = mzd_init (2048,2048);
mzd_t* Bbase = mzd_init (2048,2048);
mzd_randomize (Lbase);
mzd_randomize (Bbase);
mzd_t* Bbasecopy = mzd_copy (NULL, Bbase);
mzd_t* L = mzd_init_window (Lbase, 0, offsetL, m, offsetL + m);
mzd_t* B = mzd_init_window (Bbase, 0, offsetB, m, offsetB + n);
mzd_t* W = mzd_copy (NULL, B);
size_t i,j;
for (i=0; i<m; ++i){
for (j=i+1; j<m;++j)
mzd_write_bit(L,i,j, 0);
mzd_write_bit(L,i,i, 1);
}
mzd_trsm_lower_left(L, B, 2048);
mzd_addmul(W, L, B, 2048);
int status = 0;
for ( i=0; i<m; ++i)
for ( j=0; j<n; ++j){
if (mzd_read_bit (W,i,j)){
status = 1;
}
}
// Verifiying that nothing has been changed around the submatrices
mzd_addmul(W, L, B, 2048);
mzd_copy (B, W);
for ( i=0; i<2048; ++i)
for ( j=0; j<2048/RADIX; ++j){
if (Bbase->rows[i][j] != Bbasecopy->rows[i][j]){
status = 1;
}
}
mzd_free_window (L);
mzd_free_window (B);
mzd_free_window (W);
mzd_free(Lbase);
mzd_free(Bbase);
mzd_free(Bbasecopy);
if (!status)
printf(" ... passed\n");
else
printf(" ... FAILED\n");
return status;
}
/**
* Check that the results of all implemented multiplication algorithms
* match up.
*
* \param m Number of rows of A
* \param l Number of columns of A/number of rows of B
* \param n Number of columns of B
* \param k Parameter k of M4RM algorithm, may be 0 for automatic choice.
* \param cutoff Cut off parameter at which dimension to switch from
* Strassen to M4RM
*/
int mul_test_equality(rci_t m, rci_t l, rci_t n, int k, int cutoff) {
int ret = 0;
mzd_t *A, *B, *C, *D, *E;
printf(" mul: m: %4d, l: %4d, n: %4d, k: %2d, cutoff: %4d", m, l, n, k, cutoff);
/* we create two random matrices */
A = mzd_init(m, l);
B = mzd_init(l, n);
mzd_randomize(A);
mzd_randomize(B);
/* C = A*B via Strassen */
C = mzd_mul(NULL, A, B, cutoff);
/* D = A*B via M4RM, temporary buffers are managed internally */
D = mzd_mul_m4rm( NULL, A, B, k);
/* E = A*B via naive cubic multiplication */
E = mzd_mul_naive( NULL, A, B);
mzd_free(A);
mzd_free(B);
if (mzd_equal(C, D) != TRUE) {
printf(" Strassen != M4RM");
ret -=1;
}
if (mzd_equal(D, E) != TRUE) {
printf(" M4RM != Naiv");
ret -= 1;
}
if (mzd_equal(C, E) != TRUE) {
printf(" Strassen != Naiv");
ret -= 1;
}
mzd_free(C);
mzd_free(D);
mzd_free(E);
if(ret==0) {
printf(" ... passed\n");
} else {
printf(" ... FAILED\n");
}
return ret;
}
int addsqr_test_equality(rci_t m, int k, int cutoff) {
int ret = 0;
mzd_t *A, *C, *D, *E, *F;
printf("addsqr: m: %4d, k: %2d, cutoff: %4d", m, k, cutoff);
/* we create two random matrices */
A = mzd_init(m, m);
C = mzd_init(m, m);
mzd_randomize(A);
mzd_randomize(C);
/* D = C + A*B via M4RM, temporary buffers are managed internally */
D = mzd_copy(NULL, C);
D = mzd_addmul_m4rm(D, A, A, k);
/* E = C + A*B via naive cubic multiplication */
E = mzd_mul_m4rm(NULL, A, A, k);
mzd_add(E, E, C);
/* F = C + A*B via naive cubic multiplication */
F = mzd_copy(NULL, C);
F = mzd_addmul(F, A, A, cutoff);
mzd_free(A);
mzd_free(C);
if (mzd_equal(D, E) != TRUE) {
printf(" M4RM != add,mul");
ret -=1;
}
if (mzd_equal(E, F) != TRUE) {
printf(" add,mul = addmul");
ret -=1;
}
if (mzd_equal(F, D) != TRUE) {
printf(" M4RM != addmul");
ret -=1;
}
if (ret==0)
printf(" ... passed\n");
else
printf(" ... FAILED\n");
mzd_free(D);
mzd_free(E);
mzd_free(F);
return ret;
}
int test_pluq_solve_left(rci_t m, rci_t n, int offsetA, int offsetB) {
mzd_t* Abase = mzd_init(2048, 2048);
mzd_t* Bbase = mzd_init(2048, 2048);
mzd_randomize(Abase);
mzd_randomize(Bbase);
mzd_t* A = mzd_init_window(Abase, 0, offsetA, m, m + offsetA);
mzd_t* B = mzd_init_window(Bbase, 0, offsetB, m, n + offsetB);
// copy B
mzd_t* Bcopy = mzd_init(B->nrows, B->ncols);
for (rci_t i = 0; i < B->nrows; ++i)
for (rci_t j = 0; j < B->ncols; ++j)
mzd_write_bit(Bcopy,i,j, mzd_read_bit (B,i,j));
for (rci_t i = 0; i < m; ++i) {
mzd_write_bit(A,i,i, 1);
}
mzd_t *Acopy = mzd_copy(NULL, A);
rci_t r = mzd_echelonize(Acopy,1);
printf("solve_left m: %4d, n: %4d, r: %4d da: %4d db: %4d ", m, n, r, offsetA, offsetB);
mzd_free(Acopy);
Acopy = mzd_copy(NULL, A);
int consistency = mzd_solve_left(A, B, 0, 1);
//copy B
mzd_t *X = mzd_init(B->nrows,B->ncols);
for (rci_t i = 0; i < B->nrows; ++i)
for (rci_t j = 0; j < B->ncols; ++j)
mzd_write_bit(X,i,j, mzd_read_bit (B,i,j));
mzd_t *B1 = mzd_mul(NULL, Acopy, X, 0);
mzd_t *Z = mzd_add(NULL, Bcopy, B1);
int status = 0;
if(consistency == 0) {
status = 1 - mzd_is_zero(Z);
if (status == 0) {
printf("passed\n");
} else {
printf("FAILED\n");
}
} else {
printf("skipped (no solution)\n");
}
mzd_free(Bcopy);
mzd_free(B1);
mzd_free(Z);
mzd_free_window(A);
mzd_free_window(B);
mzd_free(Acopy);
mzd_free(Abase);
mzd_free(Bbase);
mzd_free(X);
return status;
}
int test_kernel_left_pluq(size_t m, size_t n) {
mzd_t* A = mzd_init(m, n);
mzd_randomize(A);
mzd_t *Acopy = mzd_copy(NULL, A);
size_t r = mzd_echelonize_m4ri(A, 0, 0);
printf("kernel_left m: %4zu, n: %4zu, r: %4zu ",m, n, r);
mzd_free(Acopy);
Acopy = mzd_copy(NULL, A);
mzd_t *X = mzd_kernel_left_pluq(A, 0);
if (X == NULL) {
printf("passed\n");
mzd_free(A);
mzd_free(Acopy);
return 0;
}
mzd_t *Z = mzd_mul(NULL, Acopy, X, 0);
int status = 1 - mzd_is_zero(Z);
if (!status)
printf("passed\n");
else
printf("FAILED\n");
mzd_free(A);
mzd_free(Acopy);
mzd_free(X);
mzd_free(Z);
return status;
}
int run(void *_p, unsigned long long *data, int *data_len) {
struct trsm_params *p = (struct trsm_params *)_p;
*data_len = 2;
mzd_t *B = mzd_init(p->m, p->n);
mzd_t *L = mzd_init(p->n, p->n);
mzd_randomize(B);
mzd_randomize(L);
for (rci_t i = 0; i < p->n; ++i){
for (rci_t j = i + 1; j < p->n; ++j)
mzd_write_bit(L,i,j, 0);
mzd_write_bit(L,i,i, 1);
}
data[0] = walltime(0);
data[1] = cpucycles();
mzd_trsm_lower_right(L, B, 2048);
data[0] = walltime(data[0]);
data[1] = cpucycles() - data[1];
mzd_free(B);
mzd_free(L);
return 0;
}
mzd_slice_t *_mzd_slice_addmul_naive(mzd_slice_t *C, const mzd_slice_t *A, const mzd_slice_t *B) {
if (C == NULL)
C = mzd_slice_init(A->finite_field, A->nrows, B->ncols);
const unsigned int e = A->finite_field->degree;
mzd_t *t0 = mzd_init(A->nrows, B->ncols);
for(unsigned int i=0; i<e; i++) {
for(unsigned int j=0; j<e; j++) {
mzd_mul(t0, A->x[i], B->x[j], 0);
_mzd_ptr_add_modred(A->finite_field, t0, C->x, i+j);
}
}
mzd_free(t0);
return C;
}
MatrixF2::~MatrixF2()
{
mzd_free(matrix);
}
/*
* Class: m4rjni_Mzd
* Method: mzd_destroy
* Signature: (J)V
*/
JNIEXPORT void JNICALL Java_m4rjni_Mzd_mzd_1destroy(JNIEnv *env, jobject obj, jlong ptr) {
mzd_t *M = (mzd_t*)ptr;
mzd_free(M);
}
/**
* Implements both apply_p_right and apply_p_right_trans.
*/
void _mzd_apply_p_right_even(mzd_t *A, mzp_t *P, size_t start_row, size_t start_col, int notrans) {
assert(A->offset == 0);
if(A->nrows - start_row == 0)
return;
const size_t length = MIN(P->length,A->ncols);
const size_t width = A->width;
size_t step_size = MIN(A->nrows-start_row, MAX((CPU_L1_CACHE>>3)/A->width,1));
/* our temporary where we store the columns we want to swap around */
mzd_t *B = mzd_init(step_size, A->ncols);
word *Arow;
word *Brow;
/* setup mathematical permutation */
size_t *permutation = (size_t *)m4ri_mm_calloc(sizeof(size_t),A->ncols);
for(size_t i=0; i<A->ncols; i++)
permutation[i] = i;
if (!notrans) {
for(size_t i=start_col; i<length; i++) {
size_t t = permutation[i];
permutation[i] = permutation[P->values[i]];
permutation[P->values[i]] = t;
}
} else {
for(size_t i=start_col; i<length; i++) {
size_t t = permutation[length-i-1];
permutation[length-i-1] = permutation[P->values[length-i-1]];
permutation[P->values[length-i-1]] = t;
}
}
/* we have a bitmask to encode where to write to */
word *write_mask = (word*)m4ri_mm_calloc(sizeof(word), length);
for(size_t i=0; i<A->ncols; i+=RADIX) {
const size_t todo = MIN(RADIX,A->ncols-i);
for(size_t k=0; k<todo; k++) {
if(permutation[i+k] == i+k) {
write_mask[i/RADIX] |= ONE<<(RADIX - k - 1);
}
}
}
for(size_t i=start_row; i<A->nrows; i+=step_size) {
step_size = MIN(step_size, A->nrows-i);
for(size_t k=0; k<step_size; k++) {
Arow = A->rows[i+k];
Brow = B->rows[k];
/*copy row & clear those values which will be overwritten */
for(size_t j=0; j<width; j++) {
Brow[j] = Arow[j];
Arow[j] = Arow[j] & write_mask[j];
}
}
/* here we actually write out the permutation */
mzd_write_col_to_rows_blockd(A, B, permutation, write_mask, i, i+step_size, length);
}
m4ri_mm_free(permutation);
m4ri_mm_free(write_mask);
mzd_free(B);
}
int test_lqup_full_rank (size_t m, size_t n){
printf("pluq: testing full rank m: %5zu, n: %5zu",m,n);
mzd_t* U = mzd_init (m,n);
mzd_t* L = mzd_init (m,m);
mzd_t* U2 = mzd_init (m,n);
mzd_t* L2 = mzd_init (m,m);
mzd_t* A = mzd_init (m,n);
mzd_randomize (U);
mzd_randomize (L);
size_t i,j;
for (i=0; i<m; ++i){
for (j=0; j<i && j<n;++j)
mzd_write_bit(U,i,j, 0);
for (j=i+1; j<m;++j)
mzd_write_bit(L,i,j, 0);
if(i<n)
mzd_write_bit(U,i,i, 1);
mzd_write_bit(L,i,i, 1);
}
mzd_mul(A, L, U, 2048);
mzd_t* Acopy = mzd_copy (NULL,A);
mzp_t* P = mzp_init(m);
mzp_t* Q = mzp_init(n);
mzd_pluq(A, P, Q, 2048);
for (i=0; i<m; ++i){
for (j=0; j<i && j <n;++j)
mzd_write_bit (L2, i, j, mzd_read_bit(A,i,j));
for (j=i+1; j<n;++j)
mzd_write_bit (U2, i, j, mzd_read_bit(A,i,j));
}
for (i=0; i<n && i<m; ++i){
mzd_write_bit(L2,i,i, 1);
mzd_write_bit(U2,i,i, 1);
}
mzd_addmul(Acopy,L2,U2,0);
int status = 0;
for ( i=0; i<m; ++i)
for ( j=0; j<n; ++j){
if (mzd_read_bit (Acopy,i,j)){
status = 1;
}
}
if (status){
printf(" ... FAILED\n");
} else
printf (" ... passed\n");
mzd_free(U);
mzd_free(L);
mzd_free(U2);
mzd_free(L2);
mzd_free(A);
mzd_free(Acopy);
mzp_free(P);
mzp_free(Q);
return status;
}
int run(void *_p, unsigned long long *data, int *data_len) {
struct elim_params *p = (struct elim_params *)_p;
#ifndef HAVE_LIBPAPI
*data_len = 2;
#else
*data_len = MIN(papi_array_len + 1, *data_len);
#endif
int papi_res;
mzd_t *A = mzd_init(p->m, p->n);
if(p->r != 0) {
mzd_t *L, *U;
L = mzd_init(p->m, p->m);
U = mzd_init(p->m, p->n);
mzd_randomize(U);
mzd_randomize(L);
for (rci_t i = 0; i < p->m; ++i) {
for (rci_t j = i + 1; j < p->m; j+=m4ri_radix) {
int const length = MIN(m4ri_radix, p->m - j);
mzd_clear_bits(L, i, j, length);
}
mzd_write_bit(L,i,i, 1);
for (rci_t j = 0; j < i && j < p->n; j+=m4ri_radix) {
int const length = MIN(m4ri_radix, i - j);
mzd_clear_bits(U, i, j, length);
}
if(i < p->r) {
mzd_write_bit(U, i, i, 1);
} else {
for (rci_t j = i; j < p->n; j+=m4ri_radix) {
int const length = MIN(m4ri_radix, p->n - i);
mzd_clear_bits(U, i, j, length);
}
}
}
mzd_mul(A,L,U,0);
mzd_free(L);
mzd_free(U);
} else {
mzd_randomize(A);
}
mzp_t *P = mzp_init(A->nrows);
mzp_t *Q = mzp_init(A->ncols);
#ifndef HAVE_LIBPAPI
data[0] = walltime(0);
data[1] = cpucycles();
#else
int array_len = *data_len - 1;
unsigned long long t0 = PAPI_get_virt_usec();
papi_res = PAPI_start_counters((int*)papi_events, array_len);
if (papi_res)
m4ri_die("");
#endif
if(strcmp(p->algorithm, "m4ri") == 0)
p->r = mzd_echelonize_m4ri(A, 0, 0);
else if(strcmp(p->algorithm, "ple") == 0)
p->r = mzd_ple(A, P, Q, 0);
else if(strcmp(p->algorithm, "mmpf") == 0)
p->r = _mzd_ple_russian(A, P, Q, 0);
else
m4ri_die("unknown algorithm %s",p->algorithm);
#ifndef HAVE_LIBPAPI
data[1] = cpucycles() - data[1];
data[0] = walltime(data[0]);
#else
mzp_free(P);
mzp_free(Q);
PAPI_stop_counters((long long*)&data[1], array_len);
t0 = PAPI_get_virt_usec() - t0;
data[0] = t0;
for (int nv = 0; nv <= array_len; ++nv) {
data[nv] -= loop_calibration[nv];
}
#endif
mzd_free(A);
return 0;
}
int run_nothing(void *_p, unsigned long long *data, int *data_len) {
struct elim_params *p = (struct elim_params *)_p;
mzd_t *A = mzd_init(p->m, p->n);
if(p->r != 0) {
mzd_t *L, *U;
L = mzd_init(p->m, p->m);
U = mzd_init(p->m, p->n);
mzd_randomize(U);
mzd_randomize(L);
for (rci_t i = 0; i < p->m; ++i) {
for (rci_t j = i + 1; j < p->m; j+=m4ri_radix) {
int const length = MIN(m4ri_radix, p->m - j);
mzd_clear_bits(L, i, j, length);
}
mzd_write_bit(L,i,i, 1);
for (rci_t j = 0; j < i && j <p->n; j+=m4ri_radix) {
int const length = MIN(m4ri_radix, i - j);
mzd_clear_bits(U, i, j, length);
}
if(i < p->r) {
mzd_write_bit(U, i, i, 1);
} else {
for (rci_t j = i; j < p->n; j+=m4ri_radix) {
int const length = MIN(m4ri_radix, p->n - j);
mzd_clear_bits(U, i, j, length);
}
}
}
mzd_mul(A,L,U,0);
mzd_free(L);
mzd_free(U);
} else {
mzd_randomize(A);
}
#ifndef HAVE_LIBPAPI
*data_len = 2;
#else
*data_len = MIN(papi_array_len + 1, *data_len);
#endif
int papi_res;
#ifndef HAVE_LIBPAPI
data[0] = walltime(0);
data[1] = cpucycles();
#else
int array_len = *data_len - 1;
unsigned long long t0 = PAPI_get_virt_usec();
papi_res = PAPI_start_counters((int*)papi_events, array_len);
if(papi_res)
m4ri_die("");
#endif
#ifndef HAVE_LIBPAPI
data[1] = cpucycles() - data[1];
data[0] = walltime(data[0]);
#else
PAPI_stop_counters((long long*)&data[1], array_len);
t0 = PAPI_get_virt_usec() - t0;
data[0] = t0;
for (int nv = 0; nv <= array_len; ++nv) {
if (data[nv] < loop_calibration[nv])
loop_calibration[nv] = data[nv];
}
#endif
mzd_free(A);
return (0);
}
void
_qseive(const mp_limb_t n, const mp_limb_t B)
{
nmod_sparse_mat_t M;
mp_limb_t quad, *quads, *xs, x, i = 0, j, piB = n_prime_pi(B);
const mp_limb_t * ps = n_primes_arr_readonly(piB + 2);
const double * pinvs = n_prime_inverses_arr_readonly(piB + 2);
mzd_t *K;
/* init */
quads = (mp_limb_t *)malloc((piB + 1)*sizeof(mp_limb_t *));
xs = (mp_limb_t *)malloc((piB + 1)*sizeof(mp_limb_t *));
K = mzd_init(piB + 1, 1);
nmod_sparse_mat_init(M, piB + 1, piB + 1, 2);
printf("init done\n");
printf("using %ld primes\n", piB);
/* seive */
for (x = n_sqrt(n), i = 0; i <= piB; x++)
{
quad = x*x - n;
if (quad == 0)
continue;
for (j = 0; j < piB; j++)
n_remove2_precomp(&quad, ps[j], pinvs[j]);
if (quad == 1) /* was B-smooth */
{
quads[i] = x*x - n;
quad = x*x - n;
for (j = 0; j < piB; j++)
{
if (n_remove2_precomp(&quad, ps[j], pinvs[j]) % 2)
_nmod_sparse_mat_set_entry(M, j, i, M->row_supports[j], 1);
}
xs[i] = x;
i++;
}
}
printf("data collection done\n");
n_cleanup_primes();
_bw(K, M, 1, 2, 7, 7);
printf("procesing complete\n");
mzd_print(K);
int done = 0;
for (j = 0; !done; j++)
{
fmpz_t a, b, diff, N;
fmpz_init_set_ui(a, 1);
fmpz_init_set_ui(b, 1);
fmpz_init_set_ui(N, n);
fmpz_init(diff);
for (i = 0; i < piB; i++)
{
if (mzd_read_bit(K, i, j))
{
fmpz_mul_ui(a, a, xs[i]);
fmpz_mul_ui(b, b, quads[i]);
}
}
assert(fmpz_is_square(b));
fmpz_sqrt(b, b);
if (fmpz_mod_ui(a, a, n) != fmpz_mod_ui(b, b, n) && fmpz_mod_ui(a, a, n) != n - fmpz_mod_ui(b, b, n))
{
done = 1;
fmpz_print(a);
printf("\n");
fmpz_print(b);
printf("\n");
fmpz_sub(diff, a, b);
fmpz_gcd(a, diff, N);
fmpz_divexact(b, N, a);
fmpz_print(a);
printf("\n");
fmpz_print(b);
}
fmpz_clear(a);
fmpz_clear(b);
fmpz_clear(N);
fmpz_clear(diff);
}
/* cleanup */
free(quads);
free(xs);
mzd_free(K);
nmod_sparse_mat_clear(M);
return;
}
void free_matrix(gf2matrix *m)
{
if(!m)
return;
mzd_free(m);
}
int test_lqup_half_rank(size_t m, size_t n) {
printf("pluq: testing half rank m: %5zd, n: %5zd",m,n);
mzd_t* U = mzd_init(m, n);
mzd_t* L = mzd_init(m, m);
mzd_t* U2 = mzd_init(m, n);
mzd_t* L2 = mzd_init(m, m);
mzd_t* A = mzd_init(m, n);
mzd_randomize (U);
mzd_randomize (L);
size_t i,j;
for (i=0; i<m && i<n; ++i){
mzd_write_bit(U,i,i, 1);
for (j=0; j<i;++j)
mzd_write_bit(U,i,j, 0);
if (i%2)
for (j=i; j<n;++j)
mzd_write_bit(U,i,j, 0);
for (j=i+1; j<m;++j)
mzd_write_bit(L,i,j, 0);
mzd_write_bit(L,i,i, 1);
}
mzd_mul(A, L, U, 0);
mzd_t* Acopy = mzd_copy (NULL,A);
mzp_t* Pt = mzp_init(m);
mzp_t* Q = mzp_init(n);
int r = mzd_pluq(A, Pt, Q, 0);
for (i=0; i<r; ++i){
for (j=0; j<i;++j)
mzd_write_bit (L2, i, j, mzd_read_bit(A,i,j));
for (j=i+1; j<n;++j)
mzd_write_bit (U2, i, j, mzd_read_bit(A,i,j));
}
for (i=r; i<m; i++)
for (j=0; j<r;++j)
mzd_write_bit (L2, i, j, mzd_read_bit(A,i,j));
for (i=0; i<r; ++i){
mzd_write_bit(L2,i,i, 1);
mzd_write_bit(U2,i,i, 1);
}
mzd_apply_p_left(Acopy, Pt);
mzd_apply_p_right_trans(Acopy, Q);
mzd_addmul(Acopy,L2,U2,0);
int status = 0;
for ( i=0; i<m; ++i) {
for ( j=0; j<n; ++j){
if (mzd_read_bit(Acopy,i,j)){
status = 1;
}
}
if(status)
break;
}
if (status)
printf(" ... FAILED\n");
else
printf (" ... passed\n");
mzd_free(U);
mzd_free(L);
mzd_free(U2);
mzd_free(L2);
mzd_free(A);
mzd_free(Acopy);
mzp_free(Pt);
mzp_free(Q);
return status;
}