void ifft_radix2(mp_limb_t ** ii, mp_size_t n,
mp_bitcnt_t w, mp_limb_t ** t1, mp_limb_t ** t2)
{
mp_size_t i;
mp_size_t limbs = (w*n)/FLINT_BITS;
if (n == 1)
{
ifft_butterfly(*t1, *t2, ii[0], ii[1], 0, limbs, w);
SWAP_PTRS(ii[0], *t1);
SWAP_PTRS(ii[1], *t2);
return;
}
ifft_radix2(ii, n/2, 2*w, t1, t2);
ifft_radix2(ii+n, n/2, 2*w, t1, t2);
for (i = 0; i < n; i++)
{
ifft_butterfly(*t1, *t2, ii[i], ii[n+i], i, limbs, w);
SWAP_PTRS(ii[i], *t1);
SWAP_PTRS(ii[n+i], *t2);
}
}
void ifft_negacyclic(mp_limb_t ** ii, mp_size_t n, mp_bitcnt_t w,
mp_limb_t ** t1, mp_limb_t ** t2, mp_limb_t ** temp)
{
mp_size_t i;
mp_size_t limbs = (w*n)/FLINT_BITS;
ifft_radix2(ii, n/2, 2*w, t1, t2);
ifft_radix2(ii+n, n/2, 2*w, t1, t2);
if (w & 1)
{
for (i = 0; i < n; i++)
{
ifft_butterfly(*t1, *t2, ii[i], ii[n+i], i, limbs, w);
SWAP_PTRS(ii[i], *t1);
SWAP_PTRS(ii[n+i], *t2);
fft_adjust(*t1, ii[i], n - i/2, limbs, w);
mpn_neg_n(*t1, *t1, limbs + 1);
SWAP_PTRS(ii[i], *t1);
fft_adjust(*t2, ii[n+i], n - (n+i)/2, limbs, w);
mpn_neg_n(*t2, *t2, limbs + 1);
SWAP_PTRS(ii[n+i], *t2);
i++;
ifft_butterfly(*t1, *t2, ii[i], ii[n+i], i, limbs, w);
SWAP_PTRS(ii[i], *t1);
SWAP_PTRS(ii[n+i], *t2);
fft_adjust_sqrt2(*t1, ii[i], 2*n-i, limbs, w, *temp);
mpn_neg_n(*t1, *t1, limbs + 1);
SWAP_PTRS(ii[i], *t1);
fft_adjust_sqrt2(*t2, ii[n+i], n-i, limbs, w, *temp);
mpn_neg_n(*t2, *t2, limbs + 1);
SWAP_PTRS(ii[n+i], *t2);
}
} else
{
for (i = 0; i < n; i++)
{
ifft_butterfly(*t1, *t2, ii[i], ii[n+i], i, limbs, w);
SWAP_PTRS(ii[i], *t1);
SWAP_PTRS(ii[n+i], *t2);
fft_adjust(*t1, ii[i], 2*n-i, limbs, w/2);
mpn_neg_n(*t1, *t1, limbs + 1);
SWAP_PTRS(ii[i], *t1);
fft_adjust(*t2, ii[n+i], n-i, limbs, w/2);
mpn_neg_n(*t2, *t2, limbs + 1);
SWAP_PTRS(ii[n+i], *t2);
}
}
}
void fft_mfa_truncate_sqrt2_inner(mp_limb_t ** ii, mp_limb_t ** jj, mp_size_t n,
mp_bitcnt_t w, mp_limb_t ** t1, mp_limb_t ** t2,
mp_limb_t ** temp, mp_size_t n1, mp_size_t trunc, mp_limb_t * tt)
{
mp_size_t i, j, s;
mp_size_t n2 = (2*n)/n1;
mp_size_t trunc2 = (trunc - 2*n)/n1;
mp_size_t limbs = (n*w)/FLINT_BITS;
mp_bitcnt_t depth = 0;
mp_bitcnt_t depth2 = 0;
while ((UWORD(1)<<depth) < n2) depth++;
while ((UWORD(1)<<depth2) < n1) depth2++;
ii += 2*n;
jj += 2*n;
/* convolutions on relevant rows */
for (s = 0; s < trunc2; s++)
{
i = n_revbin(s, depth);
fft_radix2(ii + i*n1, n1/2, w*n2, t1, t2);
if (ii != jj) fft_radix2(jj + i*n1, n1/2, w*n2, t1, t2);
for (j = 0; j < n1; j++)
{
mp_size_t t = i*n1 + j;
mpn_normmod_2expp1(ii[t], limbs);
if (ii != jj) mpn_normmod_2expp1(jj[t], limbs);
fft_mulmod_2expp1(ii[t], ii[t], jj[t], n, w, tt);
}
ifft_radix2(ii + i*n1, n1/2, w*n2, t1, t2);
}
ii -= 2*n;
jj -= 2*n;
/* convolutions on rows */
for (i = 0; i < n2; i++)
{
fft_radix2(ii + i*n1, n1/2, w*n2, t1, t2);
if (ii != jj) fft_radix2(jj + i*n1, n1/2, w*n2, t1, t2);
for (j = 0; j < n1; j++)
{
mp_size_t t = i*n1 + j;
mpn_normmod_2expp1(ii[t], limbs);
if (ii != jj) mpn_normmod_2expp1(jj[t], limbs);
fft_mulmod_2expp1(ii[t], ii[t], jj[t], n, w, tt);
}
ifft_radix2(ii + i*n1, n1/2, w*n2, t1, t2);
}
}
int
main(void)
{
mp_bitcnt_t depth, w;
flint_rand_t state;
printf("fft/ifft_radix2....");
fflush(stdout);
flint_randinit(state);
_flint_rand_init_gmp(state);
for (depth = 6; depth <= 12; depth++)
{
for (w = 1; w <= 5; w++)
{
mp_size_t n = (1UL<<depth);
mp_size_t limbs = (n*w)/GMP_LIMB_BITS;
mp_size_t size = limbs + 1;
mp_size_t i;
mp_limb_t * ptr;
mp_limb_t ** ii, ** jj, *t1, *t2;
ii = flint_malloc((2*(n + n*size) + 2*size)*sizeof(mp_limb_t));
for (i = 0, ptr = (mp_limb_t *) ii + 2*n; i < 2*n; i++, ptr += size)
{
ii[i] = ptr;
random_fermat(ii[i], state, limbs);
}
t1 = ptr;
t2 = t1 + size;
for (i = 0; i < 2*n; i++)
mpn_normmod_2expp1(ii[i], limbs);
jj = flint_malloc(2*(n + n*size)*sizeof(mp_limb_t));
for (i = 0, ptr = (mp_limb_t *) jj + 2*n; i < 2*n; i++, ptr += size)
{
jj[i] = ptr;
mpn_copyi(jj[i], ii[i], size);
}
fft_radix2(ii, n, w, &t1, &t2);
ifft_radix2(ii, n, w, &t1, &t2);
for (i = 0; i < 2*n; i++)
{
mpn_div_2expmod_2expp1(ii[i], ii[i], limbs, depth + 1);
mpn_normmod_2expp1(ii[i], limbs);
}
for (i = 0; i < 2*n; i++)
{
if (mpn_cmp(ii[i], jj[i], size) != 0)
{
printf("FAIL:\n");
printf("Error in entry %ld\n", i);
abort();
}
}
flint_free(ii);
flint_free(jj);
}
}
flint_randclear(state);
printf("PASS\n");
return 0;
}