// truncated multiplication for fmpz
void fmpz_mul_trunc(fmpz_t res, fmpz_t a, fmpz_t b, unsigned long trunc)
{
unsigned long sizea = FLINT_MIN(fmpz_size(a), trunc);
unsigned long sizeb = FLINT_MIN(fmpz_size(b), trunc);
while ((!a[sizea]) && (sizea)) sizea--;
while ((!b[sizeb]) && (sizeb)) sizeb--;
if ((sizea == 0) || (sizeb == 0)) {
res[0] = 0;
return;
}
if (trunc >= sizea + sizeb) {
mp_limb_t mslimb;
if (sizea >= sizeb) mslimb = F_mpn_mul(res+1, a+1, sizea, b+1, sizeb);
else mslimb = F_mpn_mul(res+1, b+1, sizeb, a+1, sizea);
res[0] = sizea + sizeb - (mslimb == 0);
} else {
mp_limb_t mslimb;
fmpz_t temp = flint_stack_alloc(sizea + sizeb + 1);
if (sizea >= sizeb) mslimb = F_mpn_mul_trunc(temp+1, a+1, sizea, b+1, sizeb, trunc);
else mslimb = F_mpn_mul_trunc(temp+1, b+1, sizeb, a+1, sizea, trunc);
temp[0] = trunc;
if (UNLIKELY(!mslimb))
__fmpz_normalise(temp); // normalise if most significant limb == 0
fmpz_set(res, temp);
flint_stack_release();
}
if ((long) (a[0] ^ b[0]) < 0L) res[0] = -res[0];
}
void fmpz_addmul(fmpz_t res, const fmpz_t a, const fmpz_t b)
{
long a0 = a[0];
long b0 = b[0];
unsigned long sizea = FLINT_ABS(a0);
unsigned long sizeb = FLINT_ABS(b0);
while ((!a[sizea]) && (sizea)) sizea--;
while ((!b[sizeb]) && (sizeb)) sizeb--;
fmpz_t temp;
mp_limb_t mslimb;
if (sizea && sizeb)
{
if (sizea + sizeb < 100)
{
temp = (fmpz_t) flint_stack_alloc_small(sizea + sizeb + 1);
if (sizea >= sizeb) mslimb = mpn_mul(temp+1, a+1, sizea, b+1, sizeb);
else mslimb = mpn_mul(temp+1, b+1, sizeb, a+1, sizea);
temp[0] = sizea + sizeb - (mslimb == 0);
if ((long) (a[0] ^ b[0]) < 0) temp[0] = -temp[0];
fmpz_add(res, res, temp);
flint_stack_release_small();
} else
{
temp = (fmpz_t) flint_stack_alloc(sizea + sizeb + 1);
if (sizea >= sizeb) mslimb = F_mpn_mul(temp+1, a+1, sizea, b+1, sizeb);
else mslimb = F_mpn_mul(temp+1, b+1, sizeb, a+1, sizea);
temp[0] = sizea + sizeb - (mslimb == 0);
if ((long) (a[0] ^ b[0]) < 0) temp[0] = -temp[0];
fmpz_add(res, res, temp);
flint_stack_release();
}
}
}
void __fmpz_mul(fmpz_t res, const fmpz_t a, const fmpz_t b)
{
long a0 = a[0];
long b0 = b[0];
unsigned long sizea = FLINT_ABS(a0);
unsigned long sizeb = FLINT_ABS(b0);
while ((!a[sizea]) && (sizea)) sizea--;
while ((!b[sizeb]) && (sizeb)) sizeb--;
mp_limb_t mslimb;
fmpz_t temp;
if ((sizea == 0) || (sizeb == 0))
{
res[0] = 0;
}
else if (sizea + sizeb < 100)
{
if (sizea >= sizeb) mslimb = mpn_mul(res+1, a+1, sizea, b+1, sizeb);
else mslimb = mpn_mul(res+1, b+1, sizeb, a+1, sizea);
res[0] = sizea + sizeb - (mslimb == 0);
if ((long) (a[0] ^ b[0]) < 0) res[0] = -res[0];
} else
{
if (sizea >= sizeb) mslimb = F_mpn_mul(res+1, a+1, sizea, b+1, sizeb);
else mslimb = F_mpn_mul(res+1, b+1, sizeb, a+1, sizea);
res[0] = sizea+sizeb - (mslimb == 0);
if ((long) (a[0] ^ b[0]) < 0) res[0] = -res[0];
}
}
void fmpz_mul(fmpz_t res, const fmpz_t a, const fmpz_t b)
{
long a0 = a[0];
long b0 = b[0];
unsigned long sizea = FLINT_ABS(a0);
unsigned long sizeb = FLINT_ABS(b0);
while ((!a[sizea]) && (sizea)) sizea--;
while ((!b[sizeb]) && (sizeb)) sizeb--;
mp_limb_t mslimb;
fmpz_t temp;
if ((sizea == 0) || (sizeb == 0))
{
res[0] = 0;
} else if (sizea + sizeb < 100)
{
temp = (fmpz_t) flint_stack_alloc_small(sizea + sizeb + 1);
if (sizea > sizeb) mslimb = mpn_mul(temp+1, a+1, sizea, b+1, sizeb);
else if (sizea == sizeb)
{
mpn_mul_n(temp+1, a+1, b+1, sizeb);
mslimb = temp[2*sizeb];
}
else mslimb = mpn_mul(temp+1, b+1, sizeb, a+1, sizea);
temp[0] = sizea + sizeb - (mslimb == 0);
F_mpn_copy(res, temp, temp[0]+1);
if ((long) (a0 ^ b0) < 0) res[0] = -res[0];
flint_stack_release_small();
} else if (sizea + sizeb < 2*FLINT_FFT_LIMBS_CROSSOVER)
{
temp = (fmpz_t) flint_stack_alloc(sizea + sizeb + 1);
if (sizea > sizeb) mslimb = mpn_mul(temp+1, a+1, sizea, b+1, sizeb);
else if (sizea == sizeb)
{
mpn_mul_n(temp+1, a+1, b+1, sizeb);
mslimb = temp[2*sizeb];
}
else mslimb = mpn_mul(temp+1, b+1, sizeb, a+1, sizea);
temp[0] = sizea + sizeb - (mslimb == 0);
F_mpn_copy(res, temp, temp[0]+1);
if ((long) (a0 ^ b0) < 0) res[0] = -res[0];
flint_stack_release();
} else
{
if (sizea >= sizeb) mslimb = F_mpn_mul(res+1, a+1, sizea, b+1, sizeb);
else mslimb = F_mpn_mul(res+1, b+1, sizeb, a+1, sizea);
res[0] = sizea+sizeb - (mslimb == 0);
if ((long) (a0 ^ b0) < 0) res[0] = -res[0];
}
}
int test_F_mpn_mul()
{
mp_limb_t * int1, * int2, * product, * product2;
mp_limb_t msl, msl2;
int result = 1;
unsigned long count;
for (count = 0; (count < 30) && (result == 1); count++)
{
unsigned long limbs2 = randint(2*FLINT_FFT_LIMBS_CROSSOVER)+1;
unsigned long limbs1 = limbs2 + randint(1000);
int1 = (mp_limb_t *) malloc(sizeof(mp_limb_t)*limbs1);
mpn_random2(int1, limbs1);
unsigned long count2;
for (count2 = 0; (count2 < 30) && (result == 1); count2++)
{
#if DEBUG
printf("%ld, %ld\n",limbs1, limbs2);
#endif
int2 = (mp_limb_t *) malloc(sizeof(mp_limb_t)*limbs2);
product = (mp_limb_t *) malloc(sizeof(mp_limb_t)*(limbs1+limbs2));
product2 = (mp_limb_t *) malloc(sizeof(mp_limb_t)*(limbs1+limbs2));
F_mpn_clear(int2, limbs2);
mpn_random2(int2, randint(limbs2-1)+1);
msl = F_mpn_mul(product, int1, limbs1, int2, limbs2);
msl2 = mpn_mul(product2, int1, limbs1, int2, limbs2);
unsigned long j;
for (j = 0; j < limbs1+limbs2 - (msl == 0); j++)
{
if (product[j] != product2[j]) result = 0;
}
result &= (msl == msl2);
free(product2);
free(product);
free(int2);
}
free(int1);
}
return result;
}
