36#define CBRT_DIVIDE_MIN (0x6000000000000000ull)
44 dst[2] =
t[1] + (dst[1] <
t[0] ? 1 : 0);
110 }
else if (ns == 2) {
114#define Ahr (dst + lo)
115#define rk (numa + 3 * lo)
118#define Alr (tp + 2 * hi)
119#define Alr2 (tp + 2 * hi + lo)
120#define scratch (tp + 2 * hi + 3 * lo)
146 if (calr == 0)
return;
170 (
R + 2 * ns)[0] += cy;
171 r -= (
R + 2 * ns)[0] < cy;
186 }
else if (calr == 0) {
211 while (i > hi + 3 *
lo &&
R[i] == 0) --i;
212 if (i > hi + 3 *
lo) {
226 if (rh > th || (rh == th && (rm > tm || (rm == tm && rl >= tl)))) {
231 if (rh < qh2 || (rh == qh2 && (rm < qm || (rm == qm && rl < ql)))) {
277 (
R + 2 * ns)[0] += cy;
278 r -= (
R + 2 * ns)[0] < cy;
304 if (r == 1)
return -1;
305 if (r == 2 && !top2)
return -1;
306 mp_size_t L = (r == 0) ? (bl + 63) / 64 : (bl + 64) / 64;
308 return (r == 0 ? 64 * L : 63 + 64 * (L - 1)) - bl;
324 mp_limb_t hlo = na >= 2 ? numa[na - 2] : 0;
326 if (
k > 1 && (
k & (
k - 1)) == 0)
continue;
344 q2 = t1[1] + (q1 < t0[1]);
349 mp_limb_t qtop = wq == 2 ? q2 : (wq == 1 ? q1 : q0);
352 int top2 = hb >= 2 ? qtop >= 3ULL << (hb - 2)
353 : (qtop & 1) && ((wq == 2 ? q1 : q0) >> (
LIMB_BITS - 1));
367 if (cy) xk[na] = cy, ++nak;
374 top2 = xhb >= 2 ? xtop >= 3ULL << (xhb - 2)
375 : (nak >= 2 && (xk[nak - 2] >> (
LIMB_BITS - 1)));
470 numa2[4] = na > 1 ? numa[-2] : 0;
498 mp_size_t alloc_size = 2 * (ns + namax + 4) + 6 * nrmax + 16;
520 mp_ptr dp = tb, t2 = tb +
r0, w1 = tb +
r0 +
r1,
xp = w1, dip = tb;
536 lmmp_mul_(w1, t2, nsqr, numa - naz, naz);
537 if (nsqr + naz > mn) {
538 if (
lmmp_add_(w1, w1, mn, w1 + mn, nsqr + naz - mn))
558 lmmp_mul_(w1, t2, nsqr, numa - naz, naz);
559 lmmp_mul_(dp, w1, nres, dstis - nr, nr + 1);
584 lmmp_mul_(dip,
xp, 2 * nr + 2, dstis - nr, nr + 1);
588 lmmp_copy(dstis - na, dip + 4 * nr - na, na - nr);
597 lmmp_not_(dstis - na, dip + 4 * nr - na, na - nr);
602 }
while (sizp != sizes);
710 mp_size_t e0 = 2 * ns + 2 * (na / 3) - nf;
725 mp_limb_t xb[6] = {0, 0, 0, 0, 0, 0}, r[2] = {0, 0};
726 for (
mp_size_t i = 0; i < na; ++i) xb[3 * nf + i] = numa[i];
730 }
else if (nl == 3) {
737 if (ns > 1) dsts[1] = r[1];
819 int remnorm = dstr != NULL &&
k == 1 &&
t == 0;
823 if (dstr && !remnorm) {
849 e0 = elo | (ehi <<
t);
855 lmmp_zero(N + 2 * ns2 + 1, L - (2 * ns2 + 1));
871 cy =
lmmp_add_1_(N + 2 * ns2 + 2, N + 2 * ns2 + 2, L - 2 * ns2 - 2, cy);
874 mp_limb_t eb[2] = {e0, e1}, e2b[4], e3b[6];
882 while (ne2 > 1 && e2b[ne2 - 1] == 0) --ne2;
886 bo =
lmmp_sub_1_(N + ns2 + ne2, N + ns2 + ne2, L - ns2 - ne2, bo);
#define CBRT_DIVIDE_MIN
Copyright (C) 2026 HJimmyK (Jericho Knox)
static mp_size_t lmmp_cbrt_cls_shift_(mp_size_t bl, int top2)
static mp_size_t lmmp_cbrt_snorm_(mp_srcptr numa, mp_size_t na, mp_ptr a2, mp_size_t *na2p)
static void lmmp_cbrt_shapply_(mp_srcptr src, mp_size_t n, mp_size_t s, mp_ptr a2)
static void lmmp_cube_6_(mp_ptr restrict t, mp_srcptr restrict r)
计算 [t,6]=[r,2]^3
static mp_size_t lmmp_cbrt_knorm_(mp_srcptr numa, mp_size_t na, mp_ptr a2, mp_size_t *na2p, mp_limb_t *kp)
static void lmmp_cube_3_(mp_ptr restrict dst, mp_limb_t a)
#define lmmp_copy(dst, src, n)
#define lmmp_zero(dst, n)
#define lmmp_debug_assert(x)
void lmmp_fill_n(mp_ptr dst, mp_size_t len, mp_limb_t val)
填充从dst开始的长度为len的数组,值为val
const mp_limb_t * mp_srcptr
#define LMMP_MAX(h, i)
返回两个数中的较大值
#define LMMP_MIN(l, o)
返回两个数中的较小值
#define lmmp_param_assert(x)
mp_limb_t lmmp_div_s_(mp_ptr dstq, mp_ptr numa, mp_size_t na, mp_srcptr numb, mp_size_t nb)
除法运算
mp_limb_t lmmp_div_1_(mp_ptr dstq, mp_srcptr numa, mp_size_t na, mp_limb_t x)
单精度数除法
void lmmp_mul_mersenne_(mp_ptr dst, mp_size_t rn, mp_srcptr numa, mp_size_t na, mp_srcptr numb, mp_size_t nb)
梅森数模乘法 [dst,rn] = [numa,na]*[numb,nb] mod B^rn-1
static mp_limb_t lmmp_add_(mp_ptr dst, mp_srcptr numa, mp_size_t na, mp_srcptr numb, mp_size_t nb)
加法静态内联函数 [dst,na]=[numa,na]+[numb,nb]
#define lmmp_dec(p)
减1宏(预期无借位)
static mp_limb_t lmmp_add_1_(mp_ptr dst, mp_srcptr numa, mp_size_t na, mp_limb_t x)
加单精度数静态内联函数 [dst,na]=[numa,na]+x
void lmmp_sqr_mersenne_(mp_ptr dst, mp_size_t rn, mp_srcptr numa, mp_size_t na)
梅森数模平方 [dst,rn] = [numa,na]^2 mod B^rn-1
#define lmmp_inc(p)
加1宏(预期无进位)
mp_limb_t lmmp_shr_(mp_ptr dst, mp_srcptr numa, mp_size_t na, mp_size_t shr)
右移操作 [dst,na] = [numa,na]>>shr,dst的高shr位填充0
void lmmp_mul_basecase_(mp_ptr dst, mp_srcptr numa, mp_size_t na, mp_srcptr numb, mp_size_t nb)
基础乘法运算 [dst,na+nb] = [numa,na] * [numb,nb]
void lmmp_mul_(mp_ptr dst, mp_srcptr numa, mp_size_t na, mp_srcptr numb, mp_size_t nb)
不等长乘法操作 [dst,na+nb] = [numa,na] * [numb,nb]
int lmmp_zero_q_(mp_srcptr p, mp_size_t n)
判零函数
mp_limb_t lmmp_shl_(mp_ptr dst, mp_srcptr numa, mp_size_t na, mp_size_t shl)
左移操作 [dst,na] = [numa,na]<<shl,dst的低shl位填充0
static mp_limb_t lmmp_sub_(mp_ptr dst, mp_srcptr numa, mp_size_t na, mp_srcptr numb, mp_size_t nb)
减法静态内联函数 [dst,na]=[numa,na]-[numb,nb]
mp_limb_t lmmp_addmul_1_(mp_ptr numa, mp_srcptr numb, mp_size_t n, mp_limb_t b)
乘以单limb并累加操作 [numa,n] += [numb,n] * b
mp_limb_t lmmp_mul_1_(mp_ptr dst, mp_srcptr numa, mp_size_t na, mp_limb_t x)
乘以单limb操作 [dst,na] = [numa,na] * x
#define lmmp_dec_1(p, dec)
减指定值宏(预期无借位)
static mp_limb_t lmmp_sub_1_(mp_ptr dst, mp_srcptr numa, mp_size_t na, mp_limb_t x)
减单精度数静态内联函数 [dst,na]=[numa,na]-x
void lmmp_not_(mp_ptr dst, mp_srcptr numa, mp_size_t na)
按位取反操作 [dst,na] = ~[numa,na] (对每个limb执行按位非操作)
mp_limb_t lmmp_submul_1_(mp_ptr numa, mp_srcptr numb, mp_size_t n, mp_limb_t b)
乘以单limb并累减操作 [numa,n] -= [numb,n] * b
mp_limb_t lmmp_sub_n_(mp_ptr dst, mp_srcptr numa, mp_srcptr numb, mp_size_t n)
无借位的n位减法 [dst,n] = [numa,n] - [numb,n]
#define lmmp_inc_1(p, inc)
加指定值宏(预期无进位)
mp_size_t lmmp_mersenne_next_size_(mp_size_t n)
计算满足 >=n 的最小分治梅森乘法可行尺寸
void lmmp_sqr_hard_2_(mp_ptr dst, mp_srcptr numa)
#define CBRT_INVNEWTON_K_THRESHOLD
#define CBRT_INVNEWTON_NF_MIN
#define CBRT_NEWTON_MODM_THRESHOLD
mp_limb_t lmmp_cbrt_3_(mp_limb_t a0, mp_limb_t a1, mp_limb_t a2)
计算算数立方根 floor(cbrt(a0+a1*B+a2*B^2))
void lmmp_cbrt_6_fast_(mp_ptr dst, mp_ptr numa, mp_ptr tp)
计算算术立方根 [dst,2] = floor(cbrt([numa,6]))
void lmmp_cbrt_newton_(mp_ptr dsts, mp_srcptr numa, mp_size_t na, mp_size_t nf)
计算近似立方根 [dsts,na/3+nf+1]=[floor|round](cbrt([numa,na]*B^(3*nf))), 输入须已归一化:na ≡ 0 (mod 3) 且顶 limb >= B...
void lmmp_cbrt_(mp_ptr dsts, mp_ptr dstr, mp_srcptr numa, mp_size_t na, mp_size_t nf)
计算 [numa,na] * B^(3*nf) 的立方根和余数
ulong lmmp_cbrt_ulong_(ulong n)
计算算数立方根 floor(cbrt(n))
void lmmp_cbrt_6_(mp_ptr dst, mp_srcptr numa, mp_size_t na)
计算算术立方根 floor(cbrt([numa,na]))
void lmmp_invcbrt_newton_(mp_ptr dstis, mp_size_t ns, mp_srcptr numa, mp_size_t na)
计算逆立方根的定点下近似:记 I=floor(cbrt(B^(3*ns+na)/[numa,na])), 则 [dstis,ns+1] 满足 I-1 <= [dstis,...
void lmmp_cbrt_divide_(mp_ptr dst, mp_ptr numa, mp_size_t ns, mp_ptr tp, int calr)
计算算术立方根 floor(cbrt([numa,3*ns]))
#define restrict
Copyright (C) 2026 HJimmyK (Jericho Knox)
#define TALLOC_TYPE(n, type)