partially refactor and move power series code to the acb_dirichlet module

This commit is contained in:
Fredrik Johansson 2016-10-21 20:32:46 +02:00
parent 84a49ff8fd
commit c4af23b1c5
5 changed files with 228 additions and 132 deletions

View file

@ -27,6 +27,11 @@
extern "C" {
#endif
void acb_dirichlet_powsum_term(acb_ptr res, arb_t log_prev, ulong * prev,
const acb_t s, ulong k, int integer, int critical_line, slong len, slong prec);
void acb_dirichlet_powsum_sieved(acb_ptr z, const acb_t s, ulong n, slong len, slong prec);
typedef struct
{
acb_struct s;

View file

@ -0,0 +1,119 @@
/*
Copyright (C) 2016 Fredrik Johansson
This file is part of Arb.
Arb is free software: you can redistribute it and/or modify it under
the terms of the GNU Lesser General Public License (LGPL) as published
by the Free Software Foundation; either version 2.1 of the License, or
(at your option) any later version. See <http://www.gnu.org/licenses/>.
*/
#include "acb_dirichlet.h"
#include "acb_poly.h"
#define POWER(_k) (powers + (((_k)-1)/2) * (len))
#define DIVISOR(_k) (divisors[((_k)-1)/2])
void
acb_dirichlet_powsum_sieved(acb_ptr z, const acb_t s, ulong n, slong len, slong prec)
{
slong * divisors;
slong powers_alloc;
slong i, j, k, ibound, power_of_two, horner_point;
ulong kprev;
int critical_line, integer;
acb_ptr powers;
acb_ptr t, u, x;
acb_ptr p1, p2;
arb_t logk, v, w;
if (n <= 1)
{
acb_set_ui(z, n);
_acb_vec_zero(z + 1, len - 1);
return;
}
critical_line = arb_is_exact(acb_realref(s)) &&
(arf_cmp_2exp_si(arb_midref(acb_realref(s)), -1) == 0);
integer = arb_is_zero(acb_imagref(s)) && arb_is_int(acb_realref(s));
divisors = flint_calloc(n / 2 + 1, sizeof(slong));
powers_alloc = (n / 6 + 1) * len;
powers = _acb_vec_init(powers_alloc);
ibound = n_sqrt(n);
for (i = 3; i <= ibound; i += 2)
if (DIVISOR(i) == 0)
for (j = i * i; j <= n; j += 2 * i)
DIVISOR(j) = i;
t = _acb_vec_init(len);
u = _acb_vec_init(len);
x = _acb_vec_init(len);
arb_init(logk);
arb_init(v);
arb_init(w);
power_of_two = 1;
while (power_of_two * 2 <= n)
power_of_two *= 2;
horner_point = n / power_of_two;
_acb_vec_zero(z, len);
kprev = 1;
acb_dirichlet_powsum_term(x, logk, &kprev, s, 2,
integer, critical_line, len, prec);
for (k = 1; k <= n; k += 2)
{
/* t = k^(-s) */
if (DIVISOR(k) == 0)
{
acb_dirichlet_powsum_term(t, logk, &kprev, s, k,
integer, critical_line, len, prec);
}
else
{
p1 = POWER(DIVISOR(k));
p2 = POWER(k / DIVISOR(k));
if (len == 1)
acb_mul(t, p1, p2, prec);
else
_acb_poly_mullow(t, p1, len, p2, len, len, prec);
}
if (k * 3 <= n)
_acb_vec_set(POWER(k), t, len);
_acb_vec_add(u, u, t, len, prec);
while (k == horner_point && power_of_two != 1)
{
_acb_poly_mullow(t, z, len, x, len, len, prec);
_acb_vec_add(z, t, u, len, prec);
power_of_two /= 2;
horner_point = n / power_of_two;
horner_point -= (horner_point % 2 == 0);
}
}
_acb_poly_mullow(t, z, len, x, len, len, prec);
_acb_vec_add(z, t, u, len, prec);
flint_free(divisors);
_acb_vec_clear(powers, powers_alloc);
_acb_vec_clear(t, len);
_acb_vec_clear(u, len);
_acb_vec_clear(x, len);
arb_clear(logk);
arb_clear(v);
arb_clear(w);
}

View file

@ -0,0 +1,73 @@
/*
Copyright (C) 2016 Fredrik Johansson
This file is part of Arb.
Arb is free software: you can redistribute it and/or modify it under
the terms of the GNU Lesser General Public License (LGPL) as published
by the Free Software Foundation; either version 2.1 of the License, or
(at your option) any later version. See <http://www.gnu.org/licenses/>.
*/
#include "acb_dirichlet.h"
void
acb_dirichlet_powsum_term(acb_ptr res, arb_t log_prev, ulong * prev,
const acb_t s, ulong k, int integer, int critical_line, slong len, slong prec)
{
slong i;
if (integer)
{
arb_neg(acb_realref(res), acb_realref(s));
arb_set_ui(acb_imagref(res), k);
arb_pow(acb_realref(res), acb_imagref(res), acb_realref(res), prec);
arb_zero(acb_imagref(res));
if (len != 1)
{
arb_log_ui_from_prev(log_prev, k, log_prev, *prev, prec);
*prev = k;
}
}
else
{
arb_t w;
arb_init(w);
arb_log_ui_from_prev(log_prev, k, log_prev, *prev, prec);
*prev = k;
arb_mul(w, log_prev, acb_imagref(s), prec);
arb_sin_cos(acb_imagref(res), acb_realref(res), w, prec);
arb_neg(acb_imagref(res), acb_imagref(res));
if (critical_line)
{
arb_rsqrt_ui(w, k, prec);
acb_mul_arb(res, res, w, prec);
}
else
{
arb_mul(w, acb_realref(s), log_prev, prec);
arb_neg(w, w);
arb_exp(w, w, prec);
acb_mul_arb(res, res, w, prec);
}
arb_clear(w);
}
if (len > 1)
{
arb_neg(log_prev, log_prev);
for (i = 1; i < len; i++)
{
acb_mul_arb(res + i, res + i - 1, log_prev, prec);
acb_div_ui(res + i, res + i, i, prec);
}
arb_neg(log_prev, log_prev);
}
}

View file

@ -10,141 +10,11 @@
*/
#include "acb_poly.h"
#define POWER(_k) (powers + (((_k)-1)/2) * (len))
#define DIVISOR(_k) (divisors[((_k)-1)/2])
#define COMPUTE_POWER(t, k, kprev) \
do { \
if (integer) \
{ \
arb_neg(w, acb_realref(s)); \
arb_set_ui(v, k); \
arb_pow(acb_realref(t), v, w, prec); \
arb_zero(acb_imagref(t)); \
if (len != 1) \
{ \
arb_log_ui_from_prev(logk, k, logk, kprev, prec); \
kprev = k; \
arb_neg(logk, logk); \
} \
} \
else \
{ \
arb_log_ui_from_prev(logk, k, logk, kprev, prec); \
kprev = k; \
arb_neg(logk, logk); \
arb_mul(w, logk, acb_imagref(s), prec); \
arb_sin_cos(acb_imagref(t), acb_realref(t), w, prec); \
if (critical_line) \
{ \
arb_rsqrt_ui(w, k, prec); \
acb_mul_arb(t, t, w, prec); \
} \
else \
{ \
arb_mul(w, acb_realref(s), logk, prec); \
arb_exp(w, w, prec); \
acb_mul_arb(t, t, w, prec); \
} \
} \
for (i = 1; i < len; i++) \
{ \
acb_mul_arb(t + i, t + i - 1, logk, prec); \
acb_div_ui(t + i, t + i, i, prec); \
} \
arb_neg(logk, logk); \
} while (0); \
#include "acb_dirichlet.h"
void
_acb_poly_powsum_one_series_sieved(acb_ptr z, const acb_t s, slong n, slong len, slong prec)
{
slong * divisors;
slong powers_alloc;
slong i, j, k, ibound, kprev, power_of_two, horner_point;
int critical_line, integer;
acb_ptr powers;
acb_ptr t, u, x;
acb_ptr p1, p2;
arb_t logk, v, w;
critical_line = arb_is_exact(acb_realref(s)) &&
(arf_cmp_2exp_si(arb_midref(acb_realref(s)), -1) == 0);
integer = arb_is_zero(acb_imagref(s)) && arb_is_int(acb_realref(s));
divisors = flint_calloc(n / 2 + 1, sizeof(slong));
powers_alloc = (n / 6 + 1) * len;
powers = _acb_vec_init(powers_alloc);
ibound = n_sqrt(n);
for (i = 3; i <= ibound; i += 2)
if (DIVISOR(i) == 0)
for (j = i * i; j <= n; j += 2 * i)
DIVISOR(j) = i;
t = _acb_vec_init(len);
u = _acb_vec_init(len);
x = _acb_vec_init(len);
arb_init(logk);
arb_init(v);
arb_init(w);
power_of_two = 1;
while (power_of_two * 2 <= n)
power_of_two *= 2;
horner_point = n / power_of_two;
_acb_vec_zero(z, len);
kprev = 0;
COMPUTE_POWER(x, 2, kprev);
for (k = 1; k <= n; k += 2)
{
/* t = k^(-s) */
if (DIVISOR(k) == 0)
{
COMPUTE_POWER(t, k, kprev);
}
else
{
p1 = POWER(DIVISOR(k));
p2 = POWER(k / DIVISOR(k));
if (len == 1)
acb_mul(t, p1, p2, prec);
else
_acb_poly_mullow(t, p1, len, p2, len, len, prec);
}
if (k * 3 <= n)
_acb_vec_set(POWER(k), t, len);
_acb_vec_add(u, u, t, len, prec);
while (k == horner_point && power_of_two != 1)
{
_acb_poly_mullow(t, z, len, x, len, len, prec);
_acb_vec_add(z, t, u, len, prec);
power_of_two /= 2;
horner_point = n / power_of_two;
horner_point -= (horner_point % 2 == 0);
}
}
_acb_poly_mullow(t, z, len, x, len, len, prec);
_acb_vec_add(z, t, u, len, prec);
flint_free(divisors);
_acb_vec_clear(powers, powers_alloc);
_acb_vec_clear(t, len);
_acb_vec_clear(u, len);
_acb_vec_clear(x, len);
arb_clear(logk);
arb_clear(v);
arb_clear(w);
acb_dirichlet_powsum_sieved(z, s, n, len, prec);
}

View file

@ -22,6 +22,35 @@ The code in other modules for computing the Riemann zeta function,
Hurwitz zeta function and polylogarithm will possibly be migrated to this
module in the future.
Truncated L-series and power sums
-------------------------------------------------------------------------------
.. function:: void acb_dirichlet_powsum_term(acb_ptr res, arb_t log_prev, ulong * prev, const acb_t s, ulong k, int integer, int critical_line, slong len, slong prec)
Sets *res* to `k^{-(s+x)}` as a power series in *x* truncated to length *len*.
The flags *integer* and *critical_line* respectively specify optimizing
for *s* being an integer or having real part 1/2.
On input *log_prev* should contain the natural logarithm of the integer
at *prev*. If *prev* is close to *k*, this can be used to speed up
computations. If `\log(k)` is computed internally by this function, then
*log_prev* is overwritten by this value, and the integer at *prev* is
overwritten by *k*, allowing *log_prev* to be recycled for the next
term when evaluating a power sum.
.. function:: void acb_dirichlet_powsum_sieved(acb_ptr res, const acb_t s, ulong n, slong len, slong prec)
Sets *res* to `\sum_{k=1}^n k^{-(s+x)}`
as a power series in *x* truncated to length *len*.
This function stores a table of powers that have already been calculated,
computing `(ij)^r` as `i^r j^r` whenever `k = ij` is
composite. As a further optimization, it groups all even `k` and
evaluates the sum as a polynomial in `2^{-(s+x)}`.
This scheme requires about `n / \log n` powers, `n / 2` multiplications,
and temporary storage of `n / 6` power series. Due to the extra
power series multiplications, it is only faster than the naive
algorithm when *len* is small.
Hurwitz zeta function
-------------------------------------------------------------------------------