Changes commited to PARI GIT during the last two weeks

74bdd0ac5b | Karim Belabas | Mon, 15 Sep 2025 21:29:19 +0200

find_div: remove now unused parameter N

2	2	src/basemath/ellanal.c

b4cc9c5192 | Karim Belabas | Mon, 15 Sep 2025 21:23:04 +0200

listheegner: kill warning

2	1	src/basemath/ellanal.c

174c393803 | Karim Belabas | Sun, 14 Sep 2025 21:23:41 +0200

find_div: use optimal algorithm to list (pairs of) essential divisors

1) No need to check gcd(Q, N/Q) = 1: start from the factorization into
primary factors and it becomes automatic and we directly compute the 2^w
essential divisors (w = number of prime divisors of N).

2) If D is the list of sorted positive divisors (essential or not) and
l = lg(D), then N / D[i] = D[l - i].

17	11	src/basemath/ellanal.c

481f508132 | Karim Belabas | Sun, 14 Sep 2025 19:46:16 +0200

cleanup listheegner

- no need to maintain separately the hashtable size in 's': it's H.nb
- no real need to keep an exact count 't', the rough upper bound H.nb is
  good enough. And it avoids shadowing the (mostly useless) value by a
  GEN t from an inner loop

12	13	src/basemath/ellanal.c

d5cf964b8f | Bill Allombert | Sun, 14 Sep 2025 16:49:43 +0200

listheegner: rewrite to use a hash and try forms by increasing norms.

70	37	src/basemath/ellanal.c

0de3ffb7d4 | Bill Allombert | Sun, 14 Sep 2025 16:49:43 +0200

New libpari function hash_values_GEN

1	1	CHANGES
3	0	doc/usersch5.tex
1	0	src/headers/paridecl.h
14	0	src/language/hash.c

dee4399112 | Karim Belabas | Sun, 14 Sep 2025 13:47:53 +0200

lfunthetacost: avoid FPE when A = 0

1	1	src/basemath/lfun.c

1d27e3882d | Karim Belabas | Sun, 14 Sep 2025 10:57:14 +0200

polclass doc: extra whitespace

3	3	src/functions/polynomials/polclass

936455dd24 | Karim Belabas | Sun, 14 Sep 2025 10:39:59 +0200

polclass doc: extra whitespace

1	1	src/functions/polynomials/polclass

33db31530e | Karim Belabas | Sun, 14 Sep 2025 08:45:15 +0200

gphelp: don't split ! and = (see ??polclass)

1	1	doc/gphelp.in

4d12223a6e | Bill Allombert | Sat, 13 Sep 2025 23:55:58 +0200

BA  18- polclass: support for Atkin modular functions of level 3 to 19

1	1	CHANGES

2a27b55e27 | Bill Allombert | Sat, 13 Sep 2025 23:52:35 +0200

New libpari function Zn_sqrtall

1	0	CHANGES
5	0	doc/usersch5.tex
24	0	src/basemath/quad.c
1	0	src/headers/paridecl.h

718be2c9c9 | Bill Allombert | Sat, 13 Sep 2025 23:52:35 +0200

Flx_genus2charpoly_naive: unroll diffnext

14	2	src/basemath/hyperell.c

5dfbd9b6f9 | Karim Belabas | Sat, 13 Sep 2025 19:31:08 +0200

lfun: make test arch-independent (and more meaningfull)

3	3	src/test/32/lfun
2	2	src/test/in/lfun

b27144eac9 | Karim Belabas | Fri, 12 Sep 2025 20:28:53 +0200

loss of accuracy when computing lfuntheta [#2634]

take into account the magnitude of a_n K^(m)(nt/sqrt(N)) to increase
accuracy and control absolute error.

1	0	CHANGES
1	0	COMPAT
7	1	doc/usersch8.tex
2	2	src/basemath/hgm.c
13	32	src/basemath/lfun.c
4	4	src/basemath/mftrace.c
1	1	src/headers/paridecl.h

15fe1ce554 | Karim Belabas | Fri, 12 Sep 2025 15:52:55 +0200

lfunthetainit: increase extrabit when Vgaeasytheta

If parameters at infinity are of the form [a] or [a,a+1] with huge 'a',
then Fourier coefficients (which are multiplied by n^a) and t/sqrt(N)
must be computed with extra accuracy

28	5	src/basemath/lfun.c
3	1	src/test/32/lfun
5	0	src/test/in/lfun