Changes commited to PARI GIT during the last two weeks

update bench

3	3	src/test/32/ellnf

elleisnum: add example

9	7	src/functions/elliptic_curves/elleisnum

14- elleisnum(elliptic curve, k): return exact result

1	0	CHANGES
39	12	src/basemath/elltrans.c
31	14	src/functions/elliptic_curves/elleisnum
40	8	src/test/32/ell
10	1	src/test/in/ell

rewrite modular corrections to take advantage of exact [w1,w2]

improves stability

36	30	src/basemath/elltrans.c
12	12	src/test/32/ell
1	1	src/test/in/ell

47- allow elleisnum(w, [k1, ..., kn])

1	0	CHANGES
88	29	src/basemath/elltrans.c
7	2	src/functions/elliptic_curves/elleisnum
1	0	src/headers/paridecl.h
4	1	src/test/32/ell
4	0	src/test/in/ell

fix comment typo

1	1	src/test/in/ell

46- [libpari] mulimag

1	0	CHANGES
3	0	doc/usersch5.tex
3	3	src/basemath/elliptic.c
20	0	src/basemath/gen3.c
1	1	src/basemath/mftrace.c
1	0	src/headers/paridecl.h
3	3	src/modules/stark.c
2	2	src/test/32/ell
3	3	src/test/32/ellnf

cleanup ellwp_cx interface

6	9	src/basemath/elltrans.c

31- elliptic functions ignored ellperiods(,1) quasi-period component

This is the description for the 2nd and following hunks of commit
b4cb096a06 which were meant as two successive commits but were unfortunately
committed all at once.

This commits also fixes a nasty problem when w1/w2 is on the border of the
fundamental domain, breaking ellperiod([w1,w2]) idempotence. Wich
becomes a problem when the input quasi-periods are taken into
consideration since red_modSL2 could change the lattice basis.

1	0	CHANGES

Simplify computation of eta(z) in ellheight_C

solve z = aw1 + bw2, a,b real;
  D = det([w1,w2; conj(w1),conj(w2)])
  a = det([z,w2; conj(z), conj(w2)]) / D
  b = det([w1,z; conj(w1),conj(z)]) / D
We have D = 2*I*imag(w1 * conj(w2)) and likewize a and b. The 2*I terms
cancel out. Saves a few multiplication and reduces to real computation
without introducing complex numbers with (random) ~ 0 components.

6	6	src/basemath/elliptic.c
92	82	src/basemath/elltrans.c
10	10	src/test/32/ell

whitespace edit

1	1	src/basemath/elltrans.c

doc typo

1	1	src/functions/elliptic_curves/elleisnum

cxEk, ellweierstrass: finish removing code duplication

always compute E4 & E6 sumultaneously and using the same code.
Also more precise.

83	73	src/basemath/elltrans.c
5	5	src/test/32/ell

rename _elleta -> elleta_W and elleta_i -> elleta_w + add comments

11	10	src/basemath/elltrans.c

update theta bench

12	13	src/test/32/theta

ellweierstrass: remove duplicate code and ad hoc interface

now uses the get_periods interface

34	70	src/basemath/elltrans.c

log(gtodouble()) -> dbllog2

1	1	src/basemath/mftrace.c

gtodouble(glog()) -> dbllog2

4	4	src/basemath/bibli1.c

improve comment

2	2	src/basemath/aprcl.c

gtodouble(glog()) -> dbllog2

2	2	src/basemath/QX_factor.c

gisdouble: return 0 instead of raising an error when input not of real type

conforms to the documentation

1	1	src/headers/pariinl.h

fix gisdouble documentation

4	3	doc/usersch5.tex

cleanup Fincke_Pohst_bound

5	4	src/basemath/buch2.c

heuristic initial precision increase in bnfinit to help small_snorm [#2632]

Commit 784f1ae504 removed this precision estimate
1) it had become possibly wrong when Fincke_Pohst_bound was introduced
2) it was used to reduce the precision of M to a "sufficient" value,
   potentially causing problems for a minor efficiency gain

Before that commit, the bound was valid to compute norms reliably from
floating point embeddings and nothing else. But as a side effect, it
could increase the initial nfinit precision if a t_POL was input. Which
would help in other places, such as small_norm's gaussred_from_QR which
is breaking here.

1) is not a serious concern, the value is of the right order of magnitude
and this is a one-sided error: at worse we would fail to compute largish
vectors which are unlikely to be smooth anyway.
2) can be circumvented: never reduce the precision, but possibly let the
   bound increase it.

So reinstate the bound as a better heuristic guess for the initial
precision. It won't hurt, it may help.

17	0	src/basemath/buch2.c
2	1	src/test/32/bnf
3	0	src/test/in/bnf

cxEk: use E4,E6 instead of G4,G6 in wp series + add 64 bits if k> 14

avoid useless multiplications by powers of 2pi

8	6	src/basemath/elltrans.c
2	2	src/test/32/ell

rewrite and test cxEk

use Weierstrass \wp series expansion for k> 14

27	34	src/basemath/elltrans.c
13	1	src/test/32/ell
1	0	src/test/in/ell

reorder elltrans.c + minor comments updates

1501	1498	src/basemath/elltrans.c

inline ellsigma_theta in ellsigma

10	16	src/basemath/elltrans.c

make elljacobi bench less cryptic

we test v/w against A

3	3	src/test/in/elljacobi

fix elljacobi bench

1	1	src/test/in/elljacobi

elljacobi cleanups

11	3	src/basemath/elltrans.c

restore elljacobi test that fails (not sure why)

1	1	src/test/32/elljacobi

delete useless comment

0	2	src/basemath/elltrans.c

rewrite elljacobi_pol + fix BIBs

48	43	src/basemath/elltrans.c
8	1	src/test/32/elljacobi
5	0	src/test/in/elljacobi

cleanup sigma/theta code, integrate old ellsigma_cx in ellsigma

86	156	src/basemath/elltrans.c
0	1	src/headers/paridecl.h
13	13	src/test/32/ell
3	3	src/test/32/ellnf
3	3	src/test/32/theta

thetaall_ii: don't negate thetanullwerte when z is reduced

This change fixes the following bug in ellsigma_cx:

? install(ellsigma_cx,"GGD0,L,p")
? ellsіgma_cx([2*I, 1], 4/3*I)
%1 = -0.55506326344653125121966647544195623187*I

? ellsigma([2*I,1], 4/3*I)
%2 = 0.55506326344653125121966647544195623186*I

For unknown reasons, it introduces a problem in elljacobi bench.
To be fixed later.

This commit is a prerequisite for the theta / elliptic functions cleanup
patch which follows (otherwise ellsigma becomes completely broken).

1	5	src/basemath/elltrans.c

cleanup elljacobi bench

8	10	src/test/in/elljacobi

ellmanintable typo: u may be a t_FRAC

1	1	src/modules/ellmanin.c

ellmanintable: avoid useless ellchangecurve

6	3	src/modules/ellmanin.c

ellmaninconstant: add flag: use Cremona table

1	0	CHANGES
17	1	src/basemath/ellisog.c
8	4	src/functions/elliptic_curves/ellmaninconstant
2	1	src/headers/paridecl.h
4	0	src/headers/paripriv.h
146	0	src/modules/ellmanin.c
3	39	src/test/32/ellmanin
23	2	src/test/in/ellmanin

ZV_lcm: use gen_product directly

7	1	src/basemath/polarit2.c

test-help: update

1	0	src/test/32/help

doc: ?\: add \qf

1	0	src/language/gplib.c

issquare: use Zn_issquare

1	1	src/basemath/ispower.c

ellwp_cx, ellzeta_cx: simplify by assuming input is reduced

+ remove useless GC

18	19	src/basemath/elltrans.c