another angle os

[Igor Schein <igor@txc.com>]

Hi,

? setrand(7);rnfkummer(bnrinit(bnfinit(quadpoly(-391483,y)),2,1),Mat(3))
  ***   inconsistent data in hnfdivide.

How let's look at the timing pattern:

? gettime;for(k=1,30,if(k!=3,setrand(7);rnfkummer(bnrinit(bnfinit(quadpoly(-391483,y),,[k/10,k/10]),2,1),Mat(3));print(k"
"gettime)))
1 2226
2 2212
4 600
5 585
6 485
7 2218
8 2218
9 491
10 2224
11 476
12 2216
13 2219
14 501
15 502
16 2240
17 2244
18 629
19 634
20 629
21 637
22 2248
23 645
24 646
25 645
26 647
27 623
28 627
29 626
30 2264

So except .3, all other values produce a valid answer, but some more
efficiently, some less efficiently, and the timing is consistent
within each of the 2 groups.  Luck factor?

Thanks

Igor