Here's a more extreme example, this time using bnfinit. Is this just a case of being *unlucky*?

? default(parisize,128m)

*** Warning: new stack size = 128000000 (122.070 Mbytes).

? \e1

echo = 1 (on)

? P=x^36-18*x^30+423*x^24-1836*x^18+3807*x^12-1458*x^6+729;

? v=

{

Vecsmall([-4546443891339392291, -4772614274774294850, -2246288600641157998,

2184519784396885847, 3855338830230007904, 8969375944544608132, 8608265968228

183757, 6273415500681370970, 5885195252242656489, 4767171491269281381, -1272

058068841457002, -8522898805058578675, -3462995918748747522, -37440823470562

73941, -7588280442232278606, -800330673691223428, 1601000787917156101, -6169

786316111353257, 6158961020464245362, 7259667069719318640, -5591369722437293

671, -7184458323559833952, 101100876029711980, 4218807242805716106, 23395138

63279706106, -377235584165752177, -7653484989466440273, -1460164831741692248

, -5948676288876253437, -4320233427562666956, 5866042301698983785, 421713566

8661605137, -8234089220656111209, 2825767952050423408, -1128535131387028106,

7063499823286387499, 5068506984690214898, 7107288543298231525, -27827074261

0567964, 2824508415623784311, 5570453429194826236, -2831564018313687675, 747

615814731296686, -5453817830642374251, 2743082897957633410, -591049344346418

6517, -1243957834720937782, -8076888079980094075, 3289857532817318357, -8460

756364696504252, 2203151030024894821, 6956170283179131766, -1723001446203469

933, -8940722501990547006, 4515347364898046942, 7083087219250249904, 6576555

527558317268, -7708175369883607583, -8063160973016529272, -61927664745483597

16, 7376235850437509968, 2633129359848840833, 2465638921812218602, 270954623

9851907100, 19, -4255074600525148973]);

}

? #

timer = 1 (on)

? setrand(1);bnfinit(P);

time = 3,436 ms.

? setrand(v);bnfinit(P);

time = 1min, 4,577 ms.

On Tue, Nov 13, 2012 at 4:03 PM, Igor Schein <igorschein@gmail.com> wrote:

Package: pari
Version: git-30598f4

? setrand(756);while(1,gettime;rnfkummer(bnrinit(bnfinit(quadpoly(13829,y)),211,1),[5,1;0,1]);print(gettime))

5516
6340
5873

5493
5432
5617
5588
15386
5652
5617
5429
5516
5465
5712
5685
5328
^C *** at top-level: ...and(756);while(1,gettime;rnfkummer(bnrinit(bn

*** ^--------------------
*** rnfkummer: user interrupt after 1min, 40,690 ms.

Nothing catastrophic by any means, but I was curious if there's a reason for the 8th iteration to be almost 3 times slower than the average - maybe there's a room for improvement somewhere.

I prepared this example a while ago waiting to report it in case bug 910 patches wouldn't fix the stack problem. They did, but I didn't want to waste this example and decided to reuse it here.

Thanks

Igor