Thank you for the Flint reference, I will check it out, and thanks again for your hard work debugging!
Best, Håkan On Mon, May 6, 2024 at 9:57 AM Vincent Delecroix <20100.delecr...@gmail.com> wrote: > The fix was easy. Much harder was to find the root of the problem! > > If you are interested in intensive computations over QQbar I strongly > advise you to look towards https://flintlib.org/ maintained by Fredrik > Johansson. This library has *two* implementations of QQbar which are > way more performant (and probably more reliable) than what is in > SageMath. The more or less short term goal is to use flint > implementations to replace the current QQbar implementation in > SageMath.You could give it a try inside SageMath via > https://github.com/mezzarobba/flint_gr_sage (feel free to open an > issue there asking for installation help or describing the things you > want to compute). > > I do intend to make my commit part of SageMath QQbar, but it is not on > the top priority list. > > Best > Vincent > > On Mon, 6 May 2024 at 09:27, Hakan Granath <hakan.gran...@gmail.com> > wrote: > > > > Great, thanks for working out and fixing the (quite non-trivial, it > seems) root cause of this issue! > > > > Just a curious question: would it be useful to add the functionality of > your commit 8de6ba5 to potentially simplify QQbar objects? Especially doing > linear algebra over QQbar I guess you get many trivial operations (like > adding 0) and optimizing them away might maybe improve performance? Just an > idea, I haven't done any benchmarking to support that hypothesis. > > > > Best regards, > > > > Håkan > > > > On Mon, May 6, 2024 at 12:15 AM Vincent Delecroix < > 20100.delecr...@gmail.com> wrote: > >> > >> Turns out to be a serious bug in complex interval fields. The problem > >> is hopefully fixed by https://github.com/sagemath/sage/pull/37941 > >> which should make its way to the next sage release. > >> > >> Thanks again for your report. > >> Vincent > >> > >> On Fri, 3 May 2024 at 18:13, vdelecroix <20100.delecr...@gmail.com> > wrote: > >> > > >> > Thanks for your report! I simplified a bit your example and posted an > issue https://github.com/sagemath/sage/issues/37927. > >> > > >> > Vincent > >> > > >> > Le vendredi 3 mai 2024 à 15:05:03 UTC+2, Hakan Granath a écrit : > >> >> > >> >> Hi, > >> >> > >> >> I think sometimes matrices over QQbar give erroneous results (sorry > for the messy example, if I try to simplify it the problem disappears): > >> >> > >> >> R.<y> = QQ[] > >> >> v1 = QQbar.polynomial_root(AA.common_polynomial(y^2 + 1), > CIF(RIF(RR(0)), RIF(-RR(1)))) > >> >> v2 = QQbar.polynomial_root(AA.common_polynomial(y^2 + 1), > CIF(RIF(RR(0)), RIF(RR(1)))) > >> >> v3 = 4*v2 > >> >> v4 = AA.polynomial_root(AA.common_polynomial(y^2 - 2), > RIF(-RR(1.4142135623730951), -RR(1.4142135623730949))) > >> >> v5 = AA.polynomial_root(AA.common_polynomial(y^2 - 2), > RIF(RR(1.4142135623730949), RR(1.4142135623730951))) > >> >> v6 = QQbar.polynomial_root(AA.common_polynomial(y^2 + 16), > CIF(RIF(RR(0)), RIF(-RR(4)))) > >> >> v7 = v6*v6 > >> >> v8 = QQbar.polynomial_root(AA.common_polynomial(y^2 + 1), > CIF(RIF(RR(0)), RIF(RR(1)))) > >> >> > >> >> M = matrix(QQbar, [[0, 0, 1, 0, 0, 0, 0, 0, 0, 0], [0, 1, 0, 0, 0, > 0, 0, 0, 0, 0], [-4, 2*v1, 1, 64, -32*v1, -16, 8*v1, 4, -2*v1, -1], [4*v1, > 1, 0, -192*v1, -80, 32*v1, 12, -4*v1, -1, 0], [2, 0, 0, -480, 160*v1, 48, > -12*v1, -2, 0, 0], [-4, 2*v2, 1, 64, -32*v2, -16, 8*v2, 4, -2*v2, -1], [v3, > 1, 0, -192*v2, -80, 32*v2, 12, -v3, -1, 0], [2, 0, 0, -480, 160*v2, 48, > -12*v2, -2, 0, 0], [0, 0, 0, 8, QQbar(4*v4), 4, QQbar(2*v4), 2, > QQbar(AA.polynomial_root(AA.common_polynomial(y^2 - 2), > RIF(-RR(1.4142135623730951), -RR(1.4142135623730949)))), 1], [0, 0, 0, > QQbar(24*v4), 20, QQbar(8*v4), 6, QQbar(2*v4), 1, 0], [0, 0, 0, 8, > QQbar(4*v5), 4, QQbar(2*v5), 2, > QQbar(AA.polynomial_root(AA.common_polynomial(y^2 - 2), > RIF(RR(1.4142135623730949), RR(1.4142135623730951)))), 1], [0, 0, 0, > QQbar(24*v5), 20, QQbar(8*v5), 6, QQbar(2*v5), 1, 0], [0, 0, 0, v7*v7*v7, > v7*v7*v6, v7*v7, v6*v6*v6, v6*v6, v6, 1], [0, 0, 0, -4096, 1024*v8, 256, > -64*v8, -16, 4*v8, 1]]) > >> >> > >> >> With this matrix I get: > >> >> > >> >> sage: M.right_kernel_matrix() > >> >> [] > >> >> > >> >> but in fact the right kernel is generated by: > >> >> > >> >> sage: v = Matrix(QQbar, 10, 1, [-108, 0, 0, 1, 0, 12, 0, -60, 0, > 64]) > >> >> sage: M * v == 0 > >> >> True > >> >> > >> >> This is with SageMath version 10.3 using Python 3.11.1 on Ubuntu > 22.04.4. > >> >> > >> >> Best regards, > >> >> > >> >> Håkan Granath > >> > > >> > -- > >> > You received this message because you are subscribed to the Google > Groups "sage-devel" group. > >> > To unsubscribe from this group and stop receiving emails from it, > send an email to sage-devel+unsubscr...@googlegroups.com. > >> > To view this discussion on the web visit > https://groups.google.com/d/msgid/sage-devel/1c927686-bada-4943-8000-a6fdf6e71029n%40googlegroups.com > . > >> > >> -- > >> You received this message because you are subscribed to the Google > Groups "sage-devel" group. > >> To unsubscribe from this group and stop receiving emails from it, send > an email to sage-devel+unsubscr...@googlegroups.com. > >> To view this discussion on the web visit > https://groups.google.com/d/msgid/sage-devel/CAGEwAAkgsAcFmt-6NKWvuh0va_UVy-jdSJVnPrSrjGE_CA3Nog%40mail.gmail.com > . > > > > -- > > You received this message because you are subscribed to the Google > Groups "sage-devel" group. > > To unsubscribe from this group and stop receiving emails from it, send > an email to sage-devel+unsubscr...@googlegroups.com. > > To view this discussion on the web visit > https://groups.google.com/d/msgid/sage-devel/CAOnrL6%2BO0s%3D6poKYFo17xTLrrX4SdCFuj%2BQj7tpkjtQfL%2BvQNw%40mail.gmail.com > . > > -- > You received this message because you are subscribed to the Google Groups > "sage-devel" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to sage-devel+unsubscr...@googlegroups.com. > To view this discussion on the web visit > https://groups.google.com/d/msgid/sage-devel/CAGEwAAnudV9KV1daN%3DY5YCC3btCf7%3D%3D9hc9rRTmuX0Ck%2BRs8ww%40mail.gmail.com > . > -- You received this message because you are subscribed to the Google Groups "sage-devel" group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/sage-devel/CAOnrL6KYrgAt3D5OqMw8%3DRbp2zQMSqCZ0upFT14wbK5pHc8Whw%40mail.gmail.com.
