I tried to compute associated_primes and it works, maybe it is related to the bug.
El miércoles, 31 de mayo de 2023 a las 22:20:43 UTC+2, enriqu...@gmail.com escribió: > I guess it can be possible to get a smaller example and it may be possible > that this code has other problems. My goal is to compute the Fitting ideals > of the "abelian" Alexander matrix of a finitely presented group. If you > execute and hold the two last paragraphs, for me the execution time of the > first of those paragraphs is quite fast while the last one takes really > long (I have no patience to say how long if it ends). > I try to provide you a shorter code starting directly with the matrix > (print out the list of coefficients and recreate it) but in that case both > codes were fast. > So maybe the problem is not on Laurent polynomials but I still share it > just in case. > Thanks, Enrique. > > El miércoles, 31 de mayo de 2023 a las 15:24:46 UTC+2, Kiran Kedlaya > escribió: > >> I wrote that code into the Sage library and it is supposed to be doing >> exactly what you proposed (clear denominators and do the computation in the >> polynomial ring). It is possible the regression was triggered by an update >> to Singular; it would indeed be helpful to identify a minimal example of >> the problem and then post it to a ticket. >> >> I believe there are also some upstream bugs with minimal associated >> primes, e.g., https://github.com/sagemath/sage/issues/29671. >> >> Kiran >> >> On Monday, May 29, 2023 at 8:53:59 PM UTC-7 Travis Scrimshaw wrote: >> >>> Dear Enrique, >>> I am having a bit of trouble understanding exactly what computations >>> are slow and fast from your description. As Nils said, can you give us some >>> explicit code (with some comments about which parts are slow)? >>> >>> Best, >>> Travis >>> >>> On Tuesday, May 30, 2023 at 3:28:39 AM UTC+9 Nils Bruin wrote: >>> >>>> Dear Enrique, >>>> >>>> From what you write I get the impression you may be talking about a >>>> regression in performance relative to earlier versions of sage. If you >>>> want >>>> to make an actionable item out of this, you'll probably have to file a >>>> ticket with explicit code on it that can be profiled; preferably with an >>>> indication why you think the performance could be significantly improved. >>>> That doesn't guarantee someone will work on it but it at least gives them >>>> a >>>> place to start if they want to, including you yourself! You could file it >>>> as an "enhancement" or even as a "bug" if you can convincingly show it's a >>>> regression. In the latter case you would probably end up identifying a >>>> version in which performance was significantly better. A git diff on some >>>> of the relevant files could then perhaps very quickly show what's >>>> happening. >>>> >>>> >>>> >>>> On Monday, 29 May 2023 at 09:07:07 UTC-7 enriqu...@gmail.com wrote: >>>> >>>>> Some time ago I had some computations on ideals in Laurent polynomial >>>>> rings, namely looking for minimal associated primes. Basically, I >>>>> converted >>>>> any generator into a polynomial, study the ideal in the polynomial ring, >>>>> and forget the prime ideals containing monomials. From some time ago, it >>>>> is >>>>> much easier since it can be done directly in the ring of Laurent >>>>> polynomials. >>>>> Yesterday these computations on an ideal with 80 generators were >>>>> really slow, but for some reason I checked that if the generators were >>>>> converted to elements in the associated polynomial ring, and then the >>>>> ideal >>>>> in the Laurent polynomial ring is constructed, then those computations >>>>> were >>>>> solved really fast. >>>>> I checked the code but I was not able to isolate the reason. Best, >>>>> Enrique. >>>>> >>>> -- 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/9d9668c7-1bce-4aa1-ad86-9c81e3f53b60n%40googlegroups.com.
