#20693: Sage crashes when computing newforms
---------------------------------+------------------------
Reporter: ehlen | Owner:
Type: defect | Status: new
Priority: critical | Milestone: sage-7.3
Component: modular forms | Resolution:
Keywords: | Merged in:
Authors: Stephan Ehlen | Reviewers:
Report Upstream: N/A | Work issues:
Branch: | Commit:
Dependencies: | Stopgaps:
---------------------------------+------------------------
Comment (by ehlen):
@jdemeyer
I'm sorry, I guess I misinterpreted the "Author" field. I probably won't
write code for this as I think the bugs and performance problems come from
relative extensions of number fields and I don't know much about the code
(and most of it is pari in some way, I guess).
My example in http://trac.sagemath.org/ticket/20693#comment:2 was missing
lines, sorry again.
I guess what I meant was
{{{
sage: D=DirichletGroup(23)
sage: c=D.gen()^2
sage: M=ModularSymbols(c,6,sign=1)
sage: S=M.cuspidal_subspace().new_subspace()
sage: Dec = S.decomposition()
sage: A=Dec[0]
sage: v =A.dual_eigenvector(names='a', lift=False)
}}}
To get a crash you have to let it run quite some time (I don't remember
how long exactly it was, maybe 15 minutes, I can restart it and let you
know).
I'm not sure if there much simpler/faster examples that crash but I can
check.
I can come up with more examples for sure. Indeed you have to wait quite a
bit until you see a crash sometimes, I had a computation running for
Gamma0(19) and weight 14 for 2 days or so until I saw the crash (but I
didn't run out of memory or anything else trivial and it also used to work
on the same machine with the same space (and sage should really be able to
do it!).
@nbruin I'll try to find some time to check out sage 6.1 again and compute
some spaces that also run in sage 7.2 and then get back to you with some
profiling results.
I did similar profiling tests as you did, @pbruin and discovered that the
main reason for this being slow is that elements in the base ring (a
cyclotomic field) get coerced into a relative extension (obtained by
adjoining the roots of a hecke polynomial). This is damn slow for some
reason (note that it is stated in the documentation of relative number
fields that doing arithmetic in number field towers is very slow but it
seems that it got worse than it used to be). I discovered that such
operations can be accelerated by very stupid means. For instance, instead
of coercing an element e of K into an extension L directly, write it in
terms of a power basis (e.list()), coerce the generator of K into L and
create the element corresponding to e in L directly. I can document some
examples later.
--
Ticket URL: <http://trac.sagemath.org/ticket/20693#comment:10>
Sage <http://www.sagemath.org>
Sage: Creating a Viable Open Source Alternative to Magma, Maple, Mathematica,
and MATLAB
--
You received this message because you are subscribed to the Google Groups
"sage-trac" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To post to this group, send email to [email protected].
Visit this group at https://groups.google.com/group/sage-trac.
For more options, visit https://groups.google.com/d/optout.
