#16411: Memory leak in polynomial evaluation
-------------------------+-------------------------------------------------
Reporter: SimonKing | Owner:
Type: defect | Status: new
Priority: major | Milestone: sage-6.3
Component: | Keywords: memory leak polynomial
commutative algebra | evaluation libsingular
Merged in: | Authors:
Reviewers: | Report Upstream: N/A
Work issues: | Branch:
Commit: | Dependencies:
Stopgaps: |
-------------------------+-------------------------------------------------
The following was reported at sage-support:
{{{
sage: C.<x,y,z> = GF(2)[]
sage: f = x^4+x*y^3+z^6
sage: a = f(1,0,0)
sage: get_memory_usage()
176.08984375
sage: for i in xrange(1000000):
....: a = f(1,0,0)
....:
sage: get_memory_usage()
198.08984375
sage: for i in xrange(1000000):
....: a = f(1,0,0)
....:
sage: get_memory_usage()
222.08984375
}}}
In the following I am citing Leif's comments from sage-devel:
In `singular_polynomial_call(&res, self._poly, _ring, coerced_x,
MPolynomial_libsingular_get_element)` we have
{{{
struct spolyrec
{
poly next; // next needs to be the first field
number coef; // and coef the second --- do not change
this !!!
unsigned long exp[VARS]; // make sure that exp is aligned
};
}}}
where both `next` and `coef` are pointers, and `VARS` is usually zero, so
`exp` is an "open-ended" array, such that the effective size of the struct
varies.
The leak depends on the values (and the amount also on the field
and the function). My impression is also that it appears whenever
res!=NULL, i.e., the result is non-zero.
The code in singular_polynomial_call() (in
src/sage/libs/singular/polynomial.pyx) explicitly prevents Singular from
reclaiming the memory:
{{{
...
ret[0] = res_id.m[0]
from_id.m[0] = NULL
res_id.m[0] = NULL
id_Delete(&to_id, r)
id_Delete(&from_id, r)
id_Delete(&res_id, r)
return 0
}}}
(`from_id.m[0]` was set to the input parameter `p`, so that's ok.)
Either it should make a garbage-collected copy of it (the result
ret / ret[0]) instead, or the caller has to clean up afterwards.
The docstring is quite misleading w.r.t. 'ret', as only the address of a
pointer to be changed is passed to the function, while the struct it
later points to always gets allocated by the callee, not the caller.
--
Ticket URL: <http://trac.sagemath.org/ticket/16411>
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 http://groups.google.com/group/sage-trac.
For more options, visit https://groups.google.com/d/optout.
