Opened 8 years ago
Closed 8 years ago
#12903 closed defect (fixed)
Memory leaks with CPLEX
Reported by: | ncohen | Owned by: | ncohen |
---|---|---|---|
Priority: | major | Milestone: | sage-5.1 |
Component: | linear programming | Keywords: | |
Cc: | dcoudert | Merged in: | sage-5.1.beta0 |
Authors: | Nathann Cohen | Reviewers: | David Coudert |
Report Upstream: | N/A | Work issues: | |
Branch: | Commit: | ||
Dependencies: | Stopgaps: |
Description
Some memory leaks remained in Sage's CPLEX interface, as reported by Geoff on http://ask.sagemath.org/question/1170/memory-blowup-with-milp
Fixed there, after a lot of debugging :-P
You can check the difference by running the following piece of code :
while True: p = MixedIntegerLinearProgram(solver = "CPLEX") get_memory_usage()
The result is scary.
Nathann
Attachments (1)
Change History (9)
comment:1 Changed 8 years ago by
- Status changed from new to needs_review
comment:2 Changed 8 years ago by
- Status changed from needs_review to needs_work
comment:3 Changed 8 years ago by
- Status changed from needs_work to needs_review
Hellooooooo !!
Well, it actually isn't related to try/catch (your LP computes a vertex cover --> it never fails) but there is a leak indeed :-D
I reduced your code to that and the problem remains
for i in xrange(5000): p = MixedIntegerLinearProgram(solver = "CPLEX", maximization = False) b = p.new_variable(dim=1) for u,v in [(0, 1), (0, 4), (0, 5), (1, 2), (1, 6), (2, 3), (2, 7), (3, 4), (3, 8), (4, 9), (5, 7), (5, 8), (6, 8), (6, 9), (7, 9)]: p.add_constraint(b[u]+b[v] >= 1) print get_memory_usage()
That is fixed by some additional sage_free calls.... All the method had them except two, add_linear_constraint included. -_-
By the way, I am a bit scared... Do the tests pass for you when having CPLEX installed as the default solver ? I see a broken docstring in generic_graph in the traveling_salesman_problem method. The bug is not realted to this ticket (it appears regardless of whether the current patch is applied) and I have no idea where it comes from O_o
Nathann
Changed 8 years ago by
comment:4 Changed 8 years ago by
Which version are you using? I have no docstring error with 5.0.rc0.
D.
comment:5 Changed 8 years ago by
I'm using rc0 too O_o
So for you this ticket passes all tests ? Well, as I said earlier the doctests breaks on my machine regardless of whether it is applied anyway... O_o
Nathann
comment:6 Changed 8 years ago by
- Reviewers set to David Coudert
- Status changed from needs_review to positive_review
That's much better!!!
From my side the patch is ready to go (long tests, docbuild, stable memory,...) and more than useful ;)
comment:7 Changed 8 years ago by
Thaaaaaaaanks ! :-)
Nathann
comment:8 Changed 8 years ago by
- Merged in set to sage-5.1.beta0
- Resolution set to fixed
- Status changed from positive_review to closed
Hello,
the patch is working (install OK, long tests OK, docbuild OK). However, we still have memory leaks when using try except.
This example without try/except uses constant memory, but with try/except, memory is constantly increasing.