[with patch, needs work] avoid creating unneeded homsets when coercing

[mabshoff]

Burcin reported at #4639:

sage: F = GF(13)
sage: get_memory_usage()
708.02734375
sage: for _ in xrange(10000):
....:     t = F.coerce(F(234234))
....:     
sage: get_memory_usage()
728.15234375
sage: for _ in xrange(100000):
    t = F.coerce(F(234234))
....:     
sage: get_memory_usage()
932.3125
sage: for _ in xrange(100000):
    t = F.coerce(F(234234))
....:     
sage: get_memory_usage()
1136.35546875

Since the patch by RobertWB at that ticket fixes the issue Burcin reported, but not the original one I am moving it over to this ticket.

Cheers,

Michael

[mabshoff]

Patch by RobertWB, originally from #4639

[mabshoff]

Patch by RobertWB, review by William.

Cheers,

Michael

[mabshoff]

This patch causes a failure with the pickle jar:

**********************************************************************
File "/scratch/mabshoff/release-cycle/sage-3.2.2.alpha1/devel/sage/sage/structure/sage_object.pyx", line 371, in __main__.example_16
Failed example:
    sage.structure.sage_object.unpickle_all(std)###line 682:_sage_    >>> sage.structure.sage_object.unpickle_all(std)
Expected:
    doctest:...: DeprecationWarning: Your data is stored in an old format. Please use the save() function to store your data in a more recent format.
    Successfully unpickled ... objects.
    Failed to unpickle 0 objects.
Got:
    doctest:1172: DeprecationWarning: Your data is stored in an old format. Please use the save() function to store your data in a more recent format.
    ** failed:  _class__sage_rings_finite_field_morphism_FiniteFieldHomset__.sobj
    Failed:
    _class__sage_rings_finite_field_morphism_FiniteFieldHomset__.sobj
    Successfully unpickled 453 objects.
    Failed to unpickle 1 objects.
**********************************************************************
1 items had failures:
   1 of   7 in __main__.example_16

Cheers,

Michael

[mabshoff]

Once the pickling issue is fixed this should go in.

Cheers,

Michael

[mabshoff]

Robert,

the leak seems to be gone in 3.2.2:

----------------------------------------------------------------------
| Sage Version 3.2.2, Release Date: 2008-12-18                       |
| Type notebook() for the GUI, and license() for information.        |
----------------------------------------------------------------------
sage: F = GF(13)
sage: get_memory_usage()
689.02734375
sage: for _ in xrange(10000):
....:     t = F.coerce(F(234234))
....:     
sage: get_memory_usage()
689.02734375
sage: for _ in xrange(100000):
....:     t = F.coerce(F(234234))
....:     
sage: get_memory_usage()
689.02734375
sage: for _ in xrange(100000):
....:     t = F.coerce(F(234234))
....:     
sage: get_memory_usage()
689.02734375
sage: 

Should we close this ticket as won't fix or is the patch her still relevant? In that case we should update the ticket to reflect the actual issue being fixed here.

Cheers,

Michael

[robertwb]

The patch is still relevant here. What is going on is every time coerce is called, a new homset is created. Creating a new homset doesn't leak anymore, but it's still sub-optimal.

I don't know that I'll have time to look into this anytime soon, but it's not a blocker anymore as the leak is solved elsewhere.

[mabshoff]

ok,

I renamed the ticket then.

Cheers,

Michael

[SimonKing]

Hi Robert,

accidentally I found this rather old ticket.

Replying to robertwb:

The patch is still relevant here. What is going on is every time coerce is called, a new homset is created. Creating a new homset doesn't leak anymore, but it's still sub-optimal.

Is this still the case? In what situation is the homset created? Is it not cached? In what files does it occur?

Cheers, Simon