On Wednesday 10 December 2008, mabshoff wrote:
> On Dec 9, 10:36 am, Martin Albrecht <[EMAIL PROTECTED]>
>
> wrote:
> > Hi there,
>
> Hi Martin,
>
> > I'm trying to understand the following memleak without success. Consider
> > three functions:
> >
> > def f(X):
> > try:
> > return ZZ(X)
> > except TypeError:
> > X = ZZ(map(ZZ,X),2)
> > return X
> >
> > def g(X):
> > try:
> > return ZZ(X)
> > except TypeError:
> > pass
> >
> > def h(X):
> > X = ZZ(map(ZZ,X),2)
> > return X
> >
> > As you can see below, f leaks:
> >
> > m = get_memory_usage()
> > for i in xrange(10^3):
> > _ = f([1,0,1])
> > print get_memory_usage() - m
> >
> > 4.72265625
> >
> > while g and h are clean:
> >
> > m = get_memory_usage()
> > for i in xrange(10^3):
> > _ = g([1,0,1])
> > print get_memory_usage() - m
> >
> > 0.0
> >
> > m = get_memory_usage()
> > for i in xrange(10^3):
> > _ = h([1,0,1])
> >
> > 0.0
> >
> > So it seems each component of the function is clean but their combination
> > leaks. Any ideas?
>
> My guess would be that maybe another coercion is involved when doing
> running h. There have been several issues exposed recently where the
> new coercion's caching mechanism did not work and subsequently did
> start leaking memory. Two examples are #4639 and #4740. The first one
> is still open while the second one has a fix that unfortunately causes
> one failure in the pickle jar.
>
> Other than that I have no idea yet what specifically can go wrong
> here. I assume you ran valgrind on the example?
I didn't spot anything obvious in the valgrind logs so far. However, another
datapoint.
This still leaks.
def f(X):
try:
return ZZ([1,0])
except TypeError, msg:
return ZZ(2)
m = get_memory_usage()
for i in xrange(10**4):
_ = f([1,0,1])
print get_memory_usage() - m
Also, if I replace ZZ by QQ, QQbar, RDF or RR it still leaks:
ZZ: 47.453125
RR: 55.6328125
QQ: 55.5625
QQbar: 25.171875
but I assume all these call Integer.__init__ at some point. However, GF(127)
does not leak.
Cheers,
Martin
--
name: Martin Albrecht
_pgp: http://pgp.mit.edu:11371/pks/lookup?op=get&search=0x8EF0DC99
_www: http://www.informatik.uni-bremen.de/~malb
_jab: [EMAIL PROTECTED]
--~--~---------~--~----~------------~-------~--~----~
To post to this group, send email to sage-devel@googlegroups.com
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at http://groups.google.com/group/sage-devel
URLs: http://www.sagemath.org
-~----------~----~----~----~------~----~------~--~---
