On 24/11/2019 20:04, Marc Mezzarobba wrote:
Marc Mezzarobba wrote:
sage: def f():
: for i in [x]range(1):
: a+a
where a = Integer(1).
If a is made local to the function, the very expensive lookups
disappear, and we still see calls to PyLong_FromLong taking much
lon
Marc Mezzarobba wrote:
>> sage: def f():
>> : for i in [x]range(1):
>> : a+a
>
> where a = Integer(1).
>
> If a is made local to the function, the very expensive lookups
> disappear, and we still see calls to PyLong_FromLong taking much
> longer than PyInt_FromLong, as
Marc Mezzarobba wrote:
> Some profiling data (via linux-perf) for
>
> sage: def f():
> : for i in [x]range(1):
> : a+a
where a = Integer(1).
If a is made local to the function, the very expensive lookups
disappear, and we still see calls to PyLong_FromLong taking mu
Some profiling data (via linux-perf) for
sage: def f():
: for i in [x]range(1):
: a+a
Py2:
Samples: 26K of event 'cycles', Event count (approx.): 17630706503
Children Self Command Shared ObjectSymbol
+ 77,85%47,79% python2 libpython2.7.so.1.0
Vincent Delecroix wrote:
> Then I guess the reason of the slowdown comes from the change in
> the integer types in Python 3 and the way we handle the conversion
> from Python integers to Sage integers
This may be part of it, but I don't think it explains everything:
sage: a = 1
sage: %timeit a+a
Then I guess the reason of the slowdown comes from the change in
the integer types in Python 3 and the way we handle the conversion
from Python integers to Sage integers.
Namely we have low level mpz_set_pylong (for "long" integers) and
mpz_set_si (for machine integers, which is a GMP function).
On Saturday, November 23, 2019 at 12:38:35 PM UTC-8, vdelecroix wrote:
>
> Le 23/11/2019 à 11:34, John H Palmieri a écrit :
> > By the way, Integer(1r) is also faster with Python 2 than with Python 3.
>
> By how much? Does it explain the 25% slowdown of the original post?
>
Python 2:
sage: %
Le samedi 23 novembre 2019 20:34:51 UTC+1, John H Palmieri a écrit :
>
> It's not actually Sage 8.9 vs. 9.0, it is? Rather it's Python 2 vs. Python
> 3.
>
I confirm this: since at least Sage 8.5, Python 3 is roughly 10% slower
than Python 2 in running the manifold doctests via
sage -tp --long
Le 23/11/2019 à 11:34, John H Palmieri a écrit :
By the way, Integer(1r) is also faster with Python 2 than with Python 3.
By how much? Does it explain the 25% slowdown of the original post?
--
You received this message because you are subscribed to the Google Groups
"sage-devel" group.
To uns
It's not actually Sage 8.9 vs. 9.0, it is? Rather it's Python 2 vs. Python
3. On my computer, I see the same timings with Python 3 builds of 8.9 and
9.0.beta6, but things look faster with Python 2 builds of 8.9 and 9.0.beta6.
By the way, Integer(1r) is also faster with Python 2 than with Python
And could you also time compare
a = Integer(1r)
on both versions?
Le 23/11/2019 à 11:13, Marc Mezzarobba a écrit :
Vincent Delecroix wrote:
@Marc: could you perform some C profiling (it might work directly
inside Sage via %crun [2]).
Yes, I'll try to investigate a bit more, but I first
Vincent Delecroix wrote:
> @Marc: could you perform some C profiling (it might work directly
> inside Sage via %crun [2]).
Yes, I'll try to investigate a bit more, but I first wanted to ask if it
was a known issue.
--
Marc
--
You received this message because you are subscribed to the Google
12 matches
Mail list logo
