regression between 6.8 and 6.9.beta1

[Sébastien Labbé]

Can you confirm that you get the same regression as I do?

Sébastien

With sage 6.8:

$ sage nbruin.sage
Time: CPU 2.48 s, Wall: 2.49 s
Time: CPU 2.13 s, Wall: 2.16 s
Time: CPU 33.20 s, Wall: 34.11 s

With sage 6.9.beta1:

$ sage nbruin.sage
Time: CPU 7.83 s, Wall: 8.07 s
Time: CPU 8.07 s, Wall: 8.38 s
Time: CPU 78.33 s, Wall: 79.96 s

The code in nbruin.sage is taken from comment http://trac.sagemath.org/ticket/14058#comment:4

 $ cat nbruin.sage
a=ZZ(1)
b=QQ(1)
c=ZZ['x'](1)
d=b+c

def test(x,y):
  for i in xrange(10^6):
    _=x+y

time test(a,b)
time test(a,c)
time test(b,c)

[Vincent Delecroix]

Hello,

I think that we should seriously start a regression test suite as this
is not the first time!

Does anybody already started something? Has suggestion?

Regression tests might even be included in the patchbot.

Vincent

[Nathann Cohen]

Can you confirm that you get the same regression as I do?

Can you profile the code in both versions of Sage, to see where the regression comes from?

Nathann 

[Sébastien Labbé]

> Can you profile the code in both versions of Sage, to see where the regression comes from?

%prun doesn't say much (note that I am compiling sage at the same time which explains the 15s)...

With sage 6.8 :
 

sage: %prun test(a,b)
         3 function calls in 2.799 seconds

   Ordered by: internal time

   ncalls  tottime  percall  cumtime  percall filename:lineno(function)
        1    2.799    2.799    2.799    2.799 <string>:10(test)
        1    0.000    0.000    2.799    2.799 <string>:1(<module>)
        1    0.000    0.000    0.000    0.000 {method 'disable' of '_lsprof.Profiler' objects}

With sage 6.9.beta1 :

sage: %prun test(a,b)
         3 function calls in 15.118 seconds

   Ordered by: internal time

   ncalls  tottime  percall  cumtime  percall filename:lineno(function)
        1   15.118   15.118   15.118   15.118 <string>:10(test)
        1    0.000    0.000   15.118   15.118 <string>:1(<module>)
        1    0.000    0.000    0.000    0.000 {method 'disable' of '_lsprof.Profiler' objects}

[Sébastien Labbé]

>Can you confirm that you get the same regression as I do?

I need somebody to confirm because my version of sage-6.8 was not clean.

[Daniel Krenn]

On 2015-08-11 12:00, Sébastien Labbé wrote:
> Can you confirm that you get the same regression as I do?
>
> Sébastien
>
> With sage 6.8:
>
> $ sage nbruin.sage
> Time: CPU 2.48 s, Wall: 2.49 s
> Time: CPU 2.13 s, Wall: 2.16 s
> Time: CPU 33.20 s, Wall: 34.11 s
>
> With sage 6.9.beta1:
>
> $ sage nbruin.sage
> Time: CPU 7.83 s, Wall: 8.07 s
> Time: CPU 8.07 s, Wall: 8.38 s
> Time: CPU 78.33 s, Wall: 79.96 s

Cannot confirm this:

dakrenn@pc72-math /local/dakrenn/sage $ 6.8/sage nbruin.sage
Time: CPU 0.59 s, Wall: 0.59 s
Time: CPU 0.55 s, Wall: 0.55 s
Time: CPU 6.73 s, Wall: 6.73 s
dakrenn@pc72-math /local/dakrenn/sage $ 6.9.beta0/sage nbruin.sage
Time: CPU 0.57 s, Wall: 0.57 s
Time: CPU 0.54 s, Wall: 0.54 s
Time: CPU 7.03 s, Wall: 7.04 s
dakrenn@pc72-math /local/dakrenn/sage $ 6.9.beta1/sage nbruin.sage
Time: CPU 0.59 s, Wall: 0.59 s
Time: CPU 0.55 s, Wall: 0.55 s
Time: CPU 6.84 s, Wall: 6.85 s

Best

Daniel

[Sébastien Labbé]

Cannot confirm this:

Okay, thanks for checking! So the hypothesis now is that it is due to an improvement from #14058 which was applied on my sage-6.8. Thanksfully!

Sébastien

[Justin C. Walker]

I likewise don’t see what Sebastien sees:
% /Users/Sage/sage-6.9.beta1/sage x.sage
Time: CPU 0.85 s, Wall: 0.85 s
Time: CPU 0.65 s, Wall: 0.65 s
Time: CPU 7.47 s, Wall: 7.47 s
% sage x.sage
Time: CPU 0.84 s, Wall: 0.84 s
Time: CPU 0.64 s, Wall: 0.64 s
Time: CPU 7.49 s, Wall: 7.49 s
% sage -version
SageMath Version 6.8, Release Date: 2015-07-26

This is on OS X, 10.10.4 (Quad-core Core i7 (2.6 GHz).


--
Justin C. Walker, Curmudgeon at Large
Director
Institute for the Enhancement of the Director's income
-----------
--
They said it couldn't be done, but sometimes,
it doesn't work out that way.
- Casey Stengel
--