By the way, the "small" regression in my timings you posted above is actually non-existent. Those are the first one I ran (and posted on #715). Afterward, I ran other "make ptestlong" and the the variance was big enough for the above difference to be meaningless. I mean I sometimes got more than the initial test with vanilla+715 in a subsequent test with vanilla only and less than the initial test with vanilla only with vanilla+715.
The test proposed by Robert and your implementation confirms that. As I just posted on #715, I guess that we indeed delete some actions that could get reused during the test suite, but we also access dictionnaries quickly so there is no regression. On 18 jan, 12:56, Simon King <simon.k...@uni-jena.de> wrote: > Hi Robert, > > Hooray! > > On 18 Jan., 00:20, Robert Bradshaw <rober...@math.washington.edu> > wrote: > > > Both, but primarily the latter. It's a microbenchmark, but loop like > > > a = Integer(10) > > b = QQ(20) > > s = RDF(30) > > for x in range(10**n): > > s += a*b*x > > > should give us an upper bound on how expensive any changes could be. > > I did the following on my laptop: > sage: def test(n): > ....: a = Integer(10) > ....: b = QQ(20) > ....: s = RDF(30) > ....: for x in xrange(10**n): > ....: s += a*b*x > > And then, sage-5.0.prealpha0+#11780 yields > sage: %time test(6) > CPU times: user 7.25 s, sys: 0.04 s, total: 7.29 s > Wall time: 7.31 s > whereas adding #715 yields > sage: %time test(6) > CPU times: user 7.29 s, sys: 0.01 s, total: 7.31 s > Wall time: 7.31 s > > So, no difference whatsoever! > > > (And yes, people write code like this...) Maybe a similar test with a > > tower of small finite fields. > > I don't understand what that would look like. > > I'll update the trac ticket with your example. > > Thank you, > Simon -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
