I would guess that a significant amount of the gain is that he doesn't have to len() the list every iteration, plus the item unpacking occurs in C. I don't know how much JIT would affect anything unless you ran the tests in PyPy.
On Sun, May 20, 2018, 23:51 Ian Mallett <i...@geometrian.com> wrote: > On Sun, May 20, 2018 at 9:25 PM, Daniel Foerster <pydsig...@gmail.com> > wrote: > >> The relevance of N exponent versus the discarded coefficients depends on >> how big N may be. With the likely sizes of N in a Pygame class, the >> difference between the algorithms seems probably negligible. A quick test >> with 20% deletion shows that your algorithm becomes more efficient around >> N=7000, but either can handle well over N=20000 in 10ms. Of course, a list >> comprehension can handle almost 100,000 in the same 10ms so the point is >> all rather moot in real-world code ;) >> > > These results were surprising to me, so I coded up a test (see attached!). > > I measure the breakeven point at a similar level, 5500 elements. MrGumm's > version is functionally equivalent to mine, but seems to run much faster > than either of ours (mine and Daniel's)—almost as fast as the list > comprehension, which I confirm to be fastest. My guess is that my `while` > loop (instead of `for` over the elements explicitly) is confusing the > Python JIT into not optimizing somehow. > > Also, this discussion on StackOverflow is relevant > <https://stackoverflow.com/questions/5745881/fast-way-to-remove-a-few-items-from-a-list-queue> > . > > Ian >
