Programming puzzles: processing lists!
|
06-14-2017, 02:34 PM
Post: #155
|
|||
|
|||
RE: Programming puzzles: processing lists!
(06-14-2017 08:04 AM)pier4r Wrote: Nice. and 125 seconds vs 4800 is quite an improvement. Because, at least from my pov, I do believe that the average use case of rotating a list is not rotating it once, but multiple times. I agree with you about the multiple rotations, which is part of the reason I went ahead and did this. To be fair, this is (for a 50g) a fairly extreme situation due to the list size. Exploding a list and then rebuilding it leaves behind unreferenced objects; the number of discards will be (at a minimum) equal to the number of objects in the list + the original list. When you do that rapidly in succession, the "garbage" builds up quickly. It's inevitable that a GC will take place, and when it does, it has to locate/flag-objects-for-disposal/re-compress all of the space they were inhabiting. The changes I made still leave a large block of memory disposed, but it's contiguous and easily "skippable". When watching a loop make 1000 calls to LROT (v7) operating on a list of 1000 elements, you can barely discern when the GCs occur. The previous version left 1000 disposable objects behind every iteration, so the GCs would lock up the system for a painful period every time they happened. I should note that this phenomenon is not unique to my library; any activity (whether built-in or user-programmed) that explodes/implodes large lists will be vulnerable to this issue. |
|||
« Next Oldest | Next Newest »
|
User(s) browsing this thread: 1 Guest(s)