[Issue 6192] std.algorithm.sort performance
d-bugmail at puremagic.com
d-bugmail at puremagic.com
Mon Jul 11 05:03:19 PDT 2011
http://d.puremagic.com/issues/show_bug.cgi?id=6192
--- Comment #4 from bearophile_hugs at eml.cc 2011-07-11 04:58:06 PDT ---
Thank you for your work. The new timings (after pull 74):
sort-sort2 benchmarks (milliseconds), N=6000000:
Random distribution:
sort: 1261
sort2: 1201
Already sorted arrays:
sort: 300
sort2: 441
Inverted order arrays:
sort: 315
sort2: 729
Few random doubles appended to the sorted arrays:
sort: 9962
sort2: 632
The last case is a common use case in Python code. Python programmers sometimes
append unsorted items to a sorted list, and once in a while they sort it.
Because the Timsort used in Python (and Java) is able to face this case very
well, this is an efficient strategy to keep a rarely updated list sorted in
Python.
I don't know how much common this pattern is in real D code (but experience
shows that often real-world data is already partially sorted).
--
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
More information about the Digitalmars-d-bugs
mailing list