[ruby-core:44169] [ruby-trunk - Feature #6256] Slightly improve ruby_qsort performance

Issue #6256 has been updated by MartinBosslet (Martin Bosslet).

File rubyqisort.patch added

I added a patch, together with benchmarks. 

Here's the results on my machine without the optimization:

$ ruby driver.rb -e ruby -p bm_sort -r 10
-----------------------------------------------------------
benchmark results:
name	ruby 2.0.0dev (2012-04-05 trunk 35241) [x86_64-linux]
sort_int	0.378
sort_int_custom	0.327
sort_int_rev	0.111
sort_int_sorted	0.100
sort_str_rnd	0.107

and here with the patch applied:

$ ruby driver.rb -e ruby -p bm_sort -r 10
-----------------------------------------------------------
benchmark results:
name	ruby 2.0.0dev (2012-04-05 trunk 35241) [x86_64-linux]
sort_int	0.367
sort_int_custom	0.332
sort_int_rev	0.229
sort_int_sorted	0.218
sort_str_rnd	0.106

From running them several times, my impression is that it's definetely
faster for sorting Fixnums, but there's no real improvement as soon
as we compare Strings or use a custom comparison, as I assumed before,
the costlier the comparison becomes the less advantage we get, at least
that's how it seemed in my tests.

Another fact is that we definitely lose in the "already ordered" cases.
The existing Quicksort does a linear check for "ordered regions" already,
to prevent worst case behavior. I actually had to take those out in
order to get the improvement with Insertion Sort. If I put them back in,
then the whole advantage is gone, and the "optimized" version runs 
slightly slower than the old version.

Applying just one final Insertion Sort instead of sorting each sub
problem below the cut off with it also turned out to be slightly
slower unlike in my test extension.

I'm confused to see only minimal improvements, when implementing a
"textbook" Quicksort with the same optimization, the benefits are much
more obvious. 

I assume there's more optimization in Ruby's current algorithm 
that would no longer be needed with additional Insertion Sort, 
maybe you have some ideas? Or maybe there's also room to 
improve my Insertion Sort?

This is as fast as I could get it right now, at least in theory it
should be possible to improve it.

----------------------------------------
Feature #6256: Slightly improve ruby_qsort performance
https://bugs.ruby-lang.org/issues/6256#change-25691

Author: MartinBosslet (Martin Bosslet)
Status: Assigned
Priority: Low
Assignee: MartinBosslet (Martin Bosslet)
Category: core
Target version: 2.0.0


Hi all,

I think I may have found a way to slightly improve the performance of ruby_qsort. 
Quicksort running time is slightly decreased if the recursion depth (or in our 
case, rather iteration depth) is bounded by leaving sub problems larger than or 
equal to some cutoff bound k untouched and running Insertion Sort on these small 
sub problems to finalize the sorting. 

I experimented with this, but to no avail, only marginal improvements if any. Then 
I remembered that instead of running Insertion Sort on each sub problem, it is 
equivalent in terms of running time to run one single Insertion Sort on the whole 
nearly sorted array as a final step. And in practice, this turns out to run faster 
than without the optimization. In my tests, execution time dropped to about 95% on 
average with an optimal cutoff (64-bit Fedora 15) [1].

Now this ain't the world - but it is faster, and I could very well imagine that there 
is still room for improving my code. In my tests, the optimal cutoff seems to be ~13 
for Integers and ~8 for Strings and Symbols. I imagine the more costly the comparisons, 
the lower will be the optimal cutoff. I've tested only on 64 Bit yet, but I will do so 
for 32 Bit, too.

If it turns out that this runs faster regardless of the architecture in use, with an
optimal cutoff yet to be determined, do you think this would be a useful addition?

I have attached a C extension for testing and discussing, it's mostly a one-to-one copy of
the code in util.c. I just added mmassign and insertion_sort plus the few lines that respect
the cutoff in rqsort (had to rename it, otherwise it would collide with the real "ruby_qsort").

-Martin

[1] https://github.com/emboss/hybridsort/blob/master/hybrid-sort-results


-- 
http://bugs.ruby-lang.org/