why new a hash which has seven objects is much slower than six length hash?

Question

I found when I new a Hash which has seven objects is much slower than six length hash. I know that the length of hash will affect the performance. But I don't know why seven is a special one.

Here is the code of benchmark(Ruby 2.2.3):

require 'benchmark/ips'

Benchmark.ips do |x|
  x.report(5) { { a: 0, b: 1, c: 2, d: 3, e: 4 } }
  x.report(6) { { a: 0, b: 1, c: 2, d: 3, e: 4, f: 5 } }
  x.report(7) { { a: 0, b: 1, c: 2, d: 3, e: 4, f: 5, g: 6 } }
  x.report(8) { { a: 0, b: 1, c: 2, d: 3, e: 4, f: 5, g: 6, h: 7 } }
  x.report(9) { { a: 0, b: 1, c: 2, d: 3, e: 4, f: 5, g: 6, h: 7, i: 8 } }

  x.compare!
end

And the blow is the results:

Calculating -------------------------------------
                   5    65.986k i/100ms
                   6    63.966k i/100ms
                   7    30.713k i/100ms
                   8    28.991k i/100ms
                   9    27.115k i/100ms
-------------------------------------------------
                   5      1.243M (± 4.3%) i/s -      6.203M
                   6      1.202M (± 5.3%) i/s -      6.013M
                   7    373.366k (±13.7%) i/s -      1.843M
                   8    351.945k (± 8.8%) i/s -      1.768M
                   9    331.398k (± 8.2%) i/s -      1.654M

Comparison:
                   5:  1243005.5 i/s
                   6:  1202032.4 i/s - 1.03x slower
                   7:   373366.5 i/s - 3.33x slower
                   8:   351945.1 i/s - 3.53x slower
                   9:   331398.3 i/s - 3.75x slower

Answer 1

From Hash lookup in Ruby, why is it so fast?:

Ruby manages the size of the bins dynamically. It starts with 11 and as soon as one of the bins has 5 or more elements, the bin size is increased and all hash elements are reallocated to their new corresponding bin.

At some point you pay an exponentially increased time penalty while Ruby resizes the bin pool, but if you think about it, its worth the time since this will keep lookup times and memory usage as low as possible.

This mean that the more bins, the less time spent looking for a specific key in a bin.

Hope it helps to understand the behaviour.