Hi there,
I propose to replace the use of the method Range#include? to normal compared with the beginning and end of the range. For example consider the actually used method:
(11…13).include?(abs_number % 100)
and if we replace him to:
modulo_number = abs_number % 100
modulo_number >= 0) && (modulo_number <= 1)
it increase performance about 1.5 times.
This approach can be used in another real existing place.
Below are the benchmarks data:
Benchmark.ips do |x|
x.report(‘old_slow’) do
abs_number = 1234
(11..13).include?(abs_number % 100)
end
x.report(‘new_fast’) do
abs_number = 1234
modulo_number = abs_number % 100
(modulo_number >= 0) && (modulo_number <= 1)
end
x.compare!
end
Warming up --------------------------------------
old_slow 257.215k i/100ms
new_fast 334.191k i/100ms
Calculating -------------------------------------
old_slow 5.628M (± 0.9%) i/s - 28.294M in 5.027572s
new_fast 8.791M (± 0.9%) i/s - 44.113M in 5.018600s
Comparison:
new_fast: 8790704.5 i/s
old_slow: 5628188.6 i/s - 1.56x slower
Benchmark.ips do |x|
x.report(‘old_slow’) do
x = 0
(0…1).include?(0)
end
x.report(‘new_fast’) do
x = 0
(x >= 0) && (x <= 1)
end
x.compare!
end
Warming up --------------------------------------
old_slow 283.858k i/100ms
new_fast 367.498k i/100ms
Calculating -------------------------------------
old_slow 6.133M (± 1.3%) i/s - 30.657M in 4.999708s
new_fast 10.486M (± 1.6%) i/s - 52.552M in 5.012845s
Comparison:
new_fast: 10486347.5 i/s
old_slow: 6132780.2 i/s - 1.71x slower
What do you think about it?
