Project

General

Profile

Actions

Feature #10909

closed

[PATCH 2/2] math.c: Direct casting from Rational to double.

Added by gogotanaka (Kazuki Tanaka) about 9 years ago. Updated about 9 years ago.

Status:
Closed
Assignee:
-
Target version:
-
[ruby-core:68325]

Description

Hi, this proposal is continued from Feature 10800 and Feature 10785
I'd like to do cast from Rational to double unless Rational#to_f is redefined.
Along with that, I've introduced some new macros(fix2dbl_without_to_f, big2dbl_without_to_f, ..) and added test cases.

Please feel free to let me know if there are something you don't like.

thanks.


Files

direct_casting_from_Rational_to_double.patch (1.45 KB) direct_casting_from_Rational_to_double.patch gogotanaka (Kazuki Tanaka), 02/26/2015 12:41 PM
add_tests_to_test_math_rb.patch (1.79 KB) add_tests_to_test_math_rb.patch gogotanaka (Kazuki Tanaka), 02/26/2015 12:41 PM

Updated by gogotanaka (Kazuki Tanaka) about 9 years ago

I can find this proposal makes Math faster more than 2 times at most.

$ cat bench.rb
require 'benchmark'

puts Benchmark.measure { 10_000_000.times { Math.cos 1             } }
puts Benchmark.measure { 10_000_000.times { Math.cos 1 << 100      } }
puts Benchmark.measure { 10_000_000.times { Math.cos 1.0           } }
puts Benchmark.measure { 10_000_000.times { Math.cos 1/3r          } }
puts Benchmark.measure { 10_000_000.times { Math.cos (1 << 100)/1r } }

Before

  1.320000   0.010000   1.330000 (  1.326548)
  3.060000   0.000000   3.060000 (  3.064614)
  1.380000   0.000000   1.380000 (  1.383077)
  5.360000   0.010000   5.370000 (  5.435371)
 19.220000   0.050000  19.270000 ( 19.565904)

After

  1.350000   0.000000   1.350000 (  1.357702)
  3.230000   0.010000   3.240000 (  3.247377)
  1.310000   0.000000   1.310000 (  1.315977)
  2.100000   0.000000   2.100000 (  2.106907)
 15.590000   0.040000  15.630000 ( 15.898346)
Actions #2

Updated by gogotanaka (Kazuki Tanaka) about 9 years ago

  • Status changed from Open to Closed
  • % Done changed from 0 to 100

Applied in changeset r49818.


  • math.c (num2dbl_with_to_f): direct casting from Rational to double.
    [Feature #10909]
Actions

Also available in: Atom PDF

Like0
Like0Like0