The system on which I performed the test is a laptop: Dell Inspiron 9400 with Centrino Duo, Intel T7200 4MB Cache 2GHz (166×12) 2GB Ram 667Mhz. This is a physical system Windows XP Pro SP3.
The purpose: to check the exploitation of more processor cores (two in my case) by comparing a single process with a double execution simultaneously.
To do this I used the string’s bench used in a previous article .
It must be remembered that the implementation of the dual process, everyone must share the CPU with the operating system (the load is always 100%) while in the single test that does not happen, a core is dedicated only to the test. For this reason, the operating system didn’t run antivirus or other heavy processes.
Let’s start with the oldest interpreter:
Ruby 1.8.6 patch 111
C:LavoroProgettiTestBench>ruby -v
ruby 1.8.6 (2007-09-24 patchlevel 111) [i386-mswin32]
C:LavoroProgettiTestBench>ruby bench_str.rb
user system total real
Concat 1.000.000:
+ 6.641000 0.000000 6.641000 ( 6.640000)
<< 5.344000 0.000000 5.344000 ( 5.375000)
#{} 6.078000 0.000000 6.078000 ( 6.078000)
Add 100.000:
+= 3.547000 2.969000 6.516000 ( 6.579000)
<< 0.062000 0.000000 0.062000 ( 0.062000)
a = a + '.' 3.000000 3.422000 6.422000 ( 6.547000)
#{} 4.531000 1.906000 6.437000 ( 6.516000)
Other 100.000:
* 100: 0.500000 0.063000 0.563000 ( 0.562000)
capitalize: 1.719000 0.078000 1.797000 ( 1.797000)
upcase: 4.906000 0.140000 5.046000 ( 5.047000)
chomp: 0.266000 0.063000 0.329000 ( 0.328000)
include: 1.234000 0.000000 1.234000 ( 1.234000)
index: 1.235000 0.000000 1.235000 ( 1.250000)
sub: 0.875000 0.094000 0.969000 ( 0.969000)
gsub: 17.453000 0.562000 18.015000 ( 18.047000)
[x..y]: 0.547000 0.016000 0.563000 ( 0.562000)
slice: 0.562000 0.000000 0.562000 ( 0.563000)
strip: 0.156000 0.000000 0.156000 ( 0.156000)
Each: 12.282000 0.078000 12.360000 ( 12.390000)
Cast 1.000.000:
.to_i: 0.437000 0.000000 0.437000 ( 0.438000)
.to_sym: 0.531000 0.000000 0.531000 ( 0.531000)
split: 9.047000 0.140000 9.187000 ( 9.235000)
--- Total: 80.953000 9.531000 90.484000 ( 90.906000)
Ruby 1.8.6 patch 111
Single thread:
--- Total: 80.953000 9.531000 90.484000 ( 90.906000)
Double simultaneous execution:
CORE1: --- Total: 87.063000 19.531000 106.594000 (107.923000) CORE2: --- Total: 91.344000 18.375000 109.719000 (110.125000)
This is the decrease of the double simultaneous execution:
Decrease by 10%
Real decrease of 20%
Ruby 1.8.6 patch 368
C:LavoroProgettiTestBench>ruby -v
ruby 1.8.6 (2009-03-31 patchlevel 368) [i386-mingw32]
C:LavoroProgettiTestBench>ruby bench_str.rb
user system total real
Concat 1.000.000:
+ 4.828000 0.000000 4.828000 ( 4.828125)
<< 3.938000 0.000000 3.938000 ( 3.953125)
#{} 4.719000 0.016000 4.735000 ( 4.750000)
Add 100.000:
+= 3.687000 2.719000 6.406000 ( 6.468750)
<< 0.047000 0.000000 0.047000 ( 0.046875)
a = a + '.' 3.422000 3.000000 6.422000 ( 6.468750)
#{} 4.797000 1.109000 5.906000 ( 5.906250)
Other 100.000:
* 100: 0.328000 0.094000 0.422000 ( 0.421875)
capitalize: 0.890000 0.078000 0.968000 ( 1.000000)
upcase: 3.469000 0.109000 3.578000 ( 3.578125)
chomp: 0.235000 0.078000 0.313000 ( 0.312500)
include: 0.796000 0.000000 0.796000 ( 0.796875)
index: 0.797000 0.000000 0.797000 ( 0.796875)
sub: 0.750000 0.172000 0.922000 ( 0.921875)
gsub: 20.860000 0.531000 21.391000 ( 21.421875)
[x..y]: 0.437000 0.016000 0.453000 ( 0.453125)
slice: 0.438000 0.000000 0.438000 ( 0.437500)
strip: 0.109000 0.000000 0.109000 ( 0.109375)
Each: 9.078000 0.172000 9.250000 ( 9.281250)
Cast 1.000.000:
.to_i: 0.297000 0.000000 0.297000 ( 0.296875)
.to_sym: 0.344000 0.000000 0.344000 ( 0.343750)
split: 7.375000 0.109000 7.484000 ( 7.500000)
--- Total: 71.641000 8.203000 79.844000 ( 80.093750)
Ruby 1.8.6 patch 368
Single thread:
--- Total: 71.641000 8.203000 79.844000 ( 80.093750)
Double simultaneous execution:
CORE1: --- Total: 80.797000 17.750000 98.547000 ( 99.093750) CORE2: --- Total: 76.500000 19.375000 95.875000 ( 97.187500)
Decrease of 9,8%
Real decrease of 22,5%
Ruby 1.9.1
C:LavoroProgettiTestBench>ruby -v
ruby 1.9.1p129 (2009-05-12 revision 23412) [i386-mingw32]
C:LavoroProgettiTestBench>ruby bench_str.rb
user system total real
Concat 1.000.000:
+ 2.610000 0.000000 2.610000 ( 2.609375)
<< 2.125000 0.000000 2.125000 ( 2.125000)
#{} 2.796000 0.000000 2.796000 ( 2.796875)
Add 100.000:
+= 3.688000 2.875000 6.563000 ( 6.609375)
<< 0.031000 0.000000 0.031000 ( 0.031250)
a = a + '.' 3.235000 3.265000 6.500000 ( 6.578125)
#{} 5.187000 2.266000 7.453000 ( 7.546875)
Other 100.000:
* 100: 0.203000 0.125000 0.328000 ( 0.328125)
capitalize: 5.563000 0.047000 5.610000 ( 5.656250)
upcase: 0.703000 0.031000 0.734000 ( 0.734375)
chomp: 0.219000 0.063000 0.282000 ( 0.296875)
include: 0.156000 0.000000 0.156000 ( 0.156250)
index: 0.156000 0.000000 0.156000 ( 0.156250)
sub: 0.500000 0.093000 0.593000 ( 0.593750)
gsub: 8.453000 0.266000 8.719000 ( 8.718750)
[x..y]: 0.219000 0.000000 0.219000 ( 0.218750)
slice: 0.219000 0.000000 0.219000 ( 0.218750)
strip: 0.109000 0.000000 0.109000 ( 0.109375)
Each: 7.188000 0.141000 7.329000 ( 7.328125)
Cast 1.000.000:
.to_i: 0.234000 0.000000 0.234000 ( 0.234375)
.to_sym: 0.391000 0.000000 0.391000 ( 0.390625)
split: 4.609000 0.000000 4.609000 ( 4.609375)
--- Total: 48.594000 9.172000 57.766000 ( 58.046875)
Ruby 1.9.1 patch 129
Single thread:
--- Total: 48.594000 9.172000 57.766000 ( 58.046875)
Double simultaneous execution:
CORE1: --- Total: 56.345000 19.219000 75.564000 ( 76.312500) CORE2: --- Total: 52.109000 20.703000 72.812000 ( 74.156250)
Decrease of 11,5%
Real decrease of 29,6%
jRuby 1.3.1
C:LavoroProgettiTestBench>jruby -v
jruby 1.3.1 (ruby 1.8.6p287) (2009-07-24 6586) (Java HotSpot(TM) Client VM 1.6.0
_15) [x86-java]
C:LavoroProgettiTestBench>jruby bench_str.rb
user system total real
Concat 1.000.000:
+ 1.703000 0.000000 1.703000 ( 1.672000)
<< 1.375000 0.000000 1.375000 ( 1.375000)
#{} 0.625000 0.000000 0.625000 ( 0.625000)
Add 100.000:
+= 8.343000 0.000000 8.343000 ( 8.343000)
<< 0.031000 0.000000 0.031000 ( 0.031000)
a = a + '.' 8.406000 0.000000 8.406000 ( 8.406000)
#{} 18.984000 0.000000 18.984000 ( 18.984000)
Other 100.000:
* 100: 0.157000 0.000000 0.157000 ( 0.157000)
capitalize: 0.937000 0.000000 0.937000 ( 0.937000)
upcase: 1.703000 0.000000 1.703000 ( 1.703000)
chomp: 0.063000 0.000000 0.063000 ( 0.063000)
include: 0.281000 0.000000 0.281000 ( 0.281000)
index: 0.313000 0.000000 0.313000 ( 0.313000)
sub: 0.437000 0.000000 0.437000 ( 0.437000)
gsub: 4.813000 0.000000 4.813000 ( 4.813000)
[x..y]: 0.156000 0.000000 0.156000 ( 0.156000)
slice: 0.140000 0.000000 0.140000 ( 0.140000)
strip: 0.047000 0.000000 0.047000 ( 0.047000)
Each: 3.094000 0.000000 3.094000 ( 3.094000)
Cast 1.000.000:
.to_i: 0.609000 0.000000 0.609000 ( 0.609000)
.to_sym: 0.266000 0.000000 0.266000 ( 0.266000)
split: 2.203000 0.000000 2.203000 ( 2.203000)
--- Total: 54.686000 0.000000 54.686000 ( 54.655000)
jRuby 1.3.1 => 1.8.6 p287
Single thread:
--- Total: 54.686000 0.000000 54.686000 ( 54.655000)
Double simultaneous execution:
CORE1: --- Total: 65.170000 0.000000 65.170000 ( 64.920000) CORE2: --- Total: 64.234000 0.000000 64.234000 ( 64.000000)
Decrease of 17,2%
Real decrease of 17,9%
IronRuby 0.9.0
C:LavoroProgettiTestBench>ir -v
IronRuby 0.9.0.0 on .NET 2.0.0.0
C:LavoroProgettiTestBench>ir bench_str.rb
user system total real
Concat 1.000.000:
+ 3.562500 0.000000 3.562500 ( 3.234375)
<< 1.531250 0.000000 1.531250 ( 1.531250)
#{} 1.453125 0.015625 1.468750 ( 1.453125)
Add 100.000:
+= 54.656250 11.296875 65.953125 ( 66.171875)
<< 0.031250 0.000000 0.031250 ( 0.031250)
a = a + '.' 54.734375 11.718750 66.453125 ( 66.671875)
#{} 55.609375 10.843750 66.453125 ( 66.609375)
Other 100.000:
* 100: 0.671875 0.000000 0.671875 ( 0.671875)
capitalize: 2.593750 0.000000 2.593750 ( 2.609375)
upcase: 7.484375 0.000000 7.484375 ( 7.500000)
chomp: 0.250000 0.000000 0.250000 ( 0.250000)
include: 5.968750 0.015625 5.984375 ( 6.031250)
index: 5.968750 0.031250 6.000000 ( 6.031250)
sub: 1.781250 0.015625 1.796875 ( 1.781250)
gsub: 12.203125 0.031250 12.234375 ( 12.281250)
[x..y]: 0.406250 0.000000 0.406250 ( 0.390625)
slice: 0.343750 0.000000 0.343750 ( 0.343750)
strip: 0.078125 0.000000 0.078125 ( 0.078125)
Each: 31.546875 0.312500 31.859375 ( 32.203125)
Cast 1.000.000:
.to_i: 0.265625 0.000000 0.265625 ( 0.250000)
.to_sym: 0.406250 0.015625 0.421875 ( 0.421875)
split: 5.515625 0.031250 5.546875 ( 5.531250)
--- Total: 247.062500 34.328125 281.390625 (282.078125)
C:LavoroProgettiTestBench>
IronRuby 0.9.0
Single thread:
--- Total: 247.062500 34.328125 281.390625 (282.078125)
Double simultaneous execution:
CORE1: --- Total: 315.531250 40.578125 356.109375 (359.500000) CORE2: --- Total: 300.484375 46.156250 346.640625 (350.265625)
Decrease of 24,5%
Real decrease of 25,8%
Summary
Ruby 1.8.6 patch 111
Decrease of 10%
Real decrease of 20%
Ruby 1.8.6 patch 368
Decrease of 9,8%
Real decrease of 22,5%
Ruby 1.9.1
Decrease of 11,5%
Real decrease of 29,6%
jRuby 1.3.1
Decrease of 17,2%
Real decrease of 17,9%
IronRuby 0.9.0
Decrease of 24,5%
Real decrease of 25,8%
Conclusions
In this test, JRuby is the only one that has a uniform load on both cores and it does almost constantly for all the time. Only at the end moves towards a core but with more discretion. Also good performance with the two processes simultaneously.
Versions 1.8.6 and 1.9.1 are quite aligned and able to share the load solely on test "Add". The performance of the dual process would not be bad but creates problems for the system (at least mine) which prolongs the time.
IronRuby has some trouble with the strings already highlighted in a previous article. Ignoring this issue, the second core was left unloaded for all the duration. With the second core free, I would have expected a less pronounced decrease in the test process twice but it did not happen, this is strange.
In a future article, I'd like to repeat the test on a server quad core.
English
Italiano