That’s cool. I see you did it on Java 7. I wondered if for such a trivial
example if things change in Java 8 or 9. I’ve heard that JVM can do smart
things with stack allocation to eliminate GC and inlining. On your system you
saw almost 4x increase. On mine I see almost 6x. However, I don’t know how well
gbench does like JMH in terms of dead-code analysis – and this loop would
definitely count as dead code. I also tried an example to eliminate the dead
code.
Environment
===========
* Groovy: 2.4.6
* JVM: Java HotSpot(TM) 64-Bit Server VM (25.74-b02, Oracle Corporation)
* JRE: 1.8.0_74
* Total Memory: 214.5 MB
* Maximum Memory: 1794 MB
* OS: Windows 7 (6.1, amd64)
Options
=======
* Warm Up: Auto (- 60 sec)
* CPU Time Measurement: On
user system cpu real
Each 6271193022 25111 6271218133 6292037678
For 1107577805 20925 1107598730 1108344167
@CompileStatic:
Each 1013972638 22068 1013994706 1033565157
For 249595132 4566 249599698 251259849
If I change the code to do something with the result:
@groovy.transform.CompileStatic
def doEach() {{ it->
int i = 0
(1..100000000).each { int x -> i *= x }
println i
}}
Now the results are this:
user system cpu real
Each 7768820885 19783 7768840668 7832590835
For 834584044 11414 834595458 859830590
Jason
From: Bob Brown [mailto:[email protected]]
Sent: Wednesday, June 29, 2016 6:28 PM
To: [email protected]
Subject: Re: For-loop vs each.
I’ve never benchmarked it,
Weirdly enough, I have!
I wrote a quick bench benchmark for just this:
http://wordpress.transentia.com.au/wordpress/2013/03/25/gorgeous-gbench/
HTH
BOB
----------------------------------------------------------------------
This email message and any attachments are for the sole use of the intended
recipient(s). Any unauthorized review, use, disclosure or distribution is
prohibited. If you are not the intended recipient, please contact the sender by
reply email and destroy all copies of the original message and any attachments.
