Comments on: Half a half: Part 2 http://www.parolski.com/2008/12/13/half-a-half-part-2/ Faith, Solaris, and Chicken Korma, by Anton Parol Tue, 11 May 2010 09:02:32 +0000 hourly 1 http://wordpress.org/?v=3.0 By: Liam McBrien http://www.parolski.com/2008/12/13/half-a-half-part-2/comment-page-1/#comment-454 Liam McBrien Wed, 17 Dec 2008 18:09:32 +0000 http://www.parolski.com/?p=158#comment-454 OK, so here's how those two runs stack up against each other: With the I/O, the tests take 1.019s and 322.714s, meaning that the loop version is roughly 317 times slower (322.714 / 1.019). Without the I/O, the tests take 0.292s and 21.83s, meaning that the loop version is roughly 75 times slower (21.83/0.292). So, by removing the console output from the loop program, we get over four times better performance relative to the non-loop version (317/75). This means that roughly 80% of the loop version's execution time must be the I/O - just goes to show it's always worth looking at where the bottlenecks in a system are! OK, so here’s how those two runs stack up against each other:

With the I/O, the tests take 1.019s and 322.714s, meaning that the loop version is roughly 317 times slower (322.714 / 1.019).

Without the I/O, the tests take 0.292s and 21.83s, meaning that the loop version is roughly 75 times slower (21.83/0.292).

So, by removing the console output from the loop program, we get over four times better performance relative to the non-loop version (317/75). This means that roughly 80% of the loop version’s execution time must be the I/O – just goes to show it’s always worth looking at where the bottlenecks in a system are!

]]> By: Anton Parol http://www.parolski.com/2008/12/13/half-a-half-part-2/comment-page-1/#comment-453 Anton Parol Sun, 14 Dec 2008 18:55:30 +0000 http://www.parolski.com/?p=158#comment-453 Both these sets of results are with the I/O calls removed. The first results are non loop, the second are with the loop: anton@anton-laptop classes $ time java The_proper_half 4000 real 0m0.292s user 0m0.384s sys 0m0.028s anton@anton-laptop classes $ time java Half_of_a_half 4000 real 0m21.830s user 0m22.169s sys 0m0.108s Both these sets of results are with the I/O calls removed. The first results are non loop, the second are with the loop:

anton@anton-laptop classes $ time java The_proper_half 4000

real 0m0.292s
user 0m0.384s
sys 0m0.028s
anton@anton-laptop classes $ time java Half_of_a_half 4000

real 0m21.830s
user 0m22.169s
sys 0m0.108s

]]> By: Liam McBrien http://www.parolski.com/2008/12/13/half-a-half-part-2/comment-page-1/#comment-452 Liam McBrien Sun, 14 Dec 2008 13:07:25 +0000 http://www.parolski.com/?p=158#comment-452 I'd be interested to see how the loop version runs without the System.out calls. I'll bet it makes a significant difference! I’d be interested to see how the loop version runs without the System.out calls. I’ll bet it makes a significant difference!

]]>