Block slowdown?

rogerdpack · March 15, 2010, 3:56pm

I noticed that methods that call yield and a few other “micro
benchmarks” are slower in jruby than MRI [at least on doze].
Should I report this as a JIRA?

gist.github.com

https://gist.github.com/rdp/332879

block speed tests.rb

require 'benchmark'

def yields
 yield
end

def named_block &block
 block.call
end

This file has been truncated. show original

gistfile2.txt

jruby --server

straight yield   0.047000   0.000000   0.047000 (  0.047000) * slower
named block      0.094000   0.000000   0.094000 (  0.094000)
cached proc       0.093000   0.000000   0.093000 (  0.093000) * slower

jruby --server --fast

straight yield   0.031000   0.000000   0.031000 (  0.031000)
named block      0.094000   0.000000   0.094000 (  0.094000)

This file has been truncated. show original

Thanks.
-rp

rogerdpack · March 15, 2010, 3:58pm

I noticed that methods that call yield and a few other “micro
benchmarks” are slower in jruby than MRI [at least on doze].
Should I report this as a JIRA?

Looks like they’re about the same speed if I use --fast
So I guess not bug worthy

rogerdpack · March 15, 2010, 4:58pm

On Mon, Mar 15, 2010 at 8:58 AM, Roger P. [email protected]
wrote:

I noticed that methods that call yield and a few other “micro
benchmarks” are slower in jruby than MRI [at least on doze].
Should I report this as a JIRA?

Looks like they’re about the same speed if I use --fast
So I guess not bug worthy

Our block performance has been adequate but about the same for over a
year. We know things we could do to improve the performance, but it’s
been solid enough we have not prioritized it above other things (like
bug fixing and new features).

One of these days we’ll circle back around to performance stuff…

Charlie

To unsubscribe from this list, please visit:

http://xircles.codehaus.org/manage_email