From: Roger Pack <rogerdpack2@...>
Date: 2010-03-24T02:36:50+09:00
Subject: [ruby-core:28890] Re: Indentifying key MRI-on-Windows issues

> Thanks to nobu and Roger for the patches for 2742. On Vista, they were able to take my micro-benchmark's "real" results from 525.16 down to 18.76 seconds on a ~48MB test file.
>
> However, it doesn't appear root cause has been identified and there is significant room for improvement as JRuby, IronRuby, and MRI 1.8.x ran in the range of 0.14 to 0.65 seconds as the following results indicate.

If you want to examine a little more closely the "real root cause" you
can do so by downloading "process explorer", run it, double click on
ruby.exe, choose the "Threads" tab then double click on the thread
number that seems the most active.  It will show you the method it's
currently in (and chances are, the one it is bottlenecking on).

If you want to recreate 1.8.6 "being slow" you can try this ticket:

http://redmine.ruby-lang.org/issues/show/2741

> The CRLF read results: http://gist.github.com/341365

Good to see it's improved.

Further bottlenecks are probably the ones I mentioned a few posts
earlier.  Patches are welcome :)


> Anyone done any MRI 1.9.2dev performance testing on Windows with "real world" (or micro-benchmark) heavy IO read workloads? �Anyone done any Rails3 testing?

I haven't, as I would be afraid to use windows MRI for real workloads
unless the I/O stuff improves (jruby might be ok, though).  It works
for development, though, don't get me wrong, I use it all the time.  I
even get it to load quickly by reducing the number of File stat's by
using the "faster_require" gem :)

-rp