[TESTING] patched 1.9.3 performance improvements?

[Jon]

I've built at version of the unreleased 1.9.3 patched with this recent feature request

http://redmine.ruby-lang.org/issues/5427

and the 7z archive is available at

http://www.mediafire.com/?1a39sjnzd2dvs8b

I'm interested in hearing how it performs, qualitatively (good) and quantitatively (better), in your Windows environment (details summary please) especially with Rails, heavy BDD/TDD workloads, and processing a large number of input files.

Jon

---
blog: http://jonforums.github.com/
twitter: @jonforums

Most people die of a sort of creeping common sense, and discover when it
is too late that the only things one never regrets are one's mistakes.
- Oscar Wilde

[Dušan D. Majkić]

> I'm interested in hearing how it performs, qualitatively (good) and quantitatively (better)

> in your Windows environment (details summary please) especially with Rails,
> heavy BDD/TDD workloads, and processing a large number of input files.

Here is what I got for Cucumber test and SQL Server db backend.
To make things clear I downloaded your 1.9.3 7z and 1.8.7 7z from
RubyInstaler.org
Than issued these (for both clean setups)

gem install bundler
gem install json --platform=ruby
bundle install

Command line used:

timeit -f d:\t.dat bundle.bat exec cucumber features/simple.feature

Difference is huge in favor of your patched 1.9.3

14.4sec vs 21.33sec

What remains slow is that it takes more than 10sec even for 1.9.3 to start
executing.

Than I tried with 1.8.7 "workhorse install" with bunch of gems installed,
and results were in line with clean install of 1.8.7

Here is screen snapshot of 1.9.3 vs 1.8.7p334;

http://i.imgur.com/RVy1N.jpg

I wonder if there is a Rubinius windows build that can be compared.

[Jon]

Thanks for testing and providing results.

Unfortunately I haven't been able to build Rubinius due to the following and haven't had time to play with a fix. I'm curious too.

https://github.com/rubinius/rubinius/issues/1164