From: os97673@...
Date: 2014-12-04T09:35:56+00:00
Subject: [ruby-core:66680] [ruby-trunk - Bug #10511] fix for #9940 causes dramatic performance regression

Issue #10511 has been updated by Oleg Sukhodolsky.


Eric Wong wrote:
> Oops, [Feature #9425] (power-of-two hash sizes :)

If the problem is in implementation of hash function then I'd expect that simple Thread.current.hash instead of theHash[Thread.current] would show a performance degradation, but (as far as I can see) it is not.
If the cause of the problem is in Hash implementation than I'd expect that theHash[1] would show performance degradation too, but it is not.
IMHO it is something more subtle and it looks like it is somehow relates to #9940.

----------------------------------------
Bug #10511: fix for #9940 causes dramatic performance regression
https://bugs.ruby-lang.org/issues/10511#change-50285

* Author: Oleg Sukhodolsky
* Status: Feedback
* Priority: Normal
* Assignee: Koichi Sasada
* Category: core
* Target version: current: 2.2.0
* ruby -v: ruby 2.1.3p242 (2014-09-19 revision 47630) [x86_64-darwin14.0]
* Backport: 2.0.0: UNKNOWN, 2.1: UNKNOWN
----------------------------------------
Originally reported against debase gem (https://github.com/denofevil/debase/issues/16)
After fix for #9940 running anything under debugger became about 4 times slower.
I've verified that we I revert the fix performance is restored.
To reproduce you could get https://github.com/railstutorial/sample_app_rails_4 and run rspec rake task under debugger (I've used RubyMine for this because debase is used as 
debugger's backend there)
Performance with 2.1.2 it takes about 20 seconds on my laptop, with 2.1.3 - almost 80 seconds :(
The same problem is observable with 2.1.4 and 2.1.5 too.




-- 
https://bugs.ruby-lang.org/