From: ptkwt@... (Phil Tomson)
Date: 2005-07-22T02:05:56+09:00
Subject: Re: C extension makes things slower

In article <42DF4F1D.2060607@capitain.de>,
Pit Capitain  <pit@capitain.de> wrote:
>Phil Tomson schrieb:
>> And the results:
>> l% ruby point.rb
>>   Total time for pure ruby point test: 8.600752
>>   ACO_Ext initializing...
>>   Total time for extension point test: 6.086567
>> 
>> Using the extension it's 2.6 second _slower_ than the pure Ruby 
>> implementation!?
>
>Phil, 6 seconds is _faster_ than 8.6 seconds...
>

<kidding> Yeah, but it's not as much faster as I wanted it to be 
</kidding>

Sheepish grin... Ah, right you are.  Things look rather backwards at 1AM. 

However, here's the result of another run:

  Total time for pure ruby point test: 8.051388
  ACO_Ext initializing...
  Total time for extension point test: 9.212984

There, I knew I wasn't hallucinating... 

Anyway, I'm starting to think that the variance might have more to do 
with loading of the machine I'm running on.  Should probably use 
benchmark instead.  Also, it was quite consistently slower (the C 
extension version) on my Powerbook, but then again I was running low on 
battery so that might have had something to do with it (still odd, 
though).

In reality, the code above is just a small part of a much bigger program 
which includes more methods defined in C.  When I ran that it was 
consistently slower by several seconds (actually almost twice as 
slow) when the extension was 'required'.  That's what led me to start to 
time different sections of the C extenstion seperately in order to find 
the culprit.

Phil