Romano:
> I suspect (i'm sure :-) you use a not-beta Rebol release. I have similar
>  results under 1.2.3.1.3.
>  Automatic Rebol memory allocation until 1.2.3 is buggy and to limit 
problems,
> 
>  blocks need to be preallocated with make block!.

Quite right -- I haven't been using the beta. That's because the main 
application I support crashes under it. That's crashes as in....

          REBOL caused an invalid page fault in
          module REBOL.EXE at 0167:00429bb9.

....and bye, bye console. Feedback know. I'll try again in the next beta.

But I have just run my collection of benchmarks under the beta and it shows 
(in most cases) the beta is substantially faster.

I have all five "Loius sort" solutions in one file, and I ran the tests as 

   loop 5 [do %sort-test.r]

This doesn't just test the different solutions in isolation. It also sees how 
well they run in an increasing aged (and possibly memory-fragmented) console.

1.2.1.3.1 -- Current release
1.2.5.3.1 -- Latest beta

  6.82 -- Romano's sort one big block -- Current release
  7.16
  6.81
  6.92
  7.09

  0.72 -- Romano's sort one big block -- Beta
  0.88
  0.94
  0.89
  0.94

 59.04  -- Sunanda's parse in the Sort -- Current release
 96.94    
 99.19    
100.29    
103.31   

 63.07  -- Sunanda's parse in the Sort -- beta
103.43    
107.16    
106.94    
111.56   

  9.34  --  Gregg's parse before the sort -- Current release
  9.78      
  9.07      
  9.73
  9.28      


  1.65  --  Gregg's parse before the sort -- Beta
  2.03      
  2.09      
  2.03
  2.03      

  2.41 -- Scott's index sort -- Current release
  2.64 
  2.58
  2.53
  2.58  


  1.10 -- Scott's index sort -- Beta
  1.21 
  1.20
  1.15
  1.21  

  2.36  -- Joel's bridge sort -- Current release
  2.31
  2.52
  2.36
  2.53

  0.60  -- Joel's bridge sort -- Beta
  0.60
  0.60
  0.66
  0.60

As you can see, we are all faster by a factor of several, except me - I just 
keep getting slower and slower.

As you suggest, Romano, there seems to be less evidence for a systematic slow 
down in the beta. Although in all cases (except Joel) out second time round 
was much slower.

Sunanda.
-- 
To unsubscribe from this list, please send an email to
[EMAIL PROTECTED] with "unsubscribe" in the 
subject, without the quotes.

Reply via email to