I've not done a test, but it showed up in the diagnostics. The issue is  
in the
copyLoopNoSource
we have a for loop
2 to: nWords - 1 do:
which translates into a for loop that doesn't get for loop optimized by  
the compiler (gcc) because
of the nWords-1 condition. Translating that to to nWordsMinusOne _  
nWords - 1 & 2 to: nWordsMinusOne do: does
get optimized. Same code exists in copyloop which was altered by the  
original FasterCopyLoop changeset, so I'm really carrying forward that  
optimization into copyLoopNoSource.

PS Can't say if I've heard if the

[ENH] Faster CopyLoop
	FasterCopyLoop-JMM

  [ENH][VM] making FreeCell redraw 20% faster...
	FasterAlphaBlend-JMM

make any difference on wintel VMs?


On Tuesday, July 29, 2003, at 03:05  PM, Andreas Raab wrote:

> So what does BitBlt>>benchmark say? Anything noticable?
>
> Cheers,
>   - Andreas
>
>> -----Original Message-----
>> From: squeak-dev-bounces at lists.squeakfoundation.org
>> [mailto:squeak-dev-bounces at lists.squeakfoundation.org] On
>> Behalf Of John M McIntosh
>> Sent: Tuesday, July 29, 2003 11:39 PM
>> To: squeak-dev at lists.squeakfoundation.org
>> Subject: [ENH][VM] - faster copy loop part 2
>>
>>
>> localize for loop condition in copy loop for better performance
>>
>> --
>> ==============================================================
>> ==========
>> ===
>> John M. McIntosh <johnmci at smalltalkconsulting.com> 1-800-477-2659
>> Corporate Smalltalk Consulting Ltd.
>> http://www.smalltalkconsulting.com
>>
>> ==============================================================
>> ==========
>> ===
>>
>
>
>
>
--
======================================================================== 
===
John M. McIntosh <johnmci at smalltalkconsulting.com> 1-800-477-2659
Corporate Smalltalk Consulting Ltd.  http://www.smalltalkconsulting.com
======================================================================== 
===