Hi,

Reading the TN it seems as if it may be worthwhile to experiment with
different strategies in DisplayScreen's forceDamageToScreen: method. E.g.,
try if simply merging the individual rectangles into a larger one helps to
improve performance, simply replacing

	rectList do:[:r| self forceToScreen: r].

with

	self forceToScreen: (Rectangle merging: rectList).

might be a worthwhile attempt. If that does the trick I think we will want
to have a primitiveFlush which takes the composite region so that VMs can
handle it if they want to (you definitely _don't_ want to do this if there's
any munging of byte order or color space involved so it would have to be
optional).

Cheers,
  - Andreas


> -----Original Message-----
> From: squeak-dev-bounces at lists.squeakfoundation.org 
> [mailto:squeak-dev-bounces at lists.squeakfoundation.org] On 
> Behalf Of Karsten Wolf
> Sent: Tuesday, August 05, 2003 9:48 PM
> To: The general-purpose Squeak developers list
> Subject: Re: [RFI]Mac performance gap between OS-9 and OSX
> 
> 
> Hello John
> 
> Thanks for the elaborate answer. I have compiled a vm with an 
> outcommented QDFlushPortBuffer() and found out:
> 1. this is where the time goes (now it's as fast as OS9)
> 2. like the (I assume it was yours) comment said: video becomes jerky 
> (at best).
> 
> A quick search at apple lead me to  TN2051. There's a code fragment 
> that could lead in the right direction.
> 
> In a first test (FreeCell 1) video becomes animated again but window 
> movements still leave artifacts.
> 
> karsten
> 
>