David,

We are getting closer. Looking at the following:

> ====Squeak 2.7-1782
> 0.44	 avg. mSecs/op -- Displaying span buffer
> 0.34	 avg. mSecs/op -- Displaying span buffer
> 0.06	 avg. mSecs/op -- Displaying span buffer

> ====Squeak 2.9a-2773
> 0.56	 avg. mSecs/op -- Displaying span buffer
> 0.48	 avg. mSecs/op -- Displaying span buffer
> 0.14	 avg. mSecs/op -- Displaying span buffer

It seems like we have some serious overhead in the display part. Okay, let's
try to find out if that happens in BitBlt or in the engine itself. Run the
following:

| bb span |
span _ Form extent: Display width @ 1 depth: 32.
span fillWhite.
bb _ BitBlt toForm: Display.
bb 
	sourceForm: span;
	sourceRect: span boundingBox;
	colorMap: (Color colorMapIfNeededFrom: 32 to: Display depth);
	combinationRule: 34.
[	Display deferUpdates: true.
	1 to: 100 do:[:j|
		1 to: Display height do:[:i|
			bb copyBitsFrom: 0 to: Display width at: i]].
	Display deferUpdates: false.
] timeToRun.

And let's see if and how the results differ. Oh, and by the way, two
questions: Are you running the same VM with the two images?! Also, what
processor are you running on?! (e.g., Intel, AMD, or whatever).

 - Andreas