Content-aware fill and layers

Question

One thing I'd been wondering about very eagerly is whether the content-aware fill feature(s) would source their content from multiple layers, or only from the current layer, or whether there might be an option, as for the clone stamp tool for example, to choose whether the current layer, the current layer and below, or all layers were sampled for content.
I hadn't noticed any demonstration of content-aware fill on a document having multiple pixel layers.
Here are the results from my testing. (They may be wrong!)
What I want to be able to do, is to put a single "corrections" layer at the top of a layer stack, and to make local corrections using content from all layers below—with the latter being rasterized on the fly as necessary, so that the cumulative effect of masks, layer effects, blending modes, transparency, knockout, adjustment layers and anything else, are all used as content for the corrections. Until now I've done something similar but more limited with the clone stamp tool. Content awareness seemed to offer greater promise, for three reasons. (1) No back and forth: no choosing and monitoring of source pixels. (2) No visible softening of the correction at the edge of the brush. (3) The promise of imperceptible results with very little thought or work.
So, can we do this?
For fill, content-awareness only extends to pixels in the current layer. It doesn't look below. If the current layer is empty, then Photoshop comes back with a message that filling is not possible in content-aware mode, because not enough content can be found. If the current layer is blank (e.g. all white), then you just get more blank. That's fair enough. You could use content-aware fill early, on source layers, or late and destructively, after flattening. But not what I'm looking for.
The same is not true of the spot healing brush, which has both a "content aware" option and a "sample all layers option". And it works wonders. It does exactly what I want. Empty layer, sample all layers, run the the brush over dodgy parts of the image, and hey presto—it really does feel as if you have a wand in your hand.
This feature alone would be enough to persuade me to upgrade to CS5, since it will save me hours of work and produce a better result (in my context, cartography). Well done Adobe.

Answer 1

You're asking about a pretty rarified realm.  Not a whole lot of people ever have 24GB let alone beyond.
I stitch big panos all the time, though 1 GP is a fair bit larger than most I do; mine are in the 100-300 MP range.  I should mention I almost exclusively edit in 16 bits/channel mode.  I also use content aware fill fairly regularly without problems.  In my case I have 8 GB of RAM in a Dell Precision workstation.  It doesn't seem to get too winded with 8 GB, so 24 GB seems about right for what you want.
You definitely DO want a workstation that's been purpose-built for extra throughput and reliability if you're going to be working on ultra-large datasets.
I'd say buy the workstation with 24 GB of RAM and see how it goes.  Also the RAID is a very good idea (I use RAID 0).  Apparently the 6 Gb/sec SATA stuff is starting to become real too, though most drives can't even keep up with the old 3 Gb speed yet.
-Noel