Ok, thanks for the hints. Actually pulling the image out of the card
probably won't be much of an issue.
I find displacement mapping works best with a grid of around 96*96,
and the texture having the same resolution. At higher resolution
there's not much difference in quality, and I'm also normal mapping so
it looks ok even lower. That shouldn't cause too much of an issue.
So, any of the dev team care to indicate if a fix is likely in the
next few months? If not I'll give this a try.
Chris
On 24 Apr 2008, at 17:59, Christopher Wright <[EMAIL PROTECTED]> wrote:
... Publishing a processed image output and publishing the image
input to the glsl shader, and removing the colorspace data in the
main
app seems to be the most promising avenue at the moment, but I've not
looked into it much so far.
To keep stuff running as smoothly as possible (I'm also of the
opinion that the only work around at this point is to export the
image to your app, do CS-tweaking there, and pass it back to the
composition, unless there's a bugfix in the pipe), you'll want to
keep the image on the GPU if possible (that's probably obvious :).
Initially, I was thinking that you could fetch the published image
as a CIImage, swizzle the colorspace, and pass it back, but that
doesn't appear possible (CIImage is way too opaque, and even the
undocumented methods don't obviously expose a way to accomplish
this). I've not worked extensively with CoreImage (mostly filter-
side stuff), so I may be missing a simple solution there.
As a secondary solution, you could try fetching as a CoreVideo
Buffer (CVImageBufferRef), and then create a CIImage from that
(initWithCVImageBuffer:options: with options set to produce a
working colorspace). I'm not sure of the details of this approach,
or whether it's even close to optimal, but it's something to kick
around I suppose. I've had good performance using CVBuffers for a
few small projects I've worked on.
For simplicity, you could also try proof-of-concept tests using
NSImage or something less accelerated. If the performance is decent
in such situations, going the extra mile for the added speed of an
all-GPU path may not be worth it for a non-widely distributed
application.
--
[ christopher wright ]
[EMAIL PROTECTED]
http://kineme.net/
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Quartzcomposer-dev mailing list ([email protected])
Help/Unsubscribe/Update your Subscription:
http://lists.apple.com/mailman/options/quartzcomposer-dev/archive%40mail-archive.com
This email sent to [EMAIL PROTECTED]
