Article on GPU acceleration in PP at Toms Hardware
They may know that DXO supports OpenCL. When DXO released that version the amount of "it doesn't work/isn't enabled/crashes/I have high end hardware how dare it not go faster/its not faster/its slower/its the best thing/it says it works but doesn't" thrashing around was quite something to watch. If I was Adobe I might be a little more cautious than DXO.
GPU-coprocessing is still uncharted territory. It requires major rewriting using cryptic and archaic C-style APIs or dialects (until C++AMP is finalized), and the benefits are not always as obvious as it seems, for example latency issues makes it very expensive to ship data to and from the accelerator and can very well offset the gain provided by massively parallel processing on the GPU vs. multi-core processing on the CPU (most benchmarks and samples I've seen compare single-threaded vs. GPU which is unfair to begin with). GPU computing really is something that must be approached on a case-by-case basis - routine by routine - with careful cost/benefit analysis and Adobe is apparently doing just that.
|Post (hide subjects)||Posted by||When|
|Jun 13, 2012|
|Jun 13, 2012|
|Jun 15, 2012|
|Jun 16, 2012|
|Leafless Tree. by SpartanWarrior|
|50 Shades On The Paris Metro by wam7|
from Your City - Commute
|Montana Badlands by stickpointed|
from The Tree
|Kestrel by Lance B|
from My Best Photo of the Week