Talk:Shader
From Wikipedia, the free encyclopedia
 Archives |
|---|
Contents |
[edit] Unified shader architecture
Does some expert on the subject want to add information on "unified shader architectures"? As far as I can see, they involve using the same pipeline and processing units to work on both shaders and vertices, but I'm sure there are more nuances that the article needs to address. GeorgeBills 03:31, 9 November 2006 (UTC)
The GLSL specification lets implementors choose between multi-processor and single do-it-all processors, as soon as the other requirements are met. The same applies to D3D10. This is implementation-dependant and should be noted on other pages (such as your favourite vid card).
Note that the whole "unified" thing is somewhat out of place since a fragment cannot be cut (geometry shader only operation on primitives), just as a vertex cannot be discarded once has been created by the GS (fragment only operation). Obviously, you can call the "kill-vertex" operation discard, just as the "kill-fragment" operation but this is just a superficial difference.
I'm reading the G80 specifications right now, but I hardly believe there will be the need for a lot of changes. Logically speaking, the processors are distinct. If they end up in the same silicon, it's not an issue for the programmer.
I'm rather busy recently so don't hold your breath.
MaxDZ8 talk 12:02, 9 November 2006 (UTC)
So, after a quick glance, here's what it does take to write something useful on the new G8x functionalities.
Up to yesterday, the GL spec was a few hundred pages. NVidia's extensions up to NV4x/G7x was about ~1500 pages. The new G8x specs alone count roughtly 500 pages (blowing the total page count to 2075) so it does take a while before they can be digested. There's everything you were expecting: geometry shaders, stream out, new buffer formats, gamma correction, uniform buffers and such.
Rectifying my previous statement, I expect it'll take a lot of time before a major update could be done. There's now enough material for a split.
MaxDZ8 talk 14:01, 9 November 2006 (UTC)
[edit] Damaged beyond repair?
Folks, this article is largely a write-off at this stage. :( Strongly suggest re-write it offline then delete the whole thing and replace the text.
I'm also convinced it's the best thing to do. It probably has more sense now to have a whole chapter on this, as well as a shading pipeline page.
MaxDZ8 talk 09:50, 11 November 2006 (UTC)
It starts off good but then starts rambling. It begins to fall apart right about at "By design, hardware shaders are ideal candidates for parallel execution..." Way to much high level speculative stuff beyond this point. Why is the bulk of this article about Real-time shader structure? Somebody got carried away on this and I agree that a re-write wouldn't be a bad idea. I have a background in 3D graphics (artistically) so perhaps I can help. But really the first bit is OK and the rest of it is just too much info.Lomacar 09:56, 4 December 2006 (UTC)
[edit] Hella Confusing, Yo
Ok people, i could tell just by skimming it that it would be too hard to understand. wtf? u ppl are supposed to help us understand what shaders are, not confuse us even more.
do vertex shaders calculate physics?KittenKiller
One thing you should do is look up terms you do not know. Basically, only people with a average level of understanding with computers will understand this. Tockwork 03:39, 22 November 2006 (UTC)
-
- Only people with a average level of understanding with computers? May be only people with some degree of understanding Computer Graphics. Computer Graphics is notoriously famous for its strange terms... As far as I know from my professor. Anyway, vertex shaders do not calculate physics, but they can be "set" to work according physics. Draconins 14:37, 4 December 2006 (UTC)
[edit] Readability of this article
The article as such appears quite readable to me. I speak out here in a capacity of a layman, so I hope my opinion will help to restrain professionals who contribute to the contents of the article from making it a "reading written by professionals for professionals". I mean that none should expect any kind of language specifications here, just a general discussion that is capable of letting curious people know what they pay their money for when they buy a new video adapter "with the support of such and such". I believe there's no sense to make up a voluminous treatise out of any article here at Wikipedia. Of course I realize it is hard sometimes to distinguish what is pertinent to the main line of discussion and what is not but I strongly encourage the author and the prospective contributors to try their best. It is a nice style to split the article when the discussion of some important but hardly indispensable notion is needed. I believe the issue with the pipeline to be just the case: one may easily grasp the notion of a shader without knowing exactly some side things. This is important because it makes an article much easier to read and understand.
