Recent ATI Patents

rwolf

Rock Star
Regular
Recent ATI Patents - obj vis cull, shadowing, tessellation, simd compiler opt.

http://appft1.uspto.gov/netacgi/nph...68".PGNR.&OS=DN/20050198468&RS=DN/20050198468

Method and apparatus for superword register value numbering

BACKGROUND OF THE INVENTION

[0002] In a computer system, a compiler is utilized to convert a software program in a programming language into machine language. A processor then may execute the machine language to perform the operations designated by the software program. However, inefficiencies arise when using compilers due to an overlap of executable instructions within the programming language and subsequent redundancies in the machine language program.

[0003] As part of the compiler process, compilers attempt to determine the equivalence of expressions. If two expressions are equivalent, then the second computation can be removed from the program. There are existing techniques that determine equivalence using hash-based value numbers. These techniques are limited because existing techniques only identify equivalence of scalar expressions and do not find equivalence of vectors.

Looks like hash tables to cache results of SIMD vector calculations. :cool:


http://appft1.uspto.gov/netacgi/nph...87".PGNR.&OS=DN/20050195187&RS=DN/20050195187

Method and apparatus for hierarchical Z buffering and stenciling

Looks like a response to ulta-shadow on Nvidia hardware.

http://appft1.uspto.gov/netacgi/nph...88".PGNR.&OS=DN/20050195188&RS=DN/20050195188
Method and apparatus for dual pass adaptive tessellation

http://appft1.uspto.gov/netacgi/nph...86".PGNR.&OS=DN/20050195186&RS=DN/20050195186

Method and apparatus for object based visibility culling
 
Last edited by a moderator:
  • Like
Reactions: Geo
Jima13 said:
? Sept 8,2005

also, isn't Mitchell working for Valve now?

You're looking at the publish date.

Look down a little further down on the left, below the assignee name/address.
 
rwolf said:
Thats when they would be designing the next generation hardware.

Could be. All I have to go on is Russ' explanation for how his company does it, which may or may not be accurate for these companies. What he said is that the "patent scrub" tends to be about the time they are finishing up final design, and filing with the patent office about the time the product goes on sale.

Assuming you'd like to protect what is actually going to go in the hardware, rather than just what your theoretical work has told you ought to work (sure, hopefully they turn out pretty close --but I suspect it is far from unheard of for the actual critter to significantly tweak the theoretical work), it does make sense to me that you'd like to see working hardware of some sort before you file your patents. It seems ATI didn't have first working hardware on R520 until well past March 2004.

But I'll be the first to admit that I have only "glimmers in the dark" on how this process really goes with these two companies.
 
rwolf said:
Look how old this patent is. Do you thing they have unified shader hardware in 2003?

Hrrm. Reasonable point. "I think I better think it out again."

Edit: How's this: that one breaks the pattern because it is associated with the original R400 design. So they finish final design and do their "patent scrub" (resulting in this patent), but then they abandon R400 for the moment because of concerns over whether the process available can make it work . . .but go ahead and file the patents anyway. "The exception that proves the rule" as it were.

EditII: I wonder if we'll ever get a really good picture of the original R400? It'd be cool to see some of that documentation. Maybe R700ish timeframe it will have cooled off enough to share? Probably not.
 
Last edited by a moderator:
What impresses me is that ATI has had a fair number of great patents lately with efficiency in mind.

I haven't found a whole lot for Nvidia yet, but that doesn't mean they don't exist.
 
rwolf said:
What your saying is that they would need R400 silicon in order to file the 2003 patent.

Is that me or Demirug you're aiming that at? I'll assume me.

What I'm saying is that would be the normal practice, short-circuited in this case by the fact that silicon never actually happened. Altho, to be truthful, I don't know that to be true --how far did they get? Are you sure you know? But still, even if it didn't, that doesn't change my general point, it just shows that R400 was not the normal progression --which is famously clear to everyone who follows the industry.

Demirug seems to be pointing at the Xenos IP developed for sale to MS as an alternate explanation for why that patent might have followed a different path than "the usual" that I posited above (which, again, was based on Russ' own experience with his company).
 
Geo, I know so little about it that I can only guess. I might be wrong, but I would assume they create the design, perhaps simulate it and the file a patent.

Sireric would be a good person to ask about how they work.
 
rwolf said:
Geo, I know so little about it that I can only guess. I might be wrong, but I would assume they create the design, perhaps simulate it and the file a patent.

Sireric would be a good person to ask about how they work.

And you may not be wrong. ;) I'm just going with "the best source available to me" at the moment.

I'm sure we'd all love to have Sireric give us some daylight on that process and the relative timing of it in relation to the development life-cycle!

Edit: And, btw, I love it when folks bring this stuff here, and hope my being argumentive about relative timing on this thread does not obscure that fact!
 
Last edited by a moderator:
rwolf said:
Geo, I know so little about it that I can only guess. I might be wrong, but I would assume they create the design, perhaps simulate it and the file a patent.

Sireric would be a good person to ask about how they work.

No, you don't need to prove that something is working to fill a patent.

Because of this you will find even some "perpetuum mobile" (not sure if this is the right englisch word) in the patent database.
 
Demirug said:
No, you don't need to prove that something is working to fill a patent.

Because of this you will find even some "perpetuum mobile" (not sure if this is the right englisch word) in the patent database.

Perpetual motion machines.
 
Demirug said:
No, you don't need to prove that something is working to fill a patent.

Because of this you will find even some "perpetuum mobile" (not sure if this is the right englisch word) in the patent database.

I think we call those "perpetual motion machines".
 
Not that "perpetuum mobile" is an English term, anyway ;)

I don't think filing a patent when the product goes on sale is common in this industry. For example, the "CineFX patent" was filed June 19, 2001.
 
Back
Top