Jump to content

  • Log In with Google      Sign In   
  • Create Account


SIGGRAPH 2013 Master Thread

  • You cannot reply to this topic
26 replies to this topic

#1 Promit   Moderators   -  Reputation: 6108

Like
13Likes
Like

Posted 25 July 2013 - 09:57 AM

I'd like to hold a general summation/discussion of the work presented at SIGGRAPH this year. I'll update this post as we go with links to the various papers and presentations, and you're all free to have an open discussion in the comments.

 

SIGGRAPH 2013 papers on the web - Ke-Sen Huang dutifully updates his page every single year with all of the papers presented at SIGGRAPH.

 

Physically Based Shading in Theory and Practice - All slides/presentations from the 2013 Physically Based Shading course. This course is held yearly and IMO is some of the best material out there on physically based techniques and state of the art. Highly recommend looking up presentations from previous years as well if you haven't seen them.

 

Advances In Real-Time Rendering In Games - Not available yet.

 

Please let me know if anything in particular from the conference caught your eye and should be highlighted.


Edited by Promit, 25 July 2013 - 07:29 PM.


Sponsor:

#2 Promit   Moderators   -  Reputation: 6108

Like
6Likes
Like

Posted 25 July 2013 - 10:02 AM

One of the papers that caught my eye this year in a bad way was the pre-computed cloth simulation. Essentially they took a (small) set of known character animations, put a cloak on the character, and precomputed every potential pose for the cloth and saved it in some kind of compressed 70 MB graph structure. Their proposal is that instead of trying to run cloth at runtime, you do it offline (4500 hours for this, if memory serves) and look up the results in the structure.

 

The paper tries to pitch this for upcoming gen real-time, which I find comical. The set of animations is small, the data is enormous to be traversing at runtime, and betting memory against compute power when GPU compute is coming fully into its own is tone-deaf at best -- if not completely insane. (Nevermind the implications to a production pipeline when the animation requires several thousand hours to regenerate.) While the technical aspects of the work may be well done, I really hate the core idea of this work and IMO it's a good example of bad academic work being pitched to industry.


Edited by Promit, 25 July 2013 - 10:03 AM.


#3 mhagain   Crossbones+   -  Reputation: 7436

Like
0Likes
Like

Posted 25 July 2013 - 11:58 AM

One of the papers that caught my eye this year in a bad way was the pre-computed cloth simulation. Essentially they took a (small) set of known character animations, put a cloak on the character, and precomputed every potential pose for the cloth and saved it in some kind of compressed 70 MB graph structure. Their proposal is that instead of trying to run cloth at runtime, you do it offline (4500 hours for this, if memory serves) and look up the results in the structure.

 

The paper tries to pitch this for upcoming gen real-time, which I find comical. The set of animations is small, the data is enormous to be traversing at runtime, and betting memory against compute power when GPU compute is coming fully into its own is tone-deaf at best -- if not completely insane. (Nevermind the implications to a production pipeline when the animation requires several thousand hours to regenerate.) While the technical aspects of the work may be well done, I really hate the core idea of this work and IMO it's a good example of bad academic work being pitched to industry.

 

So it's keyframe animation given a fancy name?  Yuck.


It appears that the gentleman thought C++ was extremely difficult and he was overjoyed that the machine was absorbing it; he understood that good C++ is difficult but the best C++ is well-nigh unintelligible.


#4 Kaptein   Prime Members   -  Reputation: 1844

Like
3Likes
Like

Posted 25 July 2013 - 12:16 PM

Lovely stuff smile.png

Wish I was there, although I'm not quite qualified to be there tongue.png

 

My favorite so far:

Highly Adaptive Liquid Simulations on Tetrahedral Meshes


Edited by Kaptein, 25 July 2013 - 12:27 PM.


#5 Promit   Moderators   -  Reputation: 6108

Like
3Likes
Like

Posted 25 July 2013 - 02:43 PM


Highly Adaptive Liquid Simulations on Tetrahedral Meshes

Looks like a way of integrating SPH into arbitrary spaces, I guess? I'm not very familiar with fluid simulation work, but the simulation time for their demos is on the order of five hours. That's a bit out of reach sad.png Maybe if they'd precomputed it all and stored it in a giant compressed graph... :D :D


Edited by Promit, 25 July 2013 - 02:43 PM.


#6 Frenetic Pony   Members   -  Reputation: 1187

Like
1Likes
Like

Posted 25 July 2013 - 03:11 PM

One of the papers that caught my eye this year in a bad way was the pre-computed cloth simulation. Essentially they took a (small) set of known character animations, put a cloak on the character, and precomputed every potential pose for the cloth and saved it in some kind of compressed 70 MB graph structure. Their proposal is that instead of trying to run cloth at runtime, you do it offline (4500 hours for this, if memory serves) and look up the results in the structure.

 

The paper tries to pitch this for upcoming gen real-time, which I find comical. The set of animations is small, the data is enormous to be traversing at runtime, and betting memory against compute power when GPU compute is coming fully into its own is tone-deaf at best -- if not completely insane. (Nevermind the implications to a production pipeline when the animation requires several thousand hours to regenerate.) While the technical aspects of the work may be well done, I really hate the core idea of this work and IMO it's a good example of bad academic work being pitched to industry.

 

Completely agree, and having watched the video (correct me if I'm wrong) it only seems to run through it's precomputed tree, rather than offer any interaction with other obejcts, which would require a much larger data set and even more precomputation time at least.

 

Regardless, here's where the presentations for the Advances in Realtime Rendering course will be posted: http://advances.realtimerendering.com/s2013/index.html



#7 swiftcoder   Senior Moderators   -  Reputation: 9612

Like
3Likes
Like

Posted 25 July 2013 - 06:29 PM

MeshGit: Diffing and merging meshes for polygonal modeling sounds pretty cool. Version control for asset creation tools has been on my mind for a while now.

Tristam MacDonald - Software Engineer @Amazon - [swiftcoding]


#8 Promit   Moderators   -  Reputation: 6108

Like
3Likes
Like

Posted 25 July 2013 - 07:28 PM

MeshGit: Diffing and merging meshes for polygonal modeling sounds pretty cool. Version control for asset creation tools has been on my mind for a while now.

Cool idea and work, though as usual for the university crowd they only look at the geometry and ignore all other vertex attributes. The paper includes a throwaway line claiming that including other attributes would be trivial, but I'll believe it when I see it.

 

Jeez, I'm really hating on people's work today. 


Edited by Promit, 25 July 2013 - 08:22 PM.


#9 tool_2046   Members   -  Reputation: 962

Like
1Likes
Like

Posted 26 July 2013 - 11:35 AM

The selfshadow blog has a bunch more links to courses, talks and papers. The massive worlds course looks pretty cool, nice to see some talk about hardware virtual textures. 

 

EDIT: Here's the url: http://blog.selfshadow.com/2013/07/24/siggraph-2013-links/



#10 MJP   Moderators   -  Reputation: 10239

Like
4Likes
Like

Posted 26 July 2013 - 12:41 PM

If you guys have any questions about my talk (Crafting a Next-Gen Material Pipeline for The Order: 1886), feel free to ask. I can also see if I can get the other course presenters on here if you have questions for them.



#11 swiftcoder   Senior Moderators   -  Reputation: 9612

Like
0Likes
Like

Posted 26 July 2013 - 02:02 PM

Those are some mighty fine looking materials. Damn you must have a lot of GPU horsepower lying around on these new consoles...


Tristam MacDonald - Software Engineer @Amazon - [swiftcoding]


#12 MJP   Moderators   -  Reputation: 10239

Like
2Likes
Like

Posted 26 July 2013 - 06:17 PM

Yeah it's pretty nice. biggrin.png

Like most modern GPU's there's plenty of ALU available while you're waiting for reads from memory, and complex BRDF's are a pretty straightforward way to take advantage of that.



#13 Frenetic Pony   Members   -  Reputation: 1187

Like
0Likes
Like

Posted 26 July 2013 - 08:52 PM

Yeah it's pretty nice. biggrin.png

Like most modern GPU's there's plenty of ALU available while you're waiting for reads from memory, and complex BRDF's are a pretty straightforward way to take advantage of that.

 

What are you guys doing for indirect specular?



#14 David Neubelt   Members   -  Reputation: 794

Like
1Likes
Like

Posted 26 July 2013 - 09:38 PM

 

Yeah it's pretty nice. biggrin.png

Like most modern GPU's there's plenty of ALU available while you're waiting for reads from memory, and complex BRDF's are a pretty straightforward way to take advantage of that.

 

What are you guys doing for indirect specular?

 

 

There are details in the course notes but specular probes.


Graphics Programmer - Ready At Dawn Studios

#15 Promit   Moderators   -  Reputation: 6108

Like
0Likes
Like

Posted 26 July 2013 - 11:20 PM

I really don't love the specular probe thing, but I adore IBL and it seems like specular probes are the closest we're going to get in production real-time in the near future. I would like to see more work on that front though... the stuff out there still feels so hacked.



#16 MJP   Moderators   -  Reputation: 10239

Like
0Likes
Like

Posted 26 July 2013 - 11:29 PM

Yeah I'm not a fan of the probes either. You can setup cases where they work fairly well, but the standard methods for pre-integrating the specular BRDF will result in a very large error at the grazing angles. And then of course there's all of the standard issues with placement, blending between probes, incorrect parallax, memory usage, and so on which make them very difficult to work with both for artists and programmers. We've been researching alternatives, but I'm not sure if we'll come up with something in the near future. I'd been kinda hoping that someone else would have a cool idea in this area, but so far I haven't heard of anything outside of the voxel cone-tracing stuff.



#17 davepermen   Members   -  Reputation: 1007

Like
0Likes
Like

Posted 27 July 2013 - 04:46 AM

One of the papers that caught my eye this year in a bad way was the pre-computed cloth simulation. Essentially they took a (small) set of known character animations, put a cloak on the character, and precomputed every potential pose for the cloth and saved it in some kind of compressed 70 MB graph structure. Their proposal is that instead of trying to run cloth at runtime, you do it offline (4500 hours for this, if memory serves) and look up the results in the structure.

 

The paper tries to pitch this for upcoming gen real-time, which I find comical. The set of animations is small, the data is enormous to be traversing at runtime, and betting memory against compute power when GPU compute is coming fully into its own is tone-deaf at best -- if not completely insane. (Nevermind the implications to a production pipeline when the animation requires several thousand hours to regenerate.) While the technical aspects of the work may be well done, I really hate the core idea of this work and IMO it's a good example of bad academic work being pitched to industry.

 

part of my mind went towards "how about for mobile?". but then, nah.. i really think it's the wrong approach. still, the datastructure might have some interesting information in it.


If that's not the help you're after then you're going to have to explain the problem better than what you have. - joanusdmentia

My Page davepermen.net | My Music on Bandcamp and on Soundcloud


#18 Frenetic Pony   Members   -  Reputation: 1187

Like
0Likes
Like

Posted 27 July 2013 - 05:57 PM

Yeah I'm not a fan of the probes either. You can setup cases where they work fairly well, but the standard methods for pre-integrating the specular BRDF will result in a very large error at the grazing angles. And then of course there's all of the standard issues with placement, blending between probes, incorrect parallax, memory usage, and so on which make them very difficult to work with both for artists and programmers. We've been researching alternatives, but I'm not sure if we'll come up with something in the near future. I'd been kinda hoping that someone else would have a cool idea in this area, but so far I haven't heard of anything outside of the voxel cone-tracing stuff.

 

Geomerics, a third party GI thing, had this video I stumbled across a while ago: vimeo.com/geomerics/review/60838484/d2817c3548 (doesn't want to embed) Maybe it's some environment probe hack, but it sure doesn't look like it.

 

It also reminds me of something Epic did. For their Samaritan demo they managed to have static occlusion of their impostor reflections pre-baked. I remember a blog post mentioning something about tracing against a point cloud, but if so I can't find what paper that was gleaned from. Still, while it was static only and in the demo only gave occlusion info, it wasn't an environment probe, was high resolution, worked with arbitrary materials, and was assumedly faster than voxel cone tracing (though that's just an assumption).


Edited by Frenetic Pony, 27 July 2013 - 07:03 PM.


#19 Hodgman   Moderators   -  Reputation: 27685

Like
0Likes
Like

Posted 27 July 2013 - 10:06 PM


It also reminds me of something Epic did. For their Samaritan demo they managed to have static occlusion of their impostor reflections pre-baked. I remember a blog post mentioning something about tracing against a point cloud, but if so I can't find what paper that was gleaned from. Still, while it was static only and in the demo only gave occlusion info, it wasn't an environment probe, was high resolution, worked with arbitrary materials, and was assumedly faster than voxel cone tracing (though that's just an assumption).

IIRC, they bake the scene geometry into a 3D signed distance field. Tracing rays though that kind of structure is very efficient, except in cases where you have a ray that comes close to a boundary, but doesn't quite touch it. In the case where a ray is travelling parallel to a surface's tangent, very close to it, it becomes extremely inefficient, but all other cases are pretty good...

(Again, IIRC:) They trace reflections against a global cube-map, and dozens of reflection quads (either hand-drawn textures, like the neon lights, or slices of geometry that have been captured with artist-placed ortho frustums), using that 3D geometry texture to deal with occlusions.

 

[edit]P.S. congrats on the awesome presentation and work MJP :D


Edited by Hodgman, 28 July 2013 - 01:46 AM.


#20 MJP   Moderators   -  Reputation: 10239

Like
0Likes
Like

Posted 28 July 2013 - 12:33 AM

I thought I had remembered someone telling me that Geomerics was generating cubemaps at runtime, but I have no particular insight into what they're doing these days. And Hodgman is correct, Epic was tracing signed distance fields to get static occlusion for the Samaritan demo, which they coupled with an even more expensive solution for allowing dynamics to occlude reflections on planar surfaces. I'm not sure if they're still doing that for UE4, since it sounds fairly expensive.







PARTNERS