Community
Bifrost Forum
Welcome to the Bifrost Forum. This is the place for artists using Bifrost to ask and answer questions, browse popular topics, and share knowledge about creating effects procedurally using Bifrost. You can also visit the Bifrost Community on AREA to download an array of ready-to-use graphs, read Bifrost news and updates, and find the latest tutorials.
cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 

Deep Sleep

11 REPLIES 11
Reply
Message 1 of 12
max_schoenherr
894 Views, 11 Replies

Deep Sleep

hi folks, everybody here embraced BIFRÖST GRAPH when it came out. a handful of maya tutors (MHS included) published pretty enthusiastic videos, most of them with the notion: let's hook up certain nodes and hope to get something running. in most cases we don't. the problem is the really terrible lack of documentation. slick tutorials about beautiful wind simulation and strands leave us stranded, since we don't know how to connect what kinds of nodes and modules.

 

in the screenshot you see a typical example. marked in red a node – and zero info in the info-section on the right. marked in yellow: creating strands – zero help about which node to start with. the construct strands node of course does not construct strands. it needs extras. fair enough. but why not tell the users in the info section about what they could connect as input and output to get going? your info is very technical and maybe talks to programmers at dreamworks, but not to your main audience, the mass of folks who learn to use maya:

 

  • Constructs a strands geometry - creating a strands from raw data in the graph. This creates a minimal strands object, it does not set point_size, point_tangent, or any other optional properties.

 

i'm afraid bifröst graph already started a deep sleep. no tutorials in weeks, after a series of great tutorials in days, a month or two ago. wake up, folks!

 

----maximilian

11 REPLIES 11
Message 2 of 12
ihooper
in reply to: max_schoenherr

There are a number of reasons for this situation. In the example you gave, that is a "rebel pack" node, which means it may change in a future release. Given the uncertain state, documenting those nodes is usually deprioritized over more stable nodes.  Holiday season and the release of Maya 2020 have also been factors in the apparent slow down. Rest assured that things should be picking up again soon.


Ian Hooper
UX Architect
Message 3 of 12

Unfortunately this is true.

The documentation must be visual sometimes to get an understanding of for example parameters and how they affect simulations. It is also painful to look and see how different parameters affect simulations, when every input change yields in a compilation pause.

 

What I also noticed is, that often lowering resolution of sims only get faster to a certain point. But this doesn't feel very linear, wehen working with really lowres. I can't get close to the speed of Maya Fluids or other solvers.

 

I have seen a presentation ( i will share the link) where they talk about doubling the speed with adaptive sims. But I don't get it working. Tutorials aren't available and parameter change stuttering combined with sim times (on a 16c threadripper 2950x) eats up my time, that I have to stick to other tools at the moment.

 

There is also a thing that bugs me: when I drop i.e. shapes into Graph, it makes a bif shape out of it. But for working interactively, I don't want to care of the BIF shape. When I select the shape node in Graph, it would be super useful to also highlight selected Object in the viewport.

Otherwise I get lost quickly in my scene.

 

There is much to be improved in terms of usability and performance. Hopefully we get a big update soon.

 

Best 

Stefan

Message 4 of 12
max_schoenherr
in reply to: ihooper

i'm patient. i just wanted to put my finger into the wound, as we say in german. the users are eager to use BIF graph, but you left them alone. i guess you are a small team around duncan b. (say hello, we know each other). let me advise you to quickly find a balance between programming/stability/updates on one side and communication/learning on the other side.

 

example: one very brief post per workday.

 

  • day 1: a fire from 6 nodes.
  • day 2: BIF and keyframes; an example.
  • day 3: a random node, its input and output colours and what they mean for connection.
  • day 4: a simple example for "for_each".
  • day 5: why is there a motion blur node in BIF? we do have motion blur in all render engines.

actually, i could to that for you, but we'd have to negotiate time and knowledge. you better do it yourselves, with minimal effort, but once a day. blog or vlog, whatever.

 

keep in mind, it's not the connection editor nerds who are interested in BIF graph, it's the SFX folks. BIF graph can lead them into programming, because the tool is so wonderful – once you get to understand it.

 

greetings from cologne,

 

----maximilian

Message 5 of 12
stamatk
in reply to: max_schoenherr

Hello,

 

Thanks for your feedback.  I would like address some of the concerns around aero performance.

 

The superior physical accuracy of the bifrost aero solver is what we hope sets us apart from Maya Fluids and other solvers.  However, this does not come for free - the solver relies on higher order time integration and a higher number of time steps to produce the most physically accurate results, both of which affect performance, especially the latter.  

 

If physical accuracy and superior results are not targeted to begin with, then one can always set the max time steps on the aero solver to a value of 1 ( the default is 3).  This will have profound performance improvements, especially at small detail sizes, but will not produce the most physically accurate results.  The higher order time integration can also be deactivated, but this will only result in a minor performance gain.

 

As for lower resolution performance, there is a threshold after which viewport performance and graph evaluation supersedes solver times and there are no more gains by further coarsening the detail size.  This is still an area of research for us.  

 

Thanks!
Kosta



Konstantinos Stamatelos

Bifrost FX Product Owner and Designer

Message 6 of 12
max_schoenherr
in reply to: stamatk

hi kosta,

 

was your comment related to this topic? the topic addresses a general view of BIF graph.

 

----maximilian

Message 7 of 12
stamatk
in reply to: max_schoenherr

Hi Maximilian,

 

Yes I was indeed addressing part of this topic.  @doppelgaenger_digital above brought up some aero performance concerns, and I wanted to comment on that.

 

Cheers,

Kosta



Konstantinos Stamatelos

Bifrost FX Product Owner and Designer

Message 8 of 12

Thank your for your reply on this. It just shows, that the amount of good tutorials is still lacking...

 

I know that the aero solver is one of the best out there. In fact I love it. I just wish I could understand it a little better without trying for days to get the results I want. The aero tutorials out there are, at least up to date, not more than a quick overview.

 

We need documentation and tutorials on things like that. 

 

Sorry for confusing on this topic. I just wanted to show where users are stuck 😉

Message 9 of 12

Btw: found this presentation about adaptivity in bifrost! I am very impressed with the adaptivity features. But I have no idea how to use it. Another topic for tutorial on how to setup different adaptivity calculations to speed up or optimize aero sims.

 

https://youtu.be/hChH51W8pNo

Message 10 of 12

 

Hi Maximillian & others who responded to this!

 

Happy new year!

 

Let me try to unpack what you and others are saying, as this thread is kind of going into all different directions.  I will address each comment/concern by itself, as I see them:

 

1. "Deep Sleep" = lack of tutorials for the past month or so.  The answer here is twofold: (a) holidays! and (b) the people at Autodesk who make tutorials were focused on making tutorials for the Maya 2020 release, and not Bifrost.  (Bifrost has to share/try to re-use the Maya resources for things like marketing, tutorials, etc).  The good news is that those resources will start looking at making more Bifrost tutorials again.  In terms of "Bifrost-related Deep Sleep", we are waking up again.

 

2.  Lack of node documentation.  As Ian said, most of the rebel pack nodes have no docs.  The rebel pack is done on the side, it's nobody's full-time job, and it often comes down to not doing it at all, vs doing it without docs.  This is not a matter of a deep sleep, this is a time vs jobs vs resources issue.  We will try to address this however, so we can get the most popular nodes either into the standard build, or at least documented.

 

3. Speed of aero sims at very low resolution.  As the computational problem goes to zero, the biggest overhead in terms of perceived performance is not the simulation or even Bifrost, but the Maya viewport display.  You can see this by simply hiding the bifShape or disabling the viewport, and then hitting play.  You will notice a big increase in "perceived performance".   We are aware of this and will eventually fix it.

 

4. Speed of adaptive sims.  We have been discussing a possible new Masterclass video on this topic.  We will keep you posted! 🙂

 

cheers

Marcus

 

 

Message 11 of 12

Hey all,

 

I'm one of the folk who handles learning here and I can confirm that we were very busy with Maya 2020.  Things should start evening out more now that the release is done.

 

As a quick aside though, I did do a Maya tutorial that involved some Bifrost just before the break.  https://www.youtube.com/watch?v=p6EuH5o_CuQ



Matt C

Senior Content Experience Designer

Maya Documentation | Maya Learning Channel |
Message 12 of 12
max_schoenherr
in reply to: t_chanma

nice video, @t_chanma !

i published one just the other day: https://youtu.be/pxn9uqNO3ZM

Can't find what you're looking for? Ask the community or share your knowledge.

Post to forums  

Autodesk Design & Make Report