horizon1231, on 01 February 2013 - 06:47 PM, said:
as sidefx told,they don't want to implement bullet in sops.they want to keep framework of DOPs.if everything stays in sops would be faster then current version,but losing the beauty of houdini.
ppl in production really want speed and stability,thats why we all love Milan' work.
Maybe there exists a middle-ground... kinda like what the 'solver sop' does - although that is still a bit different. Or a version of bullet in dops that has a reduced overhead so it runs faster.
MilanSuk, on 02 February 2013 - 10:39 AM, said:
I tried to have deal with Orbolt, which would be the easiest way for users and me. They are interested in my work, but unfortunately they are not prepared for selling c++(HDK) plugins. They have plans for this, but its not immediate future
Too bad indeed, it is a bit of a can of worms to try and support this from sidefx point of view. They generally only do it if the demand is high enough.
Here is another idea:
Start a kickstarter, set a goal and see how far you get with the community chipping in to potentially make your source code available as open source. I would chip in -- this is worth more than a bunch of training videos and might inspire you or others to create more tools too. Hopefully the survey will give you an idea of what price-range you can aim for.
I mentioned this as well when the orbolt store first opened:
magneto, on 02 February 2013 - 09:39 PM, said:
Can you please tell relatively new people like me the story/history of cluster tools and the voronoi fracture?
Were they made by users outside SESI? My understanding was that SESI never added anything users made outside SESI, so basically anyone who is not a SESI developer.
'The history of the cluster tools' ... It could be a title for a sci-fi novel
haha. I'll induldge to go off track a bit. But let's keep this thread about Milan's great work - in a sense it will help him consider certain options he might not have been aware off.
You are correct to say they were made by users inside SESI, you are incorrect in thinking those users were always part of SESI.
*) Let's start with the voronoi tools - you have some reading to do if you want to know more, but basically John Lynch first implemented them as a digital asset. I even mentioned to him back then this should become part of houdini ( http://forums.odforc..._192#entry68583
). So he went on and joined sidefx for a few months to help develop and integrate the voronoi fracture tools. Some of the results of that is that the underlying worker nodes required, such as the tetrahedralize node ( which calculated the delaunay diagram) got implemented in the HDK.
*) Some of my own tools from my own thesis on fluids made it into houdini as well - back then there were no 'attribute to volume' or 'volume to attribute' tools. These are now part of the volume sourcing sop. You could go further to say that back then we had to set up the source relationship, rather than add the fields directly to whatever we wanted to add it to (which is a much more straightforward way of thinking about it). -- The source relationship has now been mostly replaced with the source volume dop. -- Everyone is influenced by ideas and work from others, even in my own thesis I gave thanks and had access to the mind of a friend and colleague, his name is Coen Klosters. Coen is quite an exceptional individual who started out at Bournemouth doing his masters when I was finishing my undergrad - we got along well. He went on to do the sidefx internship in Toronto (the developers office), where he did a great job working together with Jeff Lait to develop the pyrotools. Then he came back to London and joined Cinesite, and so did I when I graduated. This is where the team had to handle massive amounts of data for the scorpion sequence in the first Wrath of the Titans. And then later we had to handle even more data for the Thern in John Carter. Coen went back to Sidefx after that, this time as a developer, and that is when a lot of the cluster tools were developed as they help partition massive datasets so they can be processed in parallel.
--- Have a look at the credits, see how many names you can recognize
(I know around 30 of these names at the moment and have worked with some of them directly) - http://www.sidefx.co...ini12.1/credits
*) The fast point instancer... hmmm, a few years back I was trying to develop my own instancer for the HDK ( http://www.peterclaes.be/blog/?p=151
) - as a lot of people needed that kind of functionality and it already had been developed at a lot of studios for renderman - but was being kept propreitary. It is not a super difficult thing to implement, but it is very useful. And now it is part of houdini through the "fast (point) instancer". Back then I was thinking to turn that entire project into another tutorial dvd, but I did not have the time for it -- and I am not the best c++ programmer either
-- so sharing it with the community and igniting other people's passion and interest in rendertime geometry was pretty cool to see - and I learned a lot through the follow up discussions too.
*) OpenVDB - This has got to be one of the most exciting new collaboratively developed toolsets in recent years. I wish there were more R&D aggreements with sidefx like the one with dreamworks.
magneto, on 02 February 2013 - 09:39 PM, said:
If they do add them, then what's the procedure? Why was HOT not added?
I don't really think there is a strict procedure. All of the above shows that if there is enough interest in the community it will find it's way into houdini. It is a feedback loop of itself. You share something (tool/setup/idea) with the community, someone else picks it up, develops it some more, posts it back,... and when enough people do that it starts trending and feature requests are made to sidefx or even outside TDs directly approach sidefx because they strongly believe a certain setup and tool can make a big difference for the benefit of the software as a whole.
Same goes for HOT, Maya had some ocean tools for years, but not sure how much they were actually used in production. Then Naiad came along (as well as improvements in CPU/GPU) and it became possible to simulate and interact with really large bodies of water. This is thanks to the FLIP/PIC research and Robert Bridson willingess to share so much of this research with the community. But of course you don't want to simulate an entire ocean, so now integration between HOT and Flip fluids is a focus point. Again, when enough shots require that kind of work and it starts trending it will find it's way into houdini.
-- Sidefx as a company is quite visionary, part of that vision also comes from their user base and the attention they pay to their users. If every studio develops their own bullet sop implementation, sidefx must realize there is a huge demand for that - at that point it is only a matter of time. And when it finally gets implemented it changes things in such a way that people don't even realize how the workflow was in the past.
-- So, Milan, in my opinion, if you really want to 'force the issue', the first step is to make the source code available (kickstarter or sell it or whatever) and get this project trending as a community project. That way demand is building because when people move from one place to the next they will want to be able to create these kind of rbd effects. Granted, a lot of studios already have their own implementation... but how long do you think they will keep updating their internal in-house versions when there is a completely community supported version out there that somebody is working on for the latest version of houdini. People want to work efficiently, and if they can use the same tool between facilities without having to learn in-house tools this will save the facility time/money as well. And then when the community is used to having this tool, they will start to wonder why it is not part of houdini already? And they will complain -- as software users always do
-- but in a good and constructive fashion, the way the houdini community has a wishlist. I wonder if "bullet in sops" already made it onto the H13 wishlist. If not... only a matter of time. You might as well be the one to spearhead this -- others have joined sidefx to help develop certain things (even if it was only for half a year or so), perhaps this could be an option for you too.