nosferatu_037

Yeah i actually did exactly the same, the problem is i would get completely different results compared to a vdb analysis sops curl operator. Thats why i thought i was doing something wrong math wise, or i just didnt understand the notation really well. I just used the neighbours() vex command to extract the points sharing the same edge, and went to build a gradient for each v component. I got a 3x3 matrix with the gradient vector components in it, and did the same subtraction like in that example you sent, but it looked so different from volume curl, i couldnt figure out why.

Thanks Pusat But that vop relies on volumes and inside its just a simple volumegradient, like i mentioned, id like to find out how to do the same thing without using volumes and how the math would translate into code.

Hey guysIve been trying this for a while now, but i still am not able to solve it.Other than converting a grid to a volume and running a volume analyse sop, how would one calculate curl of velocity on a grid sop using an attribute wrangle.Ive setup a simple grid and ran it thru a pointvop in which ive just plugged in the output of a curl noise vop into v. Now i just want to calculate the curl of v using just the attribute wrangle using neighbours() vex function.So far ive managed to calculate the gradient of each v component, but the rest of math i cannot translate to proper code. Any ideas?p.s. does anyone have resources how to move from math notations to codeThanks

Just go to Edit->Aliases/Variables->Variables tab Enter the name of your variable, enter the value, say name foo value 0.25 And in any context you should be able to access that variable on a global level as a $FOO. it will always return 0.25. Its an ok thing to do however you usually want to make your assets or setups know about these variables from the parent node containing the setup, rather than the Global Alias/Variable. Its going to be very difficult keeping track of all of those global constants you have and you dont have an easy enough access for a typical user to change the value.

Jens to the rescue!

Just use an attrib wrangle sop, the code for that shouldnt be longer than two three lines and make a preset out of it.

Thanks captain I was thinking about creating a template function for the UT_ValArray i just thought that there could be another way of doing this. As for ids thanks for the info, i was generally confused about it, and have since read the header files which explain the whole thing in more detailed way. Ill give it another go and see how it turns out. Thanks again

Hey guys I have a simple enough problem i wanted to solve in HDK, and i managed to do it, however im sure there has to be a more elegant solution with enough knowledge of HDK classes. I wanted to do a simple sop that compares the same user chosen attribute from two different inputs, and either groups the difference as points group, or deletes points that have the same attrib value across both inputs. Basically i just have the same point cloud on both inputs, the second input has a timeshift node that looks up the previous frame. Its a simple enough exercise, however i have to ask if anyone knows if im reinventing the wheel, if the interface for doing stuff like this is more simple to use, or could give me some hints on how to do it better. Now the problem i came across was how would one define an array to hold values of an attribute of type we dont know in advance? Ive used a hacky solution where i would do the following in cookMySop() //get ptrs to our input geoconst GU_Detail *firstInput = inputGeo(0); const GU_Detail *secondInput = inputGeo(1); //find if attr exist in those geos, lookup is our string that holds the attr name user chose from the UI const GA_Attribute *fattr = firstInput->findAttribute(GA_ATTRIB_POINT, lookup); const GA_Attribute *sattr; //get our total numpts std::cout << "total num of pts (1): " << firstInput->getPointMap().indexSize() << std::endl; // std::cout << (GA_StorageClass)fattr->getStorageClass() << std::endl; //if we have hooked a second input, find attrs and get total numpts if(secondInput) { sattr = secondInput->findAttribute(GA_ATTRIB_POINT, lookup); std::cout << "total num of pts (2): " << secondInput->getPointMap().indexSize() << std::endl; } // If source attribute doesn't exist, error. if (!fattr || !sattr) { addError(SOP_ATTRIBUTE_INVALID, (const char *)lookup); return error(); } //define our arrays UT_ValArray<int32> ptNumListFirst; UT_ValArray<int32> ptNumListSecond; UT_ValArray<int32> ptNumListDiff; //only if our attr is not string shall we compare, have no idea how to figure out string yet if(fattr->getStorageClass() == GA_STORECLASS_INT) { std::cout << "Yes it is a numeric!" << std::endl; //feed our list of attr values in the array FIRST GEO gdp->getPointAttributeAsArray(fattr,gdp->getPointRange(), ptNumListFirst); //if we have connected the second geo if(secondInput) { //feed our list of attr values in the array SECOND GEO secondInput->getPointAttributeAsArray(fattr,secondInput->getPointRange(), ptNumListSecond); //compare our first and second input list and feed the diff in the third input //this is our class method compareUTArrays(ptNumListFirst, ptNumListSecond, ptNumListDiff); } } I actually dont allow the user to pick an attrib from the list thats not int or the node errors out, so basically im forcing the user to choose an int attr. Im sure there is a better way of doing this. Also i tried using the sortedIntersection() method of the UT_ValNumeric array type, however i could never get the difference to be calculated so i had to write my own implementation of the method (compareUTArrays()). Also in the end i use this approach to get the pts in an array that either needed to be grouped or deleted, im not entirely sure if its the "proper" way of doing things: //if we ticked new group, create a new group for pts if(newGrp || deletePts) { //create our new group GA_PointGroup *diffGrp; diffGrp = gdp->newPointGroup(grpName, false); //create a temp group we will delete later if we need to delete points GA_PointGroup *delGrp = gdp->newPointGroup("__tempDelGrp__", false); //get the handle to our int attribute //would prefer if i could somehow make the handle type recognize the attr data type GA_ROHandleI lookup_attr(fattr); int32 value; //iterate thru all points for(GA_Iterator it(gdp->getPointRange()); !it.atEnd(); ++it) { GA_Offset offset = *it; value = lookup_attr.get(offset); std::cout << value << " is value of pts attr" << std::endl; //iterate thru the array and if we have found same attr value add it to the group for(UT_ValArray<int32>::const_iterator itarr = ptNumListDiff.begin(); itarr != ptNumListDiff.end(); ++itarr) { if(value == *itarr) { diffGrp->addOffset(offset); continue; } } } //the diff between the first and second input is found in the diff array //however we need to delete everything thats not the diff, and that is the second inputs members translated into first for(GA_Iterator it(secondInput->getPointRange()); !it.atEnd(); ++it) { GA_Offset offset = *it; delGrp->addOffset(offset); } if(deletePts) { gdp->destroyPointOffsets(GA_Range(*delGrp)); gdp->destroyGroup(delGrp); } } Now i guess i can use the GA_Attribute.getStorageClass and a switch or if statement and initialize my arrays according to the type, but like i said somehow im sure there has to be a more intelligent way of doing this Also im a bit confused about the bumpDataId, what should i do in this case? If i just group my points and dont even modify the geo or attrib values, should i signal houdini to bumpDataId and whats the syntax for doing that? If i delete points however im sure i have to inform houdini that ive modified geo, which class or method should i use to do something like that? I hope all of this makes sense, let me know if it doesnt ill try and elaborate better Any help would be greatly appreciated Thanks Serge

Electron microscope slow-motion video of vinyl LP

Also max seemed to have gotten some new features that are very familiar

Hvala Srecko New renderman tutorials are quite good i have to say. It was a lot of fun trying to figure out how easy/difficult mantra can handle this type of stuff. And all i can say it performed the task very good. The final render in full HD did take a while but i was chasing quality rather than rendertime.

Hey guys Ive posted this on sidefx but thought to post it here as well: Ive stumbled upon a really good tutorial on pixars website by Leif Pedersen about renderman 18/19. http://renderman.pixar.com/view/TheGrandTour It seemed like a really cool looking project, and since prman 19 was not out as non-commercial at the time, i thought how easy/difficult would it be to try and do something like that in houdini. Now im not a lighter or a shader writer, im an fx td, but this really was a challenge for me and an opportunity to play with pbr and shader building, and i have to say i was blown away by how easy it is to set up the shaders and to get the project rolling. So ive downloaded all the project files and exported them as alembic and started building the shaders and trying to replicate them the same way the tutorial suggests. Since the project files had textures and models, it was only the shader building part and the lighting on my part. The difference also is that in the tutorial they already have the shading models and they would build the shading network, where i had to do everything. Sometimes mantrasurface was enough, other times i had to build one in shops from scratch. I also used the delay load archive as well as houdini instancing geo to try and do grass, since fur wasnt cutting it. The curve render is good but renderman has a setting to make the curves appear like they are rounded, which makes the spec look quite different. Since i wasnt able to reproduce this (i did try some hacks suggested by odforce crew, it became too complicated and wasnt giving me proper results), i modeled couple of strands myself and instanced it. Final grass count was around 50000000 and it took 8 min to start the render. Im guessing writing the ifd maybe? Also the new fur system is pretty cool, i had to learn how it works tho since ive never used it before. The guide hair system is great and i managed to replicate the top hair of the character, since guide hairs were provided with proj files. Sometimes i would have trouble to try and find some attributes that are named differently in maya, but what ive concluded is that there should be a help page illustrating how to properly use fur (something like a quick tutorial). I know there are pages in help trying to do this, however its split on per fur tool basis, and i was kinda looking more for one page to go over every tool, since i would have to navigate back and forth a lot to try and get to information i needed. And i side navigation bar would help quite a lot in help The rock shader with moss was a fun thing to do, and ive concluded that there is no way that mantrasurface shader can by default provide an ubersystem solution. Multiple times i found myself wanting to have a way to hook a ramp or a noise patter onto a parameter without diving inside and digging thru the connections.Other times i would simply find the guts of the shader too messy and spending more time finding where i should hook something up, so i would just create a shader from scratch. And the mantrasurface model did provide me with enough inside information to figure out how everything worked, so that was very good. Ive stumbled across some bugs in the meantime and informed sesi thru RFE about them, such as zbrush tangent vector displacement. All in all it was extremely fun, and very rewarding i would think to go thru this project, and i got to learn some new stuff as well. Thumbs up for SESI how much they have improved the entire system, tho there is still work to be done. (default ptex map loading in mantrasurface doesnt work due to links from uv) Also awesome work and tutorial by Leif, thanks to providing models and textures. Here are some images i managed to render let me know what you guys think (i dont know if this should be in WIP section, if so admins please move it): car_prman.v01.mp4.tar.gz

I agree. Maybe it would be a good test also to include the intel openCL test as well the regular one.

Sorry didnt quite make myself clear, thought this is about general GPU usage in simulation. I was only saying that GPU simulation in production is being used and widely at certain places, and that houdinis opencl is on the right track to utilize the gpu in the production. Even tho the 4gb of gpu memory is a bottleneck sometimes, it does improve the performance quite a lot, and now in h14 with the grain solver and intel OpenCL it does look very promising. The only problem ive experienced so far in production was the memory limit of the GPU.

I would have to disagree with this one. At ilm they have the best tool ive ever used for fluid simulation, fire smoke and explosions. Its called plume and the entire sim as well as rendering is done on a gpu. Its so fast that we barely ever used to save cache files, only when particle advection was necessary. All the fluid sims you see on transformers and most of ilm movies in the past couple of years were done on a gpu with no more than 4gb of ram. Its a CUDA gpu solver if i remember correctly, you can actually find some links on it. Its such a stripped down solver which makes it so fast and so easy to iterate. Here is a tech video on it. Sidefx is doing a great job moving dop nodes to be executed on the gpu, it does make a difference.