Found 96 posts.
Search results Show results as topic list.
Technical Discussion » OpenCL Exception: clSetKernelArg (-49)
- Tom Mangold
- 96 posts
- Offline
It might be related to Houdini AND nvidia drivers. There aren't any new drivers available for my laptop for quite some time, so there's nothing like a recent change in drivers. The PC which still works correctly runs .303 and 517.40.
Technical Discussion » OpenCL Exception: clSetKernelArg (-49)
- Tom Mangold
- 96 posts
- Offline
Installed the latest Houdini version (.403). Error remains (OpenCL Exception: clSet KernelArg (-49))
Am I the only one using Vellum pressure these days? That's what causes the error and the sim to fail.
Am I the only one using Vellum pressure these days? That's what causes the error and the sim to fail.
Technical Discussion » OpenCL Exception: clSetKernelArg (-49)
- Tom Mangold
- 96 posts
- Offline
Not yet.
Just wanted to check with the rest of you, if there are any similar problems. If anybody has H up running, please give it a try. Something subdivided into cloth and pressure constraint and a quick sim. Curious if everybody else is having the same fun as me.
Thanks for checking in advance!
Just wanted to check with the rest of you, if there are any similar problems. If anybody has H up running, please give it a try. Something subdivided into cloth and pressure constraint and a quick sim. Curious if everybody else is having the same fun as me.
Thanks for checking in advance!
Technical Discussion » OpenCL Exception: clSetKernelArg (-49)
- Tom Mangold
- 96 posts
- Offline
Switch to CPU in the Preferences for OpenCL doesn't solve it, but throws another Device error and then even cloth constraint's don't work anymore.
And like I said: Cloth constraints DO work, pressure constraints DON'T.
And like I said: Cloth constraints DO work, pressure constraints DON'T.
Technical Discussion » OpenCL Exception: clSetKernelArg (-49)
- Tom Mangold
- 96 posts
- Offline
I'm getting this error with a scene set up in 19.5.303 when I run it with a different Houdini version on another PC. In my case 19.5.368. It's a Vellum sim and the responsible node for the error seems to be the Vellum constraint one when set to pressure. The cloth constraint seems to work just fine. Both PC seem to up to date concerning the GPU driver.
Anybody else running into this problem? Any way to fix this? Bit annoying since I prepared this scene for my students who are running into the same problem now. : (
Cheers
Tom
Anybody else running into this problem? Any way to fix this? Bit annoying since I prepared this scene for my students who are running into the same problem now. : (
Cheers
Tom
Technical Discussion » Anisotropic highlight are rotating with the object. : /
- Tom Mangold
- 96 posts
- Offline
Not sure what the problem might be. Whether it's RS related or more on the Houdini side, but there's definitely something fishy.
As soon as I rotate the object on it's y axis the highlights start to rotate. Since the "brush/scratch" rotation stays relatively the same, this shouldn't happen.
What can be the cause? I thought it might be related to the tangents calculated by Houdini. That's my only guess. Or does RS need a vector map for anisotropic directions? Usually the greyscale map is the one to go for, right?
Anybody an idea what's wrong with this setup?
Cheers
Tom
As soon as I rotate the object on it's y axis the highlights start to rotate. Since the "brush/scratch" rotation stays relatively the same, this shouldn't happen.
What can be the cause? I thought it might be related to the tangents calculated by Houdini. That's my only guess. Or does RS need a vector map for anisotropic directions? Usually the greyscale map is the one to go for, right?
Anybody an idea what's wrong with this setup?
Cheers
Tom
Technical Discussion » Start Python Sop via button?
- Tom Mangold
- 96 posts
- Offline
Is there a way to start a Python Sop by calling one of its functions from a button, e.g. on the hda containig the py sop? Since the python script generates geometry I can't place it inside the hda's python section. And because the script otherwise evaluates each time I load the scene I want to start it with a button so it only evaluates if I want it to.
Is there a way to achieve that?
Cheers
Tom
Is there a way to achieve that?
Cheers
Tom
Houdini Indie and Apprentice » Run script in python node on button push
- Tom Mangold
- 96 posts
- Offline
I have more or less a similar problem. I want to call a function inside the python sop from the hda in order to start the python sop. The misery is that otherwise it wants to cook every time I load the scene which is a pain in the butt (if I don't put the parts into a function which start the other functions).
Can this be done? Call a script from the hda in the python sop?
Can this be done? Call a script from the hda in the python sop?
Houdini Indie and Apprentice » Vellum glue constraints rotation free
- Tom Mangold
- 96 posts
- Offline
Technical Discussion » Houdini crashes when about to close the file and start new
- Tom Mangold
- 96 posts
- Offline
Me, too. RS is installed (3.5.06 with 19.5.303). After two minutes H crashes.
2588: Segmentation fault.
When I took the Redshift .json out of the packages folder it seems to work. Can this please get sorted out? This seems to be a major bug.
2588: Segmentation fault.
When I took the Redshift .json out of the packages folder it seems to work. Can this please get sorted out? This seems to be a major bug.
Edited by Tom Mangold - Aug. 18, 2022 10:50:30
Technical Discussion » COP file reader problem
- Tom Mangold
- 96 posts
- Offline
O.k., tried both. Python seems to be a no go, unless there's somebody who can come up with a decent loader. Combining 9 images took 4min. Ridiculous.
Using multiple volumes and a sop import is the way to go- 5s to load all the images, arrange and save them.
Thanks for the tip and idea to go this way!
Tom
Using multiple volumes and a sop import is the way to go- 5s to load all the images, arrange and save them.
Thanks for the tip and idea to go this way!
Tom
Technical Discussion » COP file reader problem
- Tom Mangold
- 96 posts
- Offline
Wow! Sounds great. I checked the python part of the DDS loader yesterday, but thought there must be more under the hood. Would be great if it can be achieved without too much hassle.
Thank you!
Thank you!
Technical Discussion » COP file reader problem
- Tom Mangold
- 96 posts
- Offline
Technical Discussion » COP file reader problem
- Tom Mangold
- 96 posts
- Offline
Yes, you are right. COP node. : ) If one spends half his life in SOPs, everything is a SOP, even in COPs. Lol!
This would be awful. The only other node I came across is the Labs DDS node which reads files. Wonder how they created it. Could some python code on a null work as a read node?
This would be awful. The only other node I came across is the Labs DDS node which reads files. Wonder how they created it. Could some python code on a null work as a read node?
Technical Discussion » COP file reader problem
- Tom Mangold
- 96 posts
- Offline
I have some serious problems with the file SOP in COPs.
I'm using an expression to read different image sequences after each other in order to combine wedges in a tableau, like nine wedges at the same time (nine file SOPs reading simultaneously different sequences which then get arranged in a pattern). So every x frames not only the frame number changes, but also the preceding term (like preview.0.00240 will become preview.9.00001).
Unfortunately H won't care about the new file name and just sticks to the former sequence, not loading any new images. Is there a way to make the file SOP actually load the file the expression actually tells it to load? Utterly frustrating to see the correct file path, while the node refuses to accept any other file sequence than the one already processed.
Unfortunately there's not even a python SOP in COPs which enables you to write your own loader. Can't think of anything else.
The "Still image" option seemed to be a possible solution, but it will load only the first image of the sequence. Is there a trick to force the file SOP into loading properly?
Cheers
Tom
I'm using an expression to read different image sequences after each other in order to combine wedges in a tableau, like nine wedges at the same time (nine file SOPs reading simultaneously different sequences which then get arranged in a pattern). So every x frames not only the frame number changes, but also the preceding term (like preview.0.00240 will become preview.9.00001).
Unfortunately H won't care about the new file name and just sticks to the former sequence, not loading any new images. Is there a way to make the file SOP actually load the file the expression actually tells it to load? Utterly frustrating to see the correct file path, while the node refuses to accept any other file sequence than the one already processed.
Unfortunately there's not even a python SOP in COPs which enables you to write your own loader. Can't think of anything else.
The "Still image" option seemed to be a possible solution, but it will load only the first image of the sequence. Is there a trick to force the file SOP into loading properly?
Cheers
Tom
Technical Discussion » How to avoid that air bubbles get stuck in velocity grids?
- Tom Mangold
- 96 posts
- Offline
Well, in the end it's about the flip separating the different densitys properly. What I've found out is that the velocity greed seems to "swallow" single flip particles with a different density. Adding a pop force to these air bubbles doesn't help either. The up force is just to weak in order to be noticed by the grid. And if you turn it up, the whole sim looks strange because of the force getting way too strong.
What's your recommendation?
The best thing I can think of right now is a sop solver with an attribute transfer. In the sop I'd separate the liquid and the air bubbles and then transfer the up force from the bubbles to the surrounding liquid. This way the grid cell what get enough lift to move up the enclosed single air particle.
Anybody with another approach?
Cheers
Tom
What's your recommendation?
The best thing I can think of right now is a sop solver with an attribute transfer. In the sop I'd separate the liquid and the air bubbles and then transfer the up force from the bubbles to the surrounding liquid. This way the grid cell what get enough lift to move up the enclosed single air particle.
Anybody with another approach?
Cheers
Tom
Technical Discussion » IK solver trouble
- Tom Mangold
- 96 posts
- Offline
Technical Discussion » Solve IK vop solver broken? Possible bug.
- Tom Mangold
- 96 posts
- Offline
Technical Discussion » Solve IK vop solver broken? Possible bug.
- Tom Mangold
- 96 posts
- Offline
If the ik chain is aligned to the z-axis, the solver works as expected.
As soon it is not, it starts to get ugly. Turned by 90° the ikchain will go into a bridge/curve like shape.
The ik chain get even disformed if target and control goal are in the very same position.
My best guess: Rest angle mode AND/OR restransforms in the solveik node.
With "No rest angles" selected the solve malfunctions instantly. Tried to get rest_transform into the node, but either without success or it doesn't work. Neither "Compute from rest transforms", nor "Compute from targets" make it work.
Any help is highly appreciated.
Cheers
Tom
As soon it is not, it starts to get ugly. Turned by 90° the ikchain will go into a bridge/curve like shape.
The ik chain get even disformed if target and control goal are in the very same position.
My best guess: Rest angle mode AND/OR restransforms in the solveik node.
With "No rest angles" selected the solve malfunctions instantly. Tried to get rest_transform into the node, but either without success or it doesn't work. Neither "Compute from rest transforms", nor "Compute from targets" make it work.
Any help is highly appreciated.
Cheers
Tom
Technical Discussion » IK solver trouble
- Tom Mangold
- 96 posts
- Offline
Hi, Matt!
Thanks for your idea. Guess the problem is that fbik tries to reach it as a target, which keeps the chain from reaching the goal.
In the end I try to solve a spider leg with 4-5 joints. Unfortunately the regular ik chain solver can't deal with this amount of joints.
Hope the solveik vop solver would work, but it instantly changes the chain, though the goal, up and tip targets are in the same place as the corresponding joints in the chain. So it instantly looks like a curve.
Any idea? I would be fine wit the solver inside of the vop, if it worked. : )
Cheers
Tom
Thanks for your idea. Guess the problem is that fbik tries to reach it as a target, which keeps the chain from reaching the goal.
In the end I try to solve a spider leg with 4-5 joints. Unfortunately the regular ik chain solver can't deal with this amount of joints.
Hope the solveik vop solver would work, but it instantly changes the chain, though the goal, up and tip targets are in the same place as the corresponding joints in the chain. So it instantly looks like a curve.
Any idea? I would be fine wit the solver inside of the vop, if it worked. : )
Cheers
Tom
-
- Quick Links