Cloth constraint sim time

[Dispel]

So I have a static object, which I constrained a cloth object to, using the new FEM cloth solver stuff. I've been using the shelf tools..

As soon as I constrain one object to another, my sim time goes through the roof. Like from 30 seconds a frame to 60+ minutes per frame.

Having watched the SideFX cloth masterclass, it never seemed like the constraints affected performance much. Any ideas for me here?

[Guest tar]

Not sure exactly what you're doing 17fps processing time for a cloth grid 'stitch'Cloth' to a Static Object torus. Latest build

CLothSim.hip

Edited December 1, 2013 by tar

[Dispel]

What is supposed to be the major difference between using a cloth-stitch-constraint and cloth-attach-constraint?

I'll try a stitch constraint instead. I was using the "cloth attach" shelf button, unawares that stitching might yield better performance.

[Dispel]

also, when I use the stitch cloth tool, I get an error complaining about "invalid literal for int() with base 10: "

"Traceback (most recent call last):
  File "dop_stitchconstraint", line 5, in <module>
  File "C:/PROGRA~1/SIDEEF~1/HOUDIN~1.21/houdini/python2.7libs\dopclothtoolutils.py", line 906, in createClothStitchConstraint
    constrainedPTgroup)
  File "C:/PROGRA~1/SIDEEF~1/HOUDIN~1.21/houdini/python2.7libs\dopsbdtoolutils.py", line 477, in generateSBDAttachmentGoalPTGroup
    constrainedPTgroup):
  File "C:/PROGRA~1/SIDEEF~1/HOUDIN~1.21/houdini/python2.7libs\dopsbdtoolutils.py", line 64, in selectedPointNumbers
    return [ int(pt) for pt in pts.split(" ") ]
ValueError: invalid literal for int() with base 10: ''

[Guest tar]

The internal difference is the 'goal surface' node in the Stitch Cloth uses points and the Attach to Body 'goal surface' uses primitives groups it appears. I'm not sure which tool is the "cloth attach" shelf tool though, that one isn't in the Cloth shelf here.

What is supposed to be the major difference between using a cloth-stitch-constraint and cloth-attach-constraint?

I'll try a stitch constraint instead. I was using the "cloth attach" shelf button, unawares that stitching might yield better performance.

The Stitch Cloth shelf tool is working on OsX 13.0.248. Might be worth logging a bug with SideFX.

also, when I use the stitch cloth tool, I get an error complaining about "invalid literal for int() with base 10: "

"Traceback (most recent call last):
File "dop_stitchconstraint", line 5, in <module>
File "C:/PROGRA~1/SIDEEF~1/HOUDIN~1.21/houdini/python2.7libs\dopclothtoolutils.py", line 906, in createClothStitchConstraint
constrainedPTgroup)
File "C:/PROGRA~1/SIDEEF~1/HOUDIN~1.21/houdini/python2.7libs\dopsbdtoolutils.py", line 477, in generateSBDAttachmentGoalPTGroup
constrainedPTgroup):
File "C:/PROGRA~1/SIDEEF~1/HOUDIN~1.21/houdini/python2.7libs\dopsbdtoolutils.py", line 64, in selectedPointNumbers
return [ int(pt) for pt in pts.split(" ") ]
ValueError: invalid literal for int() with base 10: ''

Edit: Also just tested the Attach to Body and have the same speed processing - 19fps on the attached file.

AttachToBody.hip

Edited December 1, 2013 by tar

[Dispel]

I've been on 13.0.198, I'll see if moving up a version fixes it.

[Dispel]

Preliminary result, but moving to the new version seems to make at least one of my cloth attach constraints simulate MUCH faster. In fact, the impact of the constraint is barely noticeable now, I'm finding. Adding on the other constraints shortly.

[Dispel]

Looking at the bug fixes on the sidefx website, I find this:

Tuesday, November 5, 2013

Houdini 13.0.225: Fix for slowdown that sometimes occurred when the finite element solver runs out of linear solve iterations.

I'm guessing this is the culprit, as I was getting a warning to the effect of running past the linear solve limit.

[Guest tar]

Great! The SideFx are rad and daily builds rock!