Houdini-NvFlex icon indicating copy to clipboard operation
Houdini-NvFlex copied to clipboard

simple integration of Nvidia Flex technology into SideFX Houdini

Houdini-NvFlex

simple integration of Nvidia Flex technology into SideFX Houdini

see animated preview here

preview

installation (Windows):

  • copy dll from x64 folder (nvFlexDop_165.dll for houdini 16.5, nvFlexDop_160.dll for houdini 16.0) to any dso location, like for example into C:\Users\<YOUR USER>\Documents\houdini16.5\dso\
  • copy all *.otl files from otls folder to your otls folder, like for example into C:\Users\<YOUR USER>\Documents\houdini16.5\otls\
  • copy all *.dll files from dll folder to any path in your system PATH environment variable. The most straightforward way would be to copy them into your system32 folder, like C:\Windows\system32\ , but i would suggest to copy them into a new folder, for example C:\mystuff\bin\nvflex\ and add that path to system PATH variable. Please, see here for instructions how to access environment variables. just add your folder to the end of the PATH variable. So with example folder above it will looks like <CONTENT OF YOUR PATH VARIABLE>;C:\mystuff\bin\nvflex\

installation (Linux):

  • copy so appropriate for your houdini version from x64/linux64/dso folder to your <home>/houdini16.X/dso/ folder (create dso subfolder if necessary). These so were compiled with gcc 6.3.0, glibc 2.24-11.

  • if you want to compile it yourself:

    • edit linux_build_16X.sh so the helper variables point to the locations of the libraries it requires
    • launch linux_build_16X.sh and if you have all dependencies - build will succeed, and your new so will be put into x64/linux64/dso folder
    • note: depending on your linux distribution you might require different packages. You might also require full Cuda Toolkit 8.0.44 to be able to build, in this case you will have to add paths to your Cuda toolkit to the Makefile. Although some distributions, like debian, have core libs from that toolkit available in reps, so for example for debian - package nvidia-cuda-dev will be enough and you don't have to download full Cuda Toolkit and set any paths manually.

That should do it.

If you encounter problems, like houdini crashes on startup or plugin not working in general: try this work in progress branch

it will at least print out the problem into the console

If the plugin does not load and no console message is printed, but houdini itself launches - you probably forgot to make some of dlls from the bin folder available (see pt.3 of installation guide, and make sure you have microsoft vc 2015 redistributable installed (it would be fun if you don't cuz houdini needs it to launch as well))

Examples

please, see hip files in examples folder