sanostol

I wonder if this is due to the mismatch of the gcc version. in the houdini system requirements they mention gcc 6.3.1, but debian 10 has gcc version 8.3.0

ok, nvidia is now workign, but the cpu drivers are still breaking: this is the output of hgpuinfo wen HOUDINI_OCL_DEVICETYPE is set to CPU: OpenGL Vendor: NVIDIA Corporation OpenGL Renderer: GeForce GTX TITAN/PCIe/SSE2 OpenGL Version: 4.6.0 NVIDIA 418.74 OpenGL Shading Language: 4.60 NVIDIA Detected: NVidia Consumer 6144 MB 418.74.0.0 Supported Extensions not in OpenGL 3.3 (Version): GLX_SGI_swap_control GL_ARB_base_instance (4.2) GL_ARB_bindless_texture GL_ARB_bindless_texture GL_ARB_buffer_storage (4.4) GL_ARB_clear_buffer (4.3) GL_ARB_compatibility GL_ARB_compute_shader (4.3) GL_ARB_copy_texture (4.3) GL_ARB_debug_output GL_ARB_gpu_shader5 (4.0) GL_ARB_gpu_shader_fp64 (4.0) GL_ARB_multi_bind (4.4) GL_ARB_program_interface_query (4.3) GL_ARB_robustness GL_ARB_sample_shading (4.0) GL_ARB_separate_shader_objects (4.1) GL_ARB_shader_image_load_store (4.2) GL_ARB_shader_storage_buffer_object (4.3) GL_ARB_shading_language_packing (4.2) GL_ARB_tessellation_shader (4.0) GL_ARB_texture_buffer_object_rgb32 (4.0) GL_ARB_texture_cube_map_array (4.0) GL_ARB_texture_filter_anisotropic (4.6) GL_ARB_transform_feedback3 (4.0) GL_ARB_vertex_attrib_64bit (4.1) GL_EXT_abgr GL_EXT_blend_minmax GL_EXT_direct_state_access GL_KHR_debug (4.3) GL_NVX_gpu_memory_info GL_NV_gpu_shader5 Unsupported OpenGL Extensions used by Houdini: EXT_extensions_string GL_ATI_meminfo OpenGL Limits: ELEMENTS_VERTICES 1048576 ELEMENTS_INDICES 1048576 VERTEX_UNITS 32 TESS_CONTROL_TEXTURE_IMAGE_UNITS 32 TESS_EVALUATION_TEXTURE_IMAGE_UNITS 32 GEOMETRY_UNITS 32 FRAGMENT_IMAGE_UNITS 32 COMBINED_UNITS 192 TEXTURE_SIZE 16384 3D_TEXTURE_SIZE 2048 CUBE_MAP_TEXTURE_SIZE 16384 RECTANGLE_TEXTURE_SIZE 16384 TEXTURE_ARRAY_LAYERS 2048 MAX_COLOR_TEXTURE_SAMPLES 32 MAX_DEPTH_TEXTURE_SAMPLES 32 FBO_ATTACHMENTS 8 DRAW_BUFFERS 8 VIEWPORT_DIMS 16384x16384 VERTEX_UNIFORMS 4096 FRAGMENT_UNIFORMS 4096 VERTEX_ATTRIBS 16 VERTEX_OUTPUT_COMPONENTS 128 MAX_CLIP_DISTANCES 8 PATCH_VERTICES 32 TESS_GEN_LEVEL 64 TESS_PATCH_COMPONENTS 120 TESS_CONTROL_OUTPUT_COMPONENTS 128 TESS_CONTROL_TOTAL_OUTPUT_COMPONENTS 4216 TESS_EVALUATION_OUTPUT_COMPONENTS 128 GEOMETRY_UNIFORM_COMPONENTS 2048 GEOMETRY_INPUT_COMPONENTS 128 GEOMETRY_OUTPUT_COMPONENTS 128 GEOMETRY_TOTAL_OUTPUT_COMPONENTS 1024 GEOMETRY_OUTPUT_VERTICES 1024 UNIFORM_BLOCKS 84 UNIFORM_BLOCK_SIZE 65536 UNIFORM_BUFFER_BINDINGS 84 COMBINED_VERT_UNIFORM_BLOCKS 14 COMBINED_VERT_UNIFORM_COMPS 233472 TESS_CONTROL_UNIFORM_BLOCKS 14 TESS_CONTROL_UNIFORM_COMPONENTS 2048 TESS_EVALUATION_UNIFORM_BLOCKS 14 TESS_EVALUATION_UNIFORM_COMPONENTS 2048 COMBINED_GEO_UNIFORM_BLOCKS 14 COMBINED_GEO_UNIFORM_COMPS 231424 COMBINED_FRAG_UNIFORM_BLOCKS 14 COMBINED_FRAG_UNIFORM_COMPS 233472 MAX_FRAGMENT_STORAGE_BLOCKS 16 MAX_STORAGE_BLOCK_BINDINGS 96 TRANSFORM_FEEDBACK_INT_COMPS 128 TRANSFORM_FEEDBACK_SEP_COMPS 4 TRANSFORM_FEEDBACK_SEP_ATTRIBS 4 LINE_WIDTH_RANGE 0.5 - 10 LINE_WIDTH_GRANULARITY 0.125 POINT_SIZE_RANGE 1 - 189.875 POINT_SIZE_GRANULARITY 0.125 Segmentation fault It crashes as soon as the opencl part begins Does anybody know what packages I have to install?

solved: apt install nvidia-opencl-common apt install nvidia-libopencl1 now it works fine ( so far )

there are some ways to use external files. one way is to load Your script into the hda in the Scripts tab. and then You can access it with toolutils.createModuleFromSection("bar", kwargs["type"], "bar_PythonModule") you can then access the methods by pwd().hdaModule().bar.... if I remember right, stalkerx777, explained this to me some time ago, thanks again this works really well Another way is to use the python2.7libs folder in a houdini path folder or in You houdini home directory every python script stored there can be imported as a module

Hello, as Debian 10 is out, I wanted to give it a try, as Debian is always a good move So far it is quite smooth, it installs houdini without any errors, Houdini starts and everything on the first look works. But (of course there is a but ) as soon as I do "help/about houdini" or try to use anything with opencl, I get a fatal error. nvidia drivers are installed and working, but my guess is that I need some opencl libs did anyone try this already? Thanks Martin

it was a bug, and sidfx fixed it.

Hi , I want to render intersection geomtry as uniform volume but as soon as it intersects I get a strange result. are there any solutions for this? for the first image I used boolean to get rid of the intersection and this is the result I'm looking for thanks in advance Martin EDIT: Last image is a mix, intersection on geoimetry causes the artefact, on object level it does not volume.hip

A small summary from the Houdini -Carbon Workflow for the german Movie Project "Die kleine Hexe". Hope You like it

thanks alot, jsut had time to check the HOUDINI_PATH way, and it works great, I just add it in my project based startscript and everything seeems to work fine.

no, I think You misunderstood me. No caching, just about $JOB and how to use it like $HSITE

Hello, as python libs stored that way work fine (like all the other stuff , otls, presets, scripts): $HSITE/houdiniX.Y/python2.7libs I was wondering how to do this $JOB related the goal is to have assets and otls available per project. As old school mentions in this post, it should be possible by adding houdini directory into the job folder. $JOB/houdini/python2.7libs but libs and otls are not recognized. the other way would be to setup all search paths manually in a startup script, if there is a $JOB way to do it it would be much cooler thanks in advance

Hi, I'm wondering where to place procedural shaders in future builds of houdini, material network or shop network. shopnetwork is old and probably dissapearing in near future, but for example in materialnetworks there is no alembic procedural and the cvex stuff did not work as expected, so I had to go back to shopnetwork any insights on this one? martin

the new opencl drivers on nvidia cards can access all or at least much more memory, but its hard to track as the ram usage changes really fast, and changes from simulation to simulation, for example when You use more grids.

Unless You have a verifiable permission from Your supervisor, just don't do it. It might bring You in really unnecessary problems. And even then, what's the benefit, do You get paid for work at home? You might reply, You are stuck with a problem and want to solve it at home so You are covered at work. That is as honorable as short-sighted. You are not a brownie, doing all the work over night (unless You are one, but then why care about legal consequences). Also in a sane production environment, You supervisors probably want You to do You work in a predictable manner. And if things get stuck, what will happen from time to time, as Your very task is to solve new problems with every task, a sane production expects this and wants to know about it, so they can react. And this is nothing, that should be seen personal. It is simply not. It is just reasonable for all involved. You have a problem with a production scene? It does make more sense to isolate the problem in a little test scene, that You can much more likely ask to take at home if You still want the fairy tale number, but the huge benefit is You can share it with others or support ( which is outstanding ) and even for You alone it is much easier to work on the problem alone and not carrying a collapsing skyscraper with You, when You fiddling deep in a vop loop.

what about the opencl performance compared to a titanx(pascal), 16 gig ram are tempting