% hconfig -h HOUDINI_DEFAULT_TEXTURE_COLOR HOUDINI_DEFAULT_TEXTURE_COLOR Specifies the default color to use when texture maps cannot be found while rendering. This can be specified as a single string containing four floating point numbers representing the red, green, blue and alpha components of the default color. For example: setenv HOUDINI_DEFAULT_TEXTURE_COLOR "0 0 0 0" to set it to zero black.
Found 662 posts.
Search results Show results as topic list.
Houdini Lounge » Mantra default missing texture color?
- mark
- 2593 posts
- Offline
Solaris and Karma » MaterialX online library
- mark
- 2593 posts
- Offline
The dev team is currently looking into this and hopefully will have some answers for you soon.
Technical Discussion » Need help, AMD CPU is much slower than Intel CPU
- mark
- 2593 posts
- Offline
When we multi-thread VEX, we try to ensure there's enough work for all threads to be fully pegged. When there are only 4,000 points, we may not be able to saturate all the processors all the time. When there are 4,000,000 points, all the threads will kick in.
I see similar performance when I test on Intel vs. AMD locally -- though I'm running on Linux, and I don't have a 5950x.
I've spoken to AMD to see if they can reproduce this issue or shed more light on it.
I see similar performance when I test on Intel vs. AMD locally -- though I'm running on Linux, and I don't have a 5950x.
I've spoken to AMD to see if they can reproduce this issue or shed more light on it.
Technical Discussion » Unable to download Optix denoiser
- mark
- 2593 posts
- Offline
If you have a modern nvidia driver (435 or later) you'll should use the Optix 7 denoiser which is built in the driver itself and doesn't need a download. It should work automatically.
Technical Discussion » Programming error
- mark
- 2593 posts
- Offline
This error is thrown when a render product has a mis-match on the number of render variables (AOVs) expected.
It looks like this might happen if you have an error loading one of the render vars. There might be more information if you render with verbosity level 9 (for debugging).
It looks like this might happen if you have an error loading one of the render vars. There might be more information if you render with verbosity level 9 (for debugging).
Technical Discussion » Houdini 18 Karma Custom Fisheye Lens Shader Revisited
- mark
- 2593 posts
- Offline
jsmack
Do CVEX shader builder vops work with Karma/husk as lens shaders?
There's no process for Houdini to pass the lens shader code and have it bound to a camera in Usd, so the VOP network still has to be saved to disk somewhere.
However, we have converted the Lens Shader SHOP to a VOP network, and broken it up into a Physical Lens Shader Core (which is used by the Physical Lens Shader). So, at least it's a little easier to edit lens shaders using VOPs now.
Technical Discussion » Houdini 18 Karma Custom Fisheye Lens Shader Revisited
- mark
- 2593 posts
- Offline
Hi @Brad - I think you need to ensure that karma/husk can find the HDA when it starts up. I think there are a few ways to do this. One way is to ensure the .hda with the shader is installed in the HOUDINI_OTL_PATH (i.e. $HOME/houdini19.0/otls/lens_shader.hda).
Solaris and Karma » ggx bsdf function
- mark
- 2593 posts
- Offline
Andy_23mark
A public C++ BSDF interface is unlikely.
With Karma now gold and a fully fledge USD renderer, is an API in sight?
A C++ interface is still pretty unlikely - especially with XPU.
Houdini Lounge » XPU tests (with sample scene)
- mark
- 2593 posts
- Offline
jsmackI don't think we've tried Nvlink, but we have tried multiple GPUs. In fact, one sharp eyed person noticed that in this Entagma video, it looks like there are 4 GPUs working with alongside the CPU to render the image: https://youtu.be/tI4zqTezniI?t=602 [youtu.be] (top right corner of the Karma render).GCharb
Since they only support Nvidia, I wonder if they'll ever support NVLINK?
XPU uses Optix, an Nvidia library. Perhaps Mark can speak to Nvlink support.
Solaris and Karma » Karma EXR data window
- mark
- 2593 posts
- Offline
Just a bit of clarification. I looked into the bug and it has to do with overscan only. AFAIK, Karma is saving the data window defined in the render settings to the .exr file.
Are you talking about the auto-crop feature? If so, please submit an RFE for it - it's unlikely to make it into H19 though.
Are you talking about the auto-crop feature? If so, please submit an RFE for it - it's unlikely to make it into H19 though.
Solaris and Karma » Karma EXR data window
- mark
- 2593 posts
- Offline
I think there's a bug already submitted for the data window. I'll try to raise its priority.
Solaris and Karma » Does Karma use RAT files if it finds them when set to EXR
- mark
- 2593 posts
- Offline
If the parameters in the principled shader refer to the EXR files, then Karma will load the EXR files. If you want them to load RAT files, you'll have to change the parameter in the principled shader.
Technical Discussion » Lens Shader Clipping Planes
- mark
- 2593 posts
- Offline
Is this for mantra or for karma? For karma, you want an
export vector2 clippingrange = {0,1}
(see physicallens.vfl)
Technical Discussion » Husk command line multiple frame error
- mark
- 2593 posts
- Offline
My guess is that the shell is expanding the $F4? There are a couple of alternative representations. From the help:
-o [ --output ] arg Output image. Variables are expanded in
the string can be represented in various
ways:
- $F, $FF $F4: Current frame number
- $N: The N'th frame in the sequence
- <F>, <FF>, <F4>: Frame (<UDIM> style)
- %d, %g, %04d: Frame (printf style)
Houdini Indie and Apprentice » Mantra takes 45 min to render 640x360 donut model
- mark
- 2593 posts
- Offline
Actually, we may have found the issue. Pixels (197,424) and (198,424) in the .exr have infinity values. This throws the importance sampling algorithm in mantra for a loop.
Houdini Indie and Apprentice » Mantra takes 45 min to render 640x360 donut model
- mark
- 2593 posts
- Offline
smbell
I am running into a very similar situation here with an AMD Ryzen 9 5950x machine that I just built.
Rendering this simple scene with a handful of spheres and a single HDR or EXR image as an environment light takes a really long time to render. I'm talking about almost two hours using the default mantra settings. If I open the same scene at my office on an 8 year old dual Intel xeon machine which is less than half the speed, it renders in about 20 seconds. Converting the HDR image to a .rat file solves the issue on my AMD machine and it renders in 8 seconds.
I understand .rat files are better optimized, but this seems like there is an issue with AMD machines and the handling of HDR/EXR images as environment lights in Mantra.
Can anyone with an AMD machine try to render this test scene?
When loading .rat files, mantra reports something like:
TBF File Cache: 2 file opens (2.13 MB read)
When loading .exr textures, mantra reports something like:
OpenImageIO Texture statistics
Options: gray_to_rgb=0 flip_t=0 max_tile_channels=6
Queries/batches :
texture : 0 queries in 0 batches
texture 3d : 0 queries in 0 batches
shadow : 0 queries in 0 batches
environment : 37217109 queries in 37217109 batches
Interpolations :
closest : 0
bilinear : 0
bicubic : 37217109
Average anisotropic probes : 1
Max anisotropy in the wild : 1
OpenImageIO ImageCache statistics (shared) ver 2.0.10
Options: max_memory_MB=8023.0 max_open_files=512 autotile=64
autoscanline=0 automip=0 forcefloat=0 accept_untiled=1
accept_unmipped=1 deduplicate=1 unassociatedalpha=0
failure_retries=0
Images : 2 unique
ImageInputs : 1 created, 1 current, 1 peak
Total pixel data size of all images referenced : 6.0 MB
Total actual file size of all images referenced : 1.6 MB
Pixel data read : 6.0 MB
File I/O time : 0.0s (0.0s average per thread, for 33 threads)
File open time only : 0.0s
Tiles: 128 created, 128 current, 128 peak
total tile requests : 45377804
micro-cache misses : 19148390 (42.1977%)
main cache misses : 8 (1.76298e-05%)
redundant reads: 0 tiles, 0 B
Peak cache memory : 6.0 MB
Image file statistics:
opens tiles MB read --redundant-- I/O time res File
BROKEN
2 1 8 6.0 0.0s 1024x 512x3.f32 symmetrical_garden_1k.
Tot: 1 8 6.0 ( 0 0.0) 0.0s
1 not tiled, 1 not MIP-mapped
1 was constant-valued in all pixels
Broken or invalid files: 0
And now rendering stats for various machines I was able to test:
18.5.525 - AMD Ryzen Threadripper 1950X 16-Core Processor
.exr texture
Frame Wall Clock Time: 0:00:09.86
Total Wall Clock Time: 0:00:10.01
Total CPU Time: 0:01:13.55
System CPU Time Only: 0:00:03.29
Peak Memory Usage: 505.19 MB
.hdr texture
Frame Wall Clock Time: 0:00:17.33
Total Wall Clock Time: 0:00:17.46
Total CPU Time: 0:05:45.45
System CPU Time Only: 0:00:02.98
Peak Memory Usage: 516.16 MB
.rat texture
Frame Wall Clock Time: 0:00:12.09
Total Wall Clock Time: 0:00:12.21
Total CPU Time: 0:02:55.49
System CPU Time Only: 0:00:03.91
Peak Memory Usage: 512.64 MB
H18.5.525 - AMD Ryzen Threadripper 2950X 16-Core Processor
.exr texture
Frame Wall Clock Time: 0:00:09.12
Total Wall Clock Time: 0:00:09.18
Total CPU Time: 0:01:21.28
System CPU Time Only: 0:00:05.09
Peak Memory Usage: 522.65 MB
.hdr texture
Frame Wall Clock Time: 0:00:16.64
Total Wall Clock Time: 0:00:16.70
Total CPU Time: 0:05:14.13
System CPU Time Only: 0:00:04.89
Peak Memory Usage: 519.21 MB
.rat texture
Frame Wall Clock Time: 0:00:11.43
Total Wall Clock Time: 0:00:11.54
Total CPU Time: 0:02:46.68
System CPU Time Only: 0:00:04.04
Peak Memory Usage: 510.90 MB
18.5.525 AMD Ryzen Threadripper 3990X 64-Core Processor
.exr texture
Frame Wall Clock Time: 0:00:14.50
Total Wall Clock Time: 0:00:14.56
Total CPU Time: 0:04:32.90
System CPU Time Only: 0:00:06.07
Peak Memory Usage: 1.21 GB
.hdr texture
Frame Wall Clock Time: 0:00:17.47
Total Wall Clock Time: 0:00:17.54
Total CPU Time: 0:14:53.45
System CPU Time Only: 0:00:05.86
Peak Memory Usage: 1.22 GB
.rat texture
Frame Wall Clock Time: 0:00:12.39
Total Wall Clock Time: 0:00:12.45
Total CPU Time: 0:04:04.92
System CPU Time Only: 0:00:05.48
Peak Memory Usage: 1.25 GB
I'll see if we can test with a more modern processor, but so far, we can't seem to reproduce this.
Houdini Indie and Apprentice » Mantra takes 45 min to render 640x360 donut model
- mark
- 2593 posts
- Offline
There was a performance improvement in very recent daily builds for non-RAT textures. It was most noticeable on Windows. Please try a recent daily build.
Houdini Indie and Apprentice » Memory leak bug?
- mark
- 2593 posts
- Offline
If you're able to convert your images to either .rat or MIP Mapped .exr files (make sure the files are tiled and MIP Mapped, not just standard .exr files), you might find better performance if you're using VEX to read the textures.
You can use
Tiled, MIP Mapped .exr and .rat image files are built to handle large numbers of texture maps in a single session.
You can use
imaketx
to convert the images.Tiled, MIP Mapped .exr and .rat image files are built to handle large numbers of texture maps in a single session.
Solaris and Karma » I kinda miss full quality bucket rendering
- mark
- 2593 posts
- Offline
Yes, rendering to mplay, you should be able to click to focus. The one tile you clicked on will be rendered to final quality (or almost) and the surrounding tiles will also get some further refinement.
Integrating this into the viewport rendering is a little tricky (mostly due to UI interaction). But we are investigating other options for future releases.
Oh, if you are writing your own delegate and want to use the mouse click information, it's described in: https://www.sidefx.com/docs/hdk/_h_d_k__u_s_d_hydra.html#HDK_USDHydraCustomSettings [www.sidefx.com]
Integrating this into the viewport rendering is a little tricky (mostly due to UI interaction). But we are investigating other options for future releases.
Oh, if you are writing your own delegate and want to use the mouse click information, it's described in: https://www.sidefx.com/docs/hdk/_h_d_k__u_s_d_hydra.html#HDK_USDHydraCustomSettings [www.sidefx.com]
Edited by mark - March 8, 2021 12:41:00
Solaris and Karma » Trouble with Window Conform Policy through Husk?
- mark
- 2593 posts
- Offline
-
- Quick Links