released vkdt 0.5.0

[difrkaguilar (Franklin Aguilar Matos)]

I’ve CUDA wonking in Linux Fedora 37 but anyway vkdt have that error

Blender screenshot.

[123sg (Steven)]

Is this ‘proper’ Fedora or in WSL?

[difrkaguilar (Franklin Aguilar Matos)]

Fedora 37 Workstation Edition

[priort (Todd Prior)]

I have this and ran through the CUDA process… still doesn’t complete the compile for me on WSL2

±----------------------------------------------------------------------------+
| NVIDIA-SMI 525.65 Driver Version: 527.56 CUDA Version: 12.0 |
|-------------------------------±---------------------±---------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|===============================+======================+======================|
| 0 NVIDIA GeForce … On | 00000000:01:00.0 On | N/A |
| 0% 49C P8 11W / 200W | 458MiB / 8192MiB | 0% Default |
| | | N/A |
±------------------------------±---------------------±---------------------+

±----------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=============================================================================|
| 0 N/A N/A 23 G /Xwayland N/A |
±----------------------------------------------------------------------------+

EDIT

This is where I get hung up

make[2]: Entering directory ‘/home/user/vkdt/src’
Package mad was not found in the pkg-config search path.
Perhaps you should add the directory containing `mad.pc’
to the PKG_CONFIG_PATH environment variable
No package ‘mad’ found
glslangValidator -Ipipe/modules -I./ --target-env vulkan1.2 -V pipe/modules/ab/main.comp -o pipe/modules/ab/main.comp.spv
make[2]: glslangValidator: No such file or directory
make[2]: *** [Makefile:161: pipe/modules/ab/main.comp.spv] Error 127
make[2]: Leaving directory ‘/home/user/vkdt/src’
make[1]: *** [Makefile:61: src] Error 2
make[1]: Leaving directory ‘/home/user/vkdt’
make: *** [Makefile:7: all] Error 2

[ggbutcher (Glenn Butcher)]

libmad is the MPEG Audio Decoder, probably used by ffmpeg or asound, both optional dependencies. If you’re not going to do video processing, you can disable those in config.dk

[betazoid (Anna)]

Apparently, noise-profile.sh is not included in the Debian package of vkdt. I will try to just copy the script from the github repo, place it in /usr/bin (apparently that’s where vkdt is on Debian) and execute it from there. Will that work?

[hanatos (jo)]

@difrkaguilar

[ERR] no WSI support on physical device

this here might indicate that it’s trying to use a gpu that does not have a screen attached to it. seems likely since you have so many. if you have the cable to the monitor attached to the 1080Ti, you could select it explicitly in the config file to instruct vkdt not to do stupid things. i suppose the automatic detection can be improved here

you’d put into your ~/.config/vkdt/config.rc one of the two options:

strqvk/device_name:null
intqvk/device_id:-1

where you replace either “null” by the device name that is echoed by vkdt if you run it via vkdt -d qvk or set the device id to the number that is in the same output and corresponds to the correct gpu (the one with the cable to the monitor).

[hanatos (jo)]

@123sg

[gui] monitor [0] XWAYLAND0 at 0 0
[gui] vk extension required by GLFW:
[gui]   VK_KHR_surface
[gui]   VK_KHR_xcb_surface

this output sounds like you installed glfw-x11 but are running wayland. if that is the case try installing the wayland version of glfw instead of the x11 version. it may be a good idea to start vkdt -d qvk too to see whether it picks a good device (since the lavapipe seems to be in the mix here as well).

[hanatos (jo)]

@betazoid

Apparently, noise-profile.sh is not included in the Debian package of vkdt. I will try to just copy the script from the github repo, place it in /usr/bin

i think /usr/bin/vkdt is just a symlink to the actual executable, in something like /usr/lib/vkdt or /usr/libexec/vkdt, depending on distro. you want to follow the symlink and place the shell script there. also double check whether in this directory you have

$ ls data/*cfg
data/noisecheck.cfg  data/noiseprofile.cfg

[hanatos (jo)]

@ggbutcher

libmad is the MPEG Audio Decoder, probably used by ffmpeg or asound

right. mad/vorbis/that stuff is actually just used by quake… so it’s like very optional for photography

[hanatos (jo)]

@priort

Package mad was not found in the pkg-config search path.

yes this should go away if you don’t build quake.

@priort

glslangValidator -Ipipe/modules -I./ --target-env vulkan1.2 -V pipe/modules/ab/main.comp -o pipe/modules/ab/main.comp.spv
make[2]: glslangValidator: No such file or directory

this one here is fatal. no idea about windows… maybe it needs installation, maybe it’s called glslc? if it comes by a different name, you can overwrite it in bin/config.mk as GLSLC=glslangValidator (replace by actual name here).

[123sg (Steven)]

@hanatos

installed glfw-x11 but are running wayland. if that is the case try installing the wayland version of glfw instead of the x11 version. it may be a good idea to start vkdt -d qvk too to see whether it picks a good device (since the lavapipe seems to be in the mix here as well).

Thanks! I think I might wait and see how @priort gets on… although I’ll probably have a go tomorrow sometime.

[difrkaguilar (Franklin Aguilar Matos)]

Thanks a lot, I’ll try this ASAP and of course I’ll give feedback

[difrkaguilar (Franklin Aguilar Matos)]

It’s working now.

With this config.

intgui/lod:1
fltgui/dpiscale:1
strsnd/alsa/pcm:default
intgui/export/wd:6000
intgui/export/ht:4000
intgui/export/format:0
fltgui/export/quality:90
strgui/export/basename:/home/franklin/Pictures/export/test/
strqvk/device_name:NVIDIA GeForce GTX 1080 Ti
intgui/ruc_num:2
strgui/ruc_entry00:/run/media/franklin/My Book/PHOTOGRAPHIC_DB/2022-05-01
strgui/ruc_entry01:/run/media/franklin/My Book/PHOTOGRAPHIC/TEST IMAGE
intqvk/device_id:-1
strgui/metadata/command:/usr/bin/exiftool -l -createdate -aperture -shutterspeed -iso

[hanatos (jo)]

great. is that a cr3 with old exiv2 or did you compile completely without exiv2? wondering why it doesn’t show exif information under the histogram.

[betazoid (Anna)]

Could we maybe create a pool where people can upload their raw samples for creating noise profiles? In theory I can create a directory in my Nextcloud. But maybe there is a better solution (I can delete the raws if I or someone else has created the profiles but storage is not unlimited there and I guess there are lots of cameras)?
Btw, it would be handy if noise-profile.sh could process several raws in a directory at once. Right now it’s a lot of copying and pasting… or is that already possible?

[hanatos (jo)]

@betazoid

Btw, it would be handy if noise-profile.sh could process several raws in a directory at once.

hm, for i in $(ls -1 *.CR2); do ./noise-profile.sh $i; done ?

not sure we’d need the raw input samples? once the fit is good we’ll only need the values, maybe the rawhistogram as proof that it’s good. but as mentioned before, you can pretty much recompute the raw histogram and fit for any image of that camera and iso combination… in most cases fitting should be possible, and in almost every image you should be able to judge whether the fit is good or not.

i think i should refactor the file layout soon though… all those individual files are unwieldy. would be good to download just one or two files for all support for a specific camera model.

[betazoid (Anna)]

@hanatos

hm, for i in $(ls -1 *.CR2); do ./noise-profile.sh $i; done ?

In this case, do the raws have to be in the same directory as noise-profile.sh or can I just put the directory path before *.CR2?

[betazoid (Anna)]

@hanatos

not sure we’d need the raw input samples?

I meant maybe people would provide raw samples even though they don’t want to use vkdt or are too lazy to execute noise-profile.sh themselves.

[hanatos (jo)]

ah. that makes sense, yes.