# build time configuration.
# select your favourite bloat here.
# if you have freetype2 and png16, you can uncomment this:
# VKDT_USE_FREETYPE=1
# export VKDT_USE_FREETYPE
# if you disable this, the i-raw module will not be built
# and you will be unable to load raw images.
# since rawspeed is a submodule this is only useful if
# you want to avoid the recursive dependencies.
VKDT_USE_RAWSPEED=1
export VKDT_USE_RAWSPEED
# exiv2 can optionally be used to load some super basic metadata (iso speed,
# shutter time etc) inside the i-raw module. that is, you don't have that you
# don't need to disable this.
VKDT_USE_EXIV2=1
export VKDT_USE_EXIV2
CC=clang-11
CXX=clang++
GLSLC=glslangValidator
export CC CXX GLSLC
Not sure what I should do …
Edit : I’ll try to check what GLSL compiler I should install …
Edit2 : installed on my pc are: gcc 10.2, glslang-dev, glslang-tools, glslangValidator, libvulkan-dev, mesa-vulkan-drivers, libopengl-dev, libopengl0, libglfw3-dev, GNU make 4.3, pkgconfig
…i just don’t care about it. you will find quite a fair amount of missing features when playing with this. and jpg support is not even on my private priority list (exr a bit more, if it wasn’t for the bloaty support library. i mean even this tinyexr/tinyexr.h at release · syoyo/tinyexr · GitHub has ~15k loc). this is mostly a matter of implementing an i-jpg module, but also there’s some default handling of input filenames which will need to distinguish between raw and jpg and whatnot when creating thumbnails or putting default processing graphs in place.
@Carmelo_DrRaw I have just installed the git master Version of Photoflow - Photoflow already uses a “pure-gpu-pipeline”, doesn’t it? Have you integrated vkdt code?
I have noticed artifacts (bugs?) here and there - when will a stable version be released?
its the developer version from August 28 (git master)
If I remember correctly, @hanatos and @Carmelo_DrRaw had a long conversation here about integrating the engine of vkdt into Photoflow or something similar. I complied vkdt some time ago. Photoflow git master “feels” very similar to vkdt. I also ran Photoflow with and without my Nvidia card and there is a significant difference. To me, Photoflow feels very pure GPU. But I don’t know, maybe I am wrong.
yes, had good discussions about library design and algorithms. this is probably why results feel similar to you? we might have similar ideas about how to use which algorithm now.
however, we did not work on code together yet. partly because i’m terrible at productising vkdt’s core: i’m still introducing breaking changes, deleting and adding parameters, figuring out how to best draw/blend/mask/align/etc. i think i’m converging on many of these things, but there is no “library” version of vkdt’s processing graph. also when using vkdt’s processing in a qt application you’ll introduce quite a bit of bloat and lag. the gui layer i’m using just directly draws the texture from the descriptor set the graph outputs. not willing to compromise speed at this point
okay 40 million bytes are no 40 megabytes after all. also you make a point that at least the copy to screen should be possible without wasting too many cycles, even using a bloaty monster.
not sure i want to continue the file size discussion, but it seems it is indeed true what you say:
$ ls -lha Filmulator_v0.9rc13.AppImage
-rwxr-xr-x 1 jo jo 39M Sep 17 12:22 Filmulator_v0.9rc13.AppImage
and yeah qt comes with many different sub-libraries, probably you don’t need all of them (and probably i haven’t installed all of them either).
smoother highlight reconstruction (not entirely happy, may want to get back to this)
raw colour matrix fuzzy maker/model matching
smoother guided filter results (discovered there’s a lot of literature on fast blurs)
usability/ui
sort by create date (via dirty hack instead of exiv2)
list recently used tags, click to switch collection
jump back to original collection button
“open directory” button so vkdt can start without cmdline args
crop with fixed aspect ratio
other
binary packaging: generic build flag option, vkdt runs independent of pwd
bugfixes (infinite busy loops, /0, etc)
bash script to create static webpage from markdown docs (https://jo.dreggn.org/vkdt/). planning to put all the large screenshots/support files for the web site into this separate repo, so the main repo will remain small
glad to hear it’s working.
because unlike yours all the integrated one is