This discussion has helped me understand how challenging it is for darktable developers to manage highlights and shadows and tonal range without increasing saturation. Other tools I have used include C1, Luminar Neo, Pixelmator Pro, Affinity Photo, DXO Photo Lab, and On1 Photo Raw - and all of them change highlights and shadow with changes in saturation. I used to over-saturate my photos and now I can see the beauty in more natural presentations, or with select use of saturation for effects.
3 Likes
Lightroom does basically the same thing with its default settings. there is a video on this, and if you would like to see it, I will Reply with the link.
I have over 120,000 images in my darktable library and they are now searchable with tags and a variety of other values and what amazes me is the speed of access. C1 was never able to hold the 2TB of photos I have with easy of indexing and search speed. That this software is also FOSS and works on Apple Silicon is fantastic.
2 Likes
I remember how zippy ACR on an old 4core/4threads Phenom felt back when I still used that.
Do you have links to read up on this? Iād be curious about the complexity of that cache structure.
IIRC, I heard some comments on a podcast recently.
Possibly it was in āAlgorithms + Data Structures = Programs: Episode 160: Rust & Safety at Adobe with Sean Parentā? But it wasnāt much. It was just a few offhand comments by an involved programmer, really.
2 Likes
I too am very experienced and comfortable working with filmic. However, trying to teach people new to DT how filmic works is very challenging and dare I say a barrier to many accepting DT. However, Sigmoid just seems to work for most images that I have tried without input from me. Some images such as sunsets definitely require input. One of the big differences with sigmoid is that it produces more colorful images straight out of the box while filmic requires the additional use of the color balance RGB module to add colorfulness. Both filmic and sigmoid can work well but I feel sigmoid is easier for a person new to DT. I use it for 99.9% of my images.
Since you are happy and knowledgeable in the use of filmic I agree that there is no point in wasting time learning sigmoid. 
1 Like
Yep, thanks for this. From 12m20s on.
Preview updates depending on how fast the slider is moved. The faster you move the lower the LOD that is calculated and presented.
2 Likes
And IIRC he said something about an intricate thread eviction logic to keep latency imperceptible.
Iād love to hear or read more about the inner workings of such pieces of software. Itās fascinating stuff. Thatās part of what I love about this community: Technical details are discussed openly.
2 Likes
While googleing about him, Sean Parent, I found talks on YT where he goes a bit more into detail on why this is important for Adobe (Memory safety, company liability of writing unsafe code).
The gist of it (what I understood of it, me being terrible at programming) is about C++ being not āmemory safeā out of the box, the senders&receivers being an attempt at managing this and how this specific implementation is not appropriate for what adobe does within their pipeline. As to why: because the eviction cannot wait until a task is finished, it has to free up processing power before a task is finished, which in sender&receiver is not possible.
Apparently they wrote their own library for this, which to me seems under a rather permissive license (Boost software licenense 1.0).
STLab-concurrency https://github.com/stlab/libraries
4 Likes