Over the month of October, my solar array produced 1.5 MWh, I used 915 kWh to power the house and charge my car. I drew only 3.7 kWh from the grid, exported back 570 kWh.

Unfortunately, modern rendering (i.e. path tracing, software rasterization, etc…) looks more like pure compute than traditional “graphics”, and implementing such complicated data structures and computation using graphics API feels exactly like GPU computing before CUDA.

iPhone 17 pro: orange is the new black 🤣

text compression

You remember you made a mistake, so you want avoid doing it again! Just this time you don’t recall exactly what you did wrong, so you end up messing up at the opposite end…

First rather naive implementation of Stochastic Light Cuts. Here comparing uniform sampling vs SLC with 1spp over 16,000 lights.

It’s been a long break away from my projects. Work deadlines + sickness + low energy slow recovery… I think I’ll try to make some progress on light transport, at least for a bit. Workbench doesn’t handle many lights (not gracefully at least).

Histogram Stratification for Spatio-Temporal Reservoir Sampling Corentin Salaün, Martin Bálint, Laurent Belcour, Eric Heitz, Gurprit Singh and Karol Myszkowski (Max Planck Institute for Informatics, Germany & Intel, France) iribis.github.io/publication/...

I really, really, really dislike year-based software version numbers.

The Vulkan porting is complete 🎉 The last missing piece was the primordial raster renderer, which I am not sure why I keep around :) All in all, it wasn't nearly as hard as I thought it would be.

The amount how pieces to track to share a texture between Vulkan and CUDA feels a bit over the top. Each of these need to be constructed and destroyed, otherwise GPU memory leaks... I wish we just had pointers and that's it.

Components selection and editing work again.

And boooom....

I am doing things I a rush lately… Hopefully a better structure will emerge from the mess and a rewrite will follow.

Some progress: the grid, selection, wireframe, light gizmos, reverse Z-buffer, path-tracer/raster compositing and a few more bits. It's all coming back together.

And... it's upside down 🤣

It's still a mess, but bit progress today, UI fully working. With an empty scene the program renders at 12kHz, which is ridiculous! This compared to 4kHz of the OpenGL implementation. To get there I had to hack through GLFW and dear imgui docking where some "expensive" windows system calls were made

Why is this texture taking so long to upload… I don’t get it… it’s just a 1k texture… Looking closer: size_t k_resolution = 1024; size_t k_channels = 1024; size_t size = k_resolution* k_resolution * k_channels; I shouldn’t write code past 8pm…

Small progress today: my GUI library system is partially working. Textures and picking are still missing though.

Our #GDC25 talk on RTX Mega Geometry and how it was used in UE5 and Alan Wake 2 is now available on YouTube: www.youtube.com/watch?v=Kblm...

The easy part of porting to VK the imgui-based portion of Workbench UI is done. Now to the hard part...

Rainbow cubes are back! It's likely most of you didn't know me when all my posts were rainbow cubes... As I am preparing to port Workbench to VK, I am back to basics.

What do you do on Easter break? ... f***

Also, this! The whole GUI draws at a ridiculous 4200 fps.

It's hard to find time and energy to work on my project lately. I haven't written a single line of code for it in months... But today I have upgraded my workstation from an old Intel 10th gen i9 to a Ryzen 9950X and how boy!.. If you thought Workbench opened quickly, wait until you see this:

My thoughts on Generative AI

When companies were increasing prices by 20-30%, blaming it on 5% inflation… how much do you think they’ll rise prices blaming it on 20-30% tariffs? Just asking for a friend.

Trying hard not to think about what is happening… the level of stupidity is hard to comprehend.

Recently I saw I video of the Boston Dynamics’ Atlas walking and running very smoothly and convincingly human. But, fluid motion is the result of what is energy efficient given weight distribution in the body. I wonder if making a humanoid robot moving like a human is energy efficient.

I have another take on “Vibe Coding”, and it doesn’t involve gen AI: Spend time coding on a project you enjoy, following a coding methodology and style that reflects you, experimenting on things you like, without overbearing process. That, is vibe.

First meeting 7:30am, last email 9:38pm

On closed source projects, code reviews shouldn’t be about correctness, they should be about conformance: is the new code dumb? Does it violate any design guideline? Is it documented? We are bad at spotting bugs by reading code. For that we should rely on tests, tools and benchmarks.

Model of cosmic time to scale youtu.be/nOVvEbH2GC0?...

I just realized that Apple AirPods Pro have a little known frequency calibration feature! Our ears begin to drop sensitivity to high end of the frequency spectrum as early as 8 years of age and it slowly and progressively degrades from there.

I saw this today: a guy wearing a t-shirt saying “make money, not friends”. Chances are it’s a failed strategy where you end up with no money and no friends.

New blog post: maxliani.wordpress.com/2025/03/01/m...

This is so not funny… I am one of those skilled immigrants who left Canada as i couldnt take that disfunctional imigration system… This is going to be a rant thread, please skip it unless you are considering moving to Canada and want to know what could happen to you…🧵 youtu.be/lapAm6v5Dgo?...

Go to clang-tidy configuration, add “-modernize-loop-convert”. You are welcome!

When I was an artist, I was cursing at the renderer, so I wrote my own renderer, for me and my lighting crew, and that swallowed me whole, sadly I pretty much stopped being a lighter. 12 years later, I curse at the language and the compiler…

Installed new drivers, upgraded to CUDA 12.8 and cuDNN 9.7. Now Workbench graphical session start and is fully functional in just 450ms. Process RSS at startup went from 570MB to 428MB.

Sometimes you cannot proceed in doing anything else because you see that a context menu is clipped two pixels at the bottom... and... you... have... to... fix... it...

I would never be able to make my mind for what the node graph should look like, and everyone has their rightful opinion of colors vs gradients. So, look controls FTW!! :D

Quick round of GUI polishing. Left is v1, right is v2. It's an improvement, I think... but I am still very much puzzled about what this should look like. I can't point my finger on what it is that I don't like about it. Maybe simply I don't like widgets in the nodes, I don't know... 🤔

So, I was thinking… with a small extra effort, I could make the node graph differentiable 🤔 The obvious use case would be to compute dPdt (aka motionblur). But in theory I could get it to compute multiple partial differentials over arbitrary input variables… I don’t know what I’d use it for though.

I quickly added a "time" node. Now I can animate!

On the one hand, nested loops work! 🎉 On the other, links still look like noodles.

Last night I had an idea for how to implement a for loop in my geometry nodes. I scribbled it in my notebook and today I gave it a go at implementing it. It introduces the concept of "blocks", jump "instructions" and loop-invariant code motion...

My first graph evaluation version was a recursive depth-first traversal which called nodes virtual function to execute. The new version I implemented today "compiles" the graph to a vector of function pointers to call. Excluding the geometry processing, this takes ~800ns.