Add Polygon second pass: "predictive snap points" and numerical input to finetune distance and direction of the current point position in respect to previous.

I need to revamp my modeling operators, especially those that require user interaction. Here is first attempt at the Add Polygon operator. It's a first pass; it needs some work, but I feel it's not a bad start.

Alright, I made some further improvements to the tree view, which now draws in constant time (in respect to the scene complexity). Here is a test with 16 million transform nodes viewable in the scroll area. Without screen recording, the whole app renders at 8kHz.

I played a bit with the UX of interacting with the tree view. The common operations are selection and re-parent. I want to control both with the same gesture. So, if the drag&drop is mostly vertical -> reparent. If it widens to the side, it becomes marquee selection. Yes, no?

I am not sure why I am doing this, but I optimized further the scene tree view as it was getting a bit slow and clunky to operate when I had many nodes expanded. It's not cheap, it's still the most expensive part of the program to draw.

A long time ago in a galaxy far away (circa 2011) I came up with a distant rect light for my lighting crew. I can't remember how I did it, but I am trying to recreate it. Here is a test for the light sample distribution in the hemispherical domain.

After how many years? I added code for distant area lights. Some call it "directional". The control "scale compensation" is meant to preserve the light power (radiant flux), so as the emissive area increases, the intensity decreases.

I cut down startup time by deferring DLSS initialization. The first second of images rendered are noisy, until DLSS init is complete, at which point I accept a one off 40ms hiccup to allocate buffer resources and begin denoising. I'd say it's decent enough!

DLSS in workbench. The quality of 1spp rendering is way higher compared to the other built-in denoisers, and for half the cost. The comparison changes at higher spps.

So good! youtu.be/ldxFjLJ3rVY?...

I am trying to add DLSS-RR mode to Workbench. In theory I already have most things from the renderer, exception made for the specific pixel jitter and the motion vectors… It shouldn’t be too hard for me to add either of those. I am curios to compare DLSS with the other denoisers

Now is when the future becomes history

Last weekend I refined the scene tree view in Workbench. I used the drag & drop support in dear imgui to implement DAG reordering and reparenting, plus renaming. Custom behavior and drop region draw to show where the node would be inserted, and display where and why it cannot.

100,000 fuzz iterations. I'd say this part of the system works :)

Oh boy, I am letting Cursor run automated mutation testing in Workbench... 🍿

I have just spent 7 hours straight, barely leaving my desk looking at the wrong problem. I thought I was lost in reindexing hell, but the complex data structure I just made works just fine, the problem was in the incomplete input data all along. How is your day?

Finally, I came around at implementing DOF and wiring it to the camera manipulator system. Click and drag to samples the scene to estimate the focus distance, while scroll to modulate the aperture.

As we speak about unit testing, this is an example of what my unit tests look like. In particular the assertion messages are AI autocompletions. So, I didn't have to type much. Personally, I find this much more appealing than what I get with macro-based systems.

Friends out there. I have a question related to the use of unit test frameworks. Most frameworks for C/C++ are macros based. Do you like using macros for the purpose? Or do you think macros make your life harder, you wish for another way?

Eventually I did a good cleanup on my unit testing framework. I added support for JUnit xml export (in case I ever want to setup a CI process, even though it would be nonsense for a personal project 😅). But the cool thing is that it produces a complete log and xml export even in case of crash.

Crossed off the procrastination list: added support for multiple cameras in the scene. Up to now, I couldn't create cameras from the GUI, couldn't select them for rendering either. It's funny how something like this takes 20 minutes to do but can stay on the list for years.

I am back to it. I decided to revamp my modeling operators to carve out a half a decent modeler. There is a lot of cool manipulators options to be explored there. I started on chamfer... clearly, it's going to take a while to get it right.

I am a social person, I love more working shoulder to shoulder and exchange ideas throughout the day than the reality that I work from home, isolated by time zones. The most unexpected thing to me about AI tools, such as Cursor, is the positive impact on this.

Earlier this year I spent a few weekends toying around the math of simulating finance and investments. The only way for me to understand was to write a program that would put the concept into practice. So, I have implemented a Montecarlo simulator that projects family finances.

OMG! My iPhone just updated to the new version with the “Liquid Glass” design. OMG, what were they thinking????????? Of course I saw some of it when it was announced, and I was skeptical… but I didn’t realize it was this bad.

I reopened my project for the first time in 6 months... I barely remember which side is up. I told myself, "Do the first thing that cross your mind": If there is geometry under a light, and its cameraGain is set to zero, the light and the geometry become invisible to camera rays.

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.