Neil Bickford @neilbickford@mastodon.gamedev.place

You can also draw the Petersen graph with 3-way symmetry. Here's what the same animation looks like if you draw the graph that way:

Why S5? Suppose you take 10 vertices and label each node with 2 distinct integers from 1 to 5 and join vertices with no common number. You get the Petersen graph! Can then label each edge with the number that its two vertices lack. Elements of S5 map to automorphisms by permuting numbers; map from automorphisms to S5 by edges. Then have to show this is a bijection.

(More math than gamedev, posting here for completeness)

The Petersen graph has 120 automorphisms (ways to rearrange the nodes of the graph to get an identical graph shape). Here's a looping GIF of all 120 of them!

The group of automorphisms turns out to be S5, the group of permutations of a 5-element set.

I did post audio mixing for 10 concerts at MAGWest this year! Here are the ones I mixed:
Mega Ran and Super Soul Bros.: https://youtu.be/2H11LruWgxo
8-Bit Jazz Heroes: https://youtu.be/SN4MJlQSCsk
NyteXing et al.: https://youtu.be/LGBf3_IaDQg
Nikola Whallon: https://youtu.be/cS6-Lvb8K2Q
Petriform: https://youtu.be/d6L4tILEWP0
chibi-tech: https://youtu.be/ZREflAiikr4
Infinite Combo: https://youtu.be/oxMpIuwljp4
Sergio and the Holograms: https://youtu.be/vYI7j-y_OGg
minibosses: https://youtu.be/ZjH6MMq3BCM
SkyBlew: https://youtu.be/iN-0uyREDI4

SPIR-mint

It looks like last year's Gathering for Gardner exchange book is now online, so I can finally release the paper I wrote for it!

It's about how unit quaternions work, and why any rotation is represented by two quaternions (spoiler: the space of rotations isn't simply connected, so you can't blend paths of rotations without this duplication!): https://www.dropbox.com/s/mxwfrgw6pv9vihj/G4G13%20-%20Why%20do%20the%20Unit%20Quaternions%20Double-Cover%20the%20Space%20of%20Rotations.pdf?dl=0

Also includes 9 quaternion recipes, including why qxq^-1 rotates a vector, Frisvad-style tangent space generation, and more.

The team also released Korgi today, which allows you to modify Q2 console variables (like sun direction!) using a MIDI controller (somewhat like PANORAMICAL's MIDI interface)!

You can find the code for Korgi at https://github.com/NVIDIA/korgi.

Quake II RTX is out today! I helped build the tone mapper for this: https://www.nvidia.com/en-gb/geforce/campaigns/quake-II-rtx/

I'm still amazed by the team that's made this—from the q2vkpt demo, to PBR, volumetrics, procgen skies, and more!

But there's more—all of the code is open-source! https://github.com/NVIDIA/Q2RTX

For instance, you can see how Q2RTX's path tracing shaders work at https://github.com/NVIDIA/Q2RTX/blob/master/src/refresh/vkpt/shader/path_tracer.h

Or you can see some of the new tone mapping techniques at https://github.com/NVIDIA/Q2RTX/blob/master/src/refresh/vkpt/shader/tone_mapping_histogram.comp —or build it and write your own RT shaders!

I’m excited to announce I’ll be joining the NVIDIA Developer Technology team in Santa Clara starting April to help people create the next wave of professional graphics!
It’s going to be amazing to be working on GPU technology back in the Bay Area, and I can’t wait to get started!

Global Game Jam, day 3: We presented! Although we overscoped a bit, it was super fun to get to build this world with the rest of the team!

Global Game Jam, day 2: Snow shaders, more texturing, and rigging!

Global Game Jam, day 1: Modeled and painted a phonebooth!

The fourth edition of Real-Time Rendering just arrived at the office this week! (Pictured here with coffee, not included)

After a bit of work, we now have a Shadertoy-style renderer which handles internal reflection and refraction for the glass and the water!

I was originally planning on adding caustics (more likely using Guardado and Sánchez-Crespo's method from GPU Gems 1 than by projecting a grid from the light to the ground) and dust (just visualizing how the water particles move), but this is probably enough for now!

Finally finished debugging compute shaders - the fluid sim now runs (almost*) entirely on the GPU! Right now, we're getting just over 30 frames per second on my laptop, running on a 64^3 grid with ~9.53*10^5 particles. (For some reason, it seems to be taking more time transferring particles to the grid than solving the pressure equations, which is surprising.)

Next: Moving away from the debug visualization and implementing voxel-based raymarching!

*Prefix sums not included.

There we go!

Favorite activities of fluid sims:

- flow (ok)
- get stuck between grid cells (??)
- disappear into the lower-left corner (???)

However, gravity and the ghost fluid method are both implemented now!