sjb3d

sjb3d, 13 days ago to random

Back to experimenting with meshing algorithms, trying to add variable resolution support while keeping the output manifold.

Left image is just directly meshing SDF samples in an octree (complete with red non-manifold edges where resolution changes). Right is a dual method, edges are a mess so it doesn't look manifold, but it is!

sjb3d, 20 days ago to random

Desktop Vulkan question: is it reasonable to rely on resizable BAR being enabled these days? Just filling write-combined buffer memory directly, without all the code for sync/async copies from staging buffers, is super nice.

dee, 29 days ago to random

My experience with the Zig community so far

sjb3d, 1 month ago to zig

Found a small #Zig issue with EnumSet, so raised an issue yesterday with a fix PR. One round of PR feedback to fix a test, and it's merged. ⚡ 10/10 would PR again.

sjb3d, 1 month ago to random

Ooh the HLSL proposal for vk::BufferPointer looks like a great upgrade over vk::RawBufferLoad: https://github.com/microsoft/hlsl-specs/blob/main/proposals/0010-vk-buffer-ref.md

Will be nice to remove a level of indirection for bindless buffers (and get parity with what is possible in GLSL).

sjb3d, 1 month ago to random

Some nice use cases for Zig defer by matklad: https://matklad.github.io/2024/03/21/defer-patterns.html

Definitely seems useful to consider defer for use cases other than resource cleanup.

sjb3d, 2 months ago to random

OK dynamic rendering Vulkan is best Vulkan, never creating a render pass again.

sjb3d, 2 months ago to zig

The #zig experiments continue, still very much liking it.

Dropping random C/C++ files in the build and having it just work is great.

Only minor issue I hit so far is missing syntax to make a slice from a single-item pointer (like slice::from_ref in Rust), but looks like this is getting fixed for 0.12 (https://github.com/ziglang/zig/issues/16075). And can work around in 0.11 with some comptime trickery.

sjb3d, 3 months ago to random

Tinkering with Zig (https://ziglang.org/) and there's a lot to like about it. Allocators required throughout the standard library is great, I wish rust had gone this route.

And comptime seems super powerful for generics. For example, want SoA storage for any struct type? No problem, here's an example in the standard library: https://ziglang.org/documentation/master/std/#A;std:MultiArrayList.

sjb3d, 6 months ago to random

Just pushed an #egui Vulkan renderer as part of my #rust Vulkan wrapper here: https://github.com/sjb3d/spark/.

I've been enjoying using egui (https://github.com/emilk/egui) with glow (https://github.com/grovesNL/glow) for a while for hobby projects, but needed #Vulkan integration for the next one, thought I'd share since it might be useful for someone else too!

ephtracy, 6 months ago to random

ui wip #magicaCSG

sjb3d, 7 months ago to random

Second attempt at handling hard seams between SDF shapes, it's not totally robust yet but starting to work. Behold my glorious SDF artwork.

Left is marching cubes. Right is dual marching cubes on the same grid, still only solving for position at each dual vertex, but now giving each dual polygon a surface ID. (The dual case also uses a bit of edge collapse to clean up the meshing around edges.)

video/mp4

sjb3d, 8 months ago to random

One marching cubes detail I've not seen mentioned anywhere: the tunnel in case 10.1.2 can be oriented along either diagonal, which can be taken into account when adding interior vertices for the triangulation.

This is the only case with two orientations, all other cases have only a single possible diagonal for a tunnel.

Alternating between the two different tunnel orientiations for marching cubes case 10.1.2.