I have changed the license of the sokol headers (sokol-gfx and sokol-time so far) from MIT to zlib/libpng: github.com/floooh/sokol (cc @rlyeh )

My evening of learning UE4....well....it didn't go quite as well as I hoped. I feel like I'm just too comfortable with Unity to get comfortable quickly. I would be more productive just going straight to a basic lib that gives me a framebuffer.

RT @ocornut@twitter.com: dear imgui: visual studio users: if you frequently use imgui types or step in the code, I have added a .natvis file to improve the debugging experience (just add the file to your msvc project) github.com/ocornut/imgui/blob/

Alright. Time to learn how to use UE4 the night before ...I feel like this is not the most optimal time.

Hmm. I should probably work out why the Bistro scene looks a bit....not right. On the plus side having a much larger scene fed through my import pipeline did find some bugs! mastodon.gamedev.place/media/C

I really need to look at hooking up a debug backend to my engine. Not useful seeing it spinning waiting on resources without a good visualization of what is going on with the worker threads.

Due to how it gets packed into blocks the overall size is larger than before. Sitting at about 8k of data vs 30.28k. Lot of wasted space in the texture, also probably partly responsible for some of the artifacts. mastodon.gamedev.place/media/p

A little more work jiggling the data around. Built an octtree to get leaves with no more than 16 positions (4x4 blocks). Blocks compress a bit better, but still pretty naff. However it's an alright improvement I reckon. mastodon.gamedev.place/media/t

Show thread

Alright, I think this is a success so far. First pass on jamming geometry data into compressed textures. First image is baseline, Second image is compressed. mastodon.gamedev.place/media/2 mastodon.gamedev.place/media/L

Much more productive day then I expected! Fixed up some unit tests, added some code to help with comparing images (for units and later work with encoding), added added some additional bits to my math library + tests. I think I might have most of what I need to actually move onto the vertex data stuff! Or will I try and squeeze ispc_texcomp in first...hmmm.

Alright, got another yak shaved this week, knock up a small image lib for my engine (load/save/process/conversion/etc): github.com/neilogd/Engine/tree

Now that that's done, will probably move onto poking at what I was originally aiming to do - much about with packing vertex data into textures, and use an appropriate BCn format + mips to see how well, and fast, squeezing more vertex data onto the GPU is.

dear imgui: more r&d on virtual viewports + platform abstraction layer = magically dragging windows outside client area, creating display contexts dynamically etc. (there's still flickering on transitions).

The road toward turning all those r&d stuff (docking + magical platform windows) into a polished/shipped version is long but we're getting there. Probably half-way now?


I've wondered more about specifying frequency of change and inferring root signature/register allocation.

Show thread
Show older
Gamedev Mastodon

Mastodon server focused on game development and related topics.