next »

[oahda]

You have to use the full youtube.com link for the embed to work. Cool! What's that constantly flickering over the screen tho?

Whoops, edited, thanks.

If you mean the cloth sometimes just jumping off, that's what I meant with the jumping - hoping that's due to the integration not being able to cope with the picking / turning sphere movement on and off. Although I should probably look into the latter again.

Huh… Last time I watched there was a different program flickering over your render window, so I thought your recording was borked, but I watched again now and it's gone. So the issue was on my end.

[JWki]

You have to use the full youtube.com link for the embed to work. Cool! What's that constantly flickering over the screen tho?

Whoops, edited, thanks.

If you mean the cloth sometimes just jumping off, that's what I meant with the jumping - hoping that's due to the integration not being able to cope with the picking / turning sphere movement on and off. Although I should probably look into the latter again.

Huh… Last time I watched there was a different program flickering over your render window, so I thought your recording was borked, but I watched again now and it's gone. So the issue was on my end.

Hmmm hahaha that's weird! Maybe youtube fucked up somehow, who knows.
Anyhow seeing how simple this was to get up and running in a somewhat acceptable fashion, already looking forward to outfit all my renderer test scenes from now on with banners and shit.

[oahda]

Is it very (CPU? GPU?) intensive? Also what's that environment? The UI doesn't look like ImGui to me.

[JWki]

Is it very (CPU? GPU?) intensive? Also what's that environment? The UI doesn't look like ImGui to me.

Runs on the CPU and at that resolution is not too intensive I don't think, I haven't done any profiling. This is an uni assignment so using the framework they provide (the UI is AntTweakBar, rendering is OpenGL).
AntTweakBar seems to be pretty popular in academics, I was provided with a similiar framework for a different class for a more long-term project (fluid simulation) but we threw away the framework for that one and replaced it with custom GL and dear imgui haha.

[oahda]

Sounds like a cool university curriculum/course! What's it called?

[JWki]

Sounds like a cool university curriculum/course! What's it called?

This one is called real time graphics, mainly sets up a bit of architectural knowledge (common design patterns in games, pretty basic stuff like main loops, etc), some basic physics simulation stuff like this (and rigidbodies, deformables) and then should delve into some rendering things using OpenGL basically, covering how PBR, different light indexing algorithms etc work in practice.

[popawheelie]

Always like cool cloth sims.  

Just out of interest, do they teach real time destruction? And if they do, to what extent?

[JWki]

Always like cool cloth sims.  

Just out of interest, do they teach real time destruction? And if they do, to what extent?

They teach the theory for that in a different set of classes but not in this one specifically.

[ferreiradaselva]

Did my first basic cloth sim for an assignment in uni, I think I got it to work as well as intended within the given framework.
There's some extreme jumps in the video that I'm just going to attribute to the super basic explicit euler integration that was already provided so none of my business.

Woah... you know those satisfying videos... that.

Edit:

The implementation of the X11 backend in my window framework is almost done. Most window operations are done and keyboard and mouse input are working. Wayland still needs some work on input and some window operations. Both have already OpenGL support.

Something that occurred to me. Not every game developer, especially small developers, add support for game controller or use tablet input, so I thought I should make these disableable during compilation-time. Using some macros, yknow (SWFW_NO_GAME_CONTROLLER, SWFW_NO_TABLET).

[ThemsAllTook]

A followup to the animation I posted a bit ago... I added a function to my armature code that can generate a debug mesh to show all of the bones on their own. Should be useful for debugging animations and seeing the structure underneath. In my test application, I can toggle between this and the skinned representation, and they both animate the same way:



This is going to be pretty cool once I get model loading implemented!

[oahda]

Do ask if you plan on using glTF and have any troubles since I've gotten that working lately, including animations.

[ferreiradaselva]

Btw, is the pose processed on GPU or client-side?

[popawheelie]

I could use a general purpose TIFF read/writer that has no dependencies. I'm using libTIFF, but it unfortunately has an unload issue that's been causing headaches.

[ThemsAllTook]

Do ask if you plan on using glTF and have any troubles since I've gotten that working lately, including animations.

Cool, will do. I've read through the glTF spec, and while I think it should be possible to get it to work for what I'm doing, it does seem pretty hefty. It's tempting to just skip it and write my own Blender exporter for a custom format, but I'll at least see how far I can get with glTF first.

Btw, is the pose processed on GPU or client-side?

This was a fun one to figure out. The pose is done on the CPU, and the skinning is done in the vertex shader. I have an Armature class that stores the bones in their default state; an Animation class that stores keyframes, intervals, and curves; and an AnimationState class that represents the intersection of the two, with a computed frame of a pose. When drawing, I call a function in Animation that calculates a matrix for each bone in the mesh's AnimationState, then I pass those into the vertex shader as a mat4 uniform array. My mesh data comes with vertex attributes for bone ID (which is just an index into the matrix array), and bone weight, up to 4 per vertex.

The process of computing a pose for a given frame has a lot more complicated logic than I would have expected. Doing it on the GPU would be a challenge, though certainly possible if I restructured my data a bit. I'm not planning on drawing hundreds of animated characters at once, so it feels like it should be efficient enough for my needs as it is. I guess I'll find out once I start drawing more complicated models.

[oahda]

Do ask if you plan on using glTF and have any troubles since I've gotten that working lately, including animations.

Cool, will do. I've read through the glTF spec, and while I think it should be possible to get it to work for what I'm doing, it does seem pretty hefty. It's tempting to just skip it and write my own Blender exporter for a custom format, but I'll at least see how far I can get with glTF first.

Cool! Realise I didn't specify so in case it was ambiguous: glTF 2.0.

[ferreiradaselva]

This was a fun one to figure out. The pose is done on the CPU, and the skinning is done in the vertex shader. I have an Armature class that stores the bones in their default state; an Animation class that stores keyframes, intervals, and curves; and an AnimationState class that represents the intersection of the two, with a computed frame of a pose. When drawing, I call a function in Animation that calculates a matrix for each bone in the mesh's AnimationState, then I pass those into the vertex shader as a mat4 uniform array. My mesh data comes with vertex attributes for bone ID (which is just an index into the matrix array), and bone weight, up to 4 per vertex.

The process of computing a pose for a given frame has a lot more complicated logic than I would have expected. Doing it on the GPU would be a challenge, though certainly possible if I restructured my data a bit. I'm not planning on drawing hundreds of animated characters at once, so it feels like it should be efficient enough for my needs as it is. I guess I'll find out once I start drawing more complicated models.

I've seen people making on GPU-side, but I would do the same way you are doing. Simple and it works. I suspect doing it on GPU-side would require passing too many data to the GPU.

[qMopey]

I could use a general purpose TIFF read/writer that has no dependencies. While SDL can do this, it depends on libTIFF. Unfortunately that dll won't always unload causing headaches.

Why tiff? Last AAA game I worked on used a million tiffs, but I couldn't figure out why that file format was chosen over others. Any insights?

[qMopey]

Did my first basic cloth sim for an assignment in uni, I think I got it to work as well as intended within the given framework.
There's some extreme jumps in the video that I'm just going to attribute to the super basic explicit euler integration that was already provided so none of my business.

Pretty nice effect! What kind of cloth solver are you using? Looks like you might be using Jakobsen style solver: https://www.gamasutra.com/view/feature/131313/advanced_character_physics.php

I see some stretching on the corner vertices. What integration is this? Symplectic euler?

Nice ant tweak bar there A lot of older uni courses use ant tweak bar. For a while it was sort of the "go to" for getting in debug UI really quickly, before Dear imgui was a thing.

[ferreiradaselva]

Afaik, TIFF supports subimages (you can obviously see why this part is very attractive for game developers) and you can add tags. I've never used, tho (always felt like an obscure format to me).

[popawheelie]

I could use a general purpose TIFF read/writer that has no dependencies. While SDL can do this, it depends on libTIFF. Unfortunately that dll won't always unload causing headaches.

Why tiff? Last AAA game I worked on used a million tiffs, but I couldn't figure out why that file format was chosen over others. Any insights?

No reason really. PNG would be fine too. I need a format that can handle 16 bits per channel, and LibTIFF is capable of that. It just has that damn unload issue.