mosra/magnum

Lightweight and modular C++11/C++14 graphics middleware for games and dataviz

Vladimír Vondruš

@mosra

clang tidy is a bit overmotivated, i'd say

the relevant GL API takes a (signed) int, and if i'd make setCount(std::size_t), then it would complain if passing an int to it again

Guillaume Jacquemin

@williamjcm

Today, I learnt Magnum types such as Color4 or Vector3 were trivially copyable, and thus made one of my struct types trivially copyable as well.

Fo Nz

@FoNz99089892_twitter

Today, I learnt Magnum types such as Color4 or Vector3 were trivially copyable, and thus made one of my struct types trivially copyable as well.

As I saw, basically they have a data method where it returns the base address for the internal array. I think you can just memcpy them.

Jordan Peck

@Auburn

@mosra I've updated my Nvidia drivers to latest version on Windows and I'm now having an issue with ImGui not correctly handling alpha

ImGui is rendering a fullscreen quad with this color data

I'm sure it is caused by the driver update, since a user was having the issue on the latest driver. So I updated then I had the same issue

Fo Nz

@FoNz99089892_twitter

I'm trying to use AnySceneImporter on iOS, but when it's loaded by the PluginManager, the application says that the plugin is not present. I can see the library being statically linked into my Xcode project. I all the dependencies by myself. What's the error? Any idea? Thanks in advance.

Jordan Peck

@Auburn

I found a fix by changing ImGui window flags, seems odd though that a driver update would cause the behaviour to change? I would have thought OpenGL doesn't get touched much these days

Vladimír Vondruš

@mosra

@Auburn what was the flag that fixed it? all i can think of is the new driver not clearing framebuffer attachments by default (or vice versa) so you end up with something different if you forgot to clear explicitly

@FoNz99089892_twitter if you're not using CMake to link the static plugins, you have to perform an additional step to register them for the plugin manager, basically the following in a file that's compiled directly to the final executable (and not any intermediate static library):

#include <MagnumPlugins/AnySceneImporter/importStaticPlugin.cpp>

Relevant docs: https://doc.magnum.graphics/magnum/plugins.html#plugins-static

Fo Nz

@FoNz99089892_twitter

@mosra Thanks a lot! I completely missed that.

Jordan Peck

@Auburn

@mosra It was this flag for a dock node, which in turn sets the fullscreen imgui window flag ImGuiWindowFlags_NoBackground https://github.com/Auburn/FastNoise2/commit/732577d86dbdf7f79de949b86c3ef3589ba82699#diff-13f4704b60755926950ac5d2b0ca1e4850f8a69d0a51ebc2cce9418b8dfefa95L550

if (dockspace_flags & ImGuiDockNodeFlags_PassthruCentralNode)
        host_window_flags |= ImGuiWindowFlags_NoBackground;

Which makes me wonder how it was working before...

Jordan Peck

@Auburn

Linux and MacOS are fine without the fix, but it seems like that behaviour is wrong

Jordan Peck

@Auburn

With the fix

Without

My only guess is that there was something in driver hiding the fullscreen quad previously

Vladimír Vondruš

@mosra

Which makes me wonder how it was working before...

huh, yeah exactly

is it possible that the fullscreen quad got clipped away for some reason? though i don't see why the quad would but not the rest of imgui

Jordan Peck

@Auburn

The fullscreen quad is a fullscreen imgui window

Vladimír Vondruš

@mosra

crazy, but .. is it possible that you hit some nvidia game-specific workaround before? :D

does anything change if you rename the broken exectuable?

Jordan Peck

@Auburn

That doesn't explain why the issue doesn't exist on Linux though?

Vladimír Vondruš

@mosra

i mean, certain drivers are known to check if the executable is named rage.exe and then applying certain shader patches etc

so you might be hitting something similar, and that it works on linux is maybe just because there's no .exe

but that's a crazy idea, i don't have anything better, sorry :)

Vladimír Vondruš

@mosra

https://github.com/Auburn/FastNoise2

huh, i wonder how did i miss your project, or did i just forget it existed? .. i remember we were discussing something related to noise and SIMD detection but don't remember seeing this fancy tool :sweat_smile:

(can you add the #magnum topic to the repo description? unless you want to keep it a secret, of course)

Jordan Peck

@Auburn

Ya sure

Just tried testing on an another PC that still had the older driver and the pre-fix builds are broken now there too

Some weird driver magic

Vladimír Vondruš

@mosra

:eyes:

renaming the exe does nothing, then?

Jordan Peck

@Auburn

Well that's what I was going to test, but it's already "correct" (broken)

Vladimír Vondruš

@mosra

i downloaded the latest release, added a checkerboard node and it's also blank (linux, mesa)

or do i need to do something special to make it show? it showed some voxel counts and got considerably slower so it looks like it's actually doing something

Jordan Peck

@Auburn

That should be it

Maybe I'm just remembering wrong but I thought I had tested it on Linux with the dock node change

Jordan Peck

@Auburn

Here is the latest release with the fix, https://github.com/Auburn/FastNoise2/releases/tag/v0.9.4-alpha

Vladimír Vondruš

@mosra

oh yeah this works :)

Jordan Peck

@Auburn

You can add FastNoise2 to the featured projects list if you think it's exciting enough :D

Vladimír Vondruš

@mosra

yep, i just added a note to do that :)

Jordan Peck

@Auburn

How are your SIMD adventures going?

Vladimír Vondruš

@mosra

the runtime detection and dispatch code is done, but stashed in a branch until i have a real code using it and proving it works

want to avoid making something available and then later redoing / deprecating most of it because it sucks

Where communities thrive

People

Repo info

Activity