next »

[JobLeonard]

Barry Sonnenfeld in Evil Dead II vibes

[Prinsessa]

  so that's how to make a horror game then?

Got some new ones...

This happened trying to fix some incorrect matrices:



This isn't really a fail as it was intentional to clearly see that my snow shader was working by making it green and thick and in every direction and on everything, but still amusing:

[Ordnas]

You should add it in the final game as a new parallel dimension 

[Prinsessa]

 

[FrankForce]

When I fooled myself into thinking there was a real life bug on my screen...

[Kyzrati]

When I fooled myself into thinking there was a real life bug on my screen...

Dude seeing that gif before reading the comment already had me thinking there was something on my screen, too! xD

[Prinsessa]

So what caused this?

[fluffrabbit]

Trying to fade up. My modulation of both the bloom threshold and amount apparently produced a very steep curve that ended up in the negative once it passed 1.0.

Code:

fadeFactor += d;
thresh = bloomThreshold * ( 1.0f - fadeFactor );
amount = bloomAmount / ( 1.0f - fadeFactor );

[JobLeonard]

Might still be a cool effect, but be careful with how different platforms might glitch in a different matter again

[fluffrabbit]

GLES2 is generally consistent everywhere. The bug is in my math. It might be okay for a lightning effect I suppose.

[Prinsessa]

3D animation system is truly the gift that keeps on giving when it comes to visual bugs…

[ThemsAllTook]

Something has gone wrong with one of my lighting shaders, and it's producing some pretty wild results. I have no idea what's going on here:

[Prinsessa]

Looks like a solid art direction to me!

[Schrompf]

Trying to unwrap every surface of my blocky dungeon to a unique texture place and render content on it. So far, my normals seem to be lost in an LSD trip.

(Click for full size)

[JobLeonard]

I have a soft spot for funky effects due to (integer) overflow, and this looks like it might be a case of one

[Schrompf]

Solved it in the meantime. It was like five errors in a row, and one of them was indeed an integer overflow while encoding a normal with >1.0f length into R8G8B8

[JobLeonard]

Good job figuring out all five of the errors - we all know how that struggle feels!

[Prinsessa]

So what does it look like fixed?

[Schrompf]

In an epsilon environment of "fixed". MipMaps still not present, apparently DX9 GenerateSubMipLevels() doesn't do anything, need to roll my own. Also Texture::LockSubRegion() triggers a GPU sync, equal what flags I pass to it, so I'll have to write it to some GPU scratch memory and copy it over with a simple shader. Or I take the bullet and finally convert my whole stuff to Vulkan.

And the Bilinear sampler pulls in neighbouring texels which haven't been written to, so you get weird artefacts on any edge. I'll probably copy over the "limit coords to inner safe region" code I wrote for my last game. The mippier the sampling, the farther in I limit every tex sampling so that the filters never catch values from outside the valid tex region. Worked great for a 2D sprite/tile based renderer, should work nicely here, too.

(Click for full size)

[JobLeonard]

That must have been satisfying the first time it rendered onto the screen correctly