P.O.P.S.- Demo 3.1 available!

[Argh]

OK, still not quite happy with fire / smoke yet. Got most of the kinks out, in terms of the most generic case, but obviously real fire needs better solutions than this. Still looking at several approaches to that.

Explosions using this run really fast when you're not looking at them, so I consider that goal(building a credible replacement for all CEGtag events) met.
I'll release another demo when I've had time to write some fire-specific behaviors, now that the general case works.


Oh, yeah, and I built management stuff and hooked it to projectile behaviors, so you're looking at a system that's basically doing this from Pieces that have fallen off my demo model, that part is done.

[Master-Athmos]

Different blending mode didn't change anything here - still no bubbles...

[Argh]

Great- no errors, but it still won't run, eh? That is frustrating.

I've fixed a few things in the shader while getting fire/smoke going, one of which may have caused the problem (I had a variable that was zero'd- this didn't cause a failure on nVidia, but it very well might have on ATi). One other thing that might be causing the problem is that it's using a DDS texture- I might revert back to TGA, it has a small speed impact but not that huge.

[Argh]

So, any luck, ATi folks? I'm almost done with fire and smoke, but I want to get this working on all hardware before releasing another demo...

[Master-Athmos]

Actually you just succeeded in breaking the fog as it's gone now but still no bubbles...
Apart from a problem with button9.wav I get no error messages though...

[Argh]

No, I just turned the fog off, I was trying to eliminate possible causes.

So, do you hear the bubble sounds?

[Master-Athmos]

Yeah I hear the bubbles and the particles seem to be there as they appear as red dots on radar:

radar bubbles.jpg (31.15 KiB) Viewed 1706 times

[Argh]

No, those are just the Projectiles that serve as spawn points (for the fire demo).

Here, try this version. It uses TGA instead of DDS. It's gotta be something related to either the texture, point objects in general (doubtful, the snow gadget has been reported working on ATi) or there's something else that's screwing up the results. Hrmm. Did you see the TexRect for the diagnostic, when that was still on?

[Master-Athmos]

Nope. I also tried it with a bmp but without any result...
I'd say you need to do an extremely simple colored particle spawner in order to test out at which point things go wrong...

[Argh]

Yeah, I'll disable everything, depth-test etc. and remove the texture instructions from the shader, please hold on for 15 minutes, I'll post a new one soon. It's really aggravating that it works perfectly over here but just doesn't render at all over there without any other signs that it failed, though.

[Argh]

OK, let's try this- all of the fragment operations (in the final shader) and OpenGL state changes are removed, it should just show simple points, which should look like white squares. If this still doesn't work, the problem's either on the vertex side of the final shader, or something is wrong with the physics or creation shaders.

[Master-Athmos]

snowy.jpg

(85.96 KiB) Downloaded 23 times

[Argh]

Well... that's good, and bad. At least the physics / creation FBO isn't the problem, that could take another week to find, test and debug.

Now, I'll add the OpenGL state changes back in. Please bear with me- there are three main things gone, all of which are important- OpenGL state changes (depth-testing, etc.), LOS testing (reads from a texture) and finally, colormap (multiple texture reads).

Here is a version with the OpenGL state changes.

[Master-Athmos]

Ok - I now have got a rectangle which is black before game start and shortly after it gets partially white...

EDIT:
The particle system itself works fine just as before...

[Argh]

Sorry, that was the LOS texture, I turned the diagnostic display back on, I'll turn that off.

OK, still working, not the OpenGL. Next up- LOS. LOS is supposed to use the discard operation if the fragment fails, to save GPU a bit. It's possible that's a problem, so if it fails now, at least I know it's not the colormap portion. It didn't result in a huge performance difference, so I can remove it without any real worries one way or another.

[Master-Athmos]

It's still working here...

[Argh]

That is not a relief, for me that is. OK... last bit is colormap, which consists of two operations- one, to chose the base texture, and then a secondary operation to cycle the colors. I'll isolate the first part now, and let's see if we can give it a texture correctly.

[Argh]

Ok, here's simple texture choice logic. You should now see bubbles.

[Master-Athmos]

Now things turned totally invisible again...

[Argh]

Well, that's good, in a sense... let's remove the choice logic, and see if the textured point objects are OK.