Any ATi users able to compile either the lighting shader or the normalmap shader?

Works fine for me with ATI:


Which Shader Model did you use btw - will the graphics card need model 2 or 3?

What hardware is that, Master Athmos? It's thrilling to think that ATi might actually be able to run this stuff, finally.

I don't see the dynamic light source, though. Did you see the lights?

ATI x1950 on current drivers on WinXp : no errors ,
i see no normal maps or any kind of selfshadowing or shader fancyness

added infolog on your request

Post your Infolog. Can't debug for ATi if I don't have 'em. Good to know it doesn't crash, of course, but I'd prefer to make it actually work.

[EDIT]Thanks![/EDIT]

OK... I think I have both of ash2Life's reported problems solved, and resolved imbazcek's issues.

ATi's very strict drivers only support the 1.X GLSL featureset. Why that means that some of their cards apparently support 2.X... who knows, and who cares.

What matters is that the "transpose" operation is not valid for (certain) ATi hardware.

Soooo... I wrote a manual transposition of the matrices (took me awhile, I am not familiar with matrix math and had to go read about it), and fixed the reference to "xyza" in the GLSL lighting shader, which ATi's overly-finicky drivers won't take (but they should take "xyzw", which is standards-correct... I hope).

Version 4 is now available. Tested, and runs on nVidia hardware as advertised, and doesn't seem to have changed the speed any.

I am hoping (fingers crossed) that, at the very least, the lighting shader works as advertised now. The normalmap shader... it depends on how ATi's hardware handles the "too many floats" issue that necessitates a downgraded version of the first shader, for people with older (but theoretically shader-supporting) hardware.

ATi users, please test again and report!

30 fps with 100 of these on dsd. 8800gt, 8x AA, 1920*1200, 4096 shadows.
Looks nice!

Cool.

Uploaded v0.5, which features a fix for the depthmap version with shadows.

If you have a GLSL 2.0 capable card, you should get all of the bells and whistles now. With depthmaps, it should look a bit different than the screens (and you'll see self-shadowing, so it'll be obvious).

Gotta get some sleep for a few hours now. Keep sending in infologs if things are broken, I really wanna know if ATi is somewhat un-borked yet.

Ah - ok. With the latest version I now also get an area where there's a colored lighting:



I'm using a HD4830 and for the latest version there are no errors in the infolog...

now theres a changing colored light on both rotating objects - the surface and texture is still a standard

Great, the lighting works for ATi, finally.

Show me the infologs, I'll see which line the normalmapping is stopping on now, ash

late comer

Sorry, I had to sleep.

Whee... WARNING: built-in varying gl_TexCoord [3] has mismatched access semantics between the vertex and fragment shader

Sounds like more "fun" with the GLSL error references, I haven't the foggiest what that one means

Version 6 is now up, which should (hopefully) resolve ash2life's reported issue. Please let me know.

On an ATI? check if the texCoord in question is being accessed in both the vertex and fragment shader. The ATI drivers are picky about that, if it's only accessed in the vertex shader and not the fragment shader it'll spam that warning and perhaps even recompile the shader every frame (that's what happened when I struggled with a similar problem at work at least).

i see a bumpmap now -
makes these spheres look like they have a more real surface than just being a shiny metalball

Sweet! Glad I got it running for ATi on the lower end, finally!

Olemar, you were right on the money. I caught that elsewhere online and was like... "oh, yeah, I'm still referring to junk I'm not using in shader2"... another example of my brilliant coding

OK... now for the hard part.

I need to get shadowmaps running, for self-shading on older hardware (such as mine).

I can't figure out how to shoehorn it into the main shader pass for older hardware, there aren't enough floats available. But... I was thinking... maybe render it as a second pass, get whether it's in shadow or not, since that's a binary state (no soft shadows or volumetric stuff to worry about thank, goodness).

Kloot told me it'd be a mess, but I'm going to try anyhow. I figure I'll just re-render the unit geometry with an offset.

OK... got the first part working. Self-shadows working as an entirely separate operation.

It took me a minute, because I didn't know how the new Unit Materials worked.

Still working on one last wrinkle.

Crashes for me after countdown...

http://pastebin.com/m6bcba7d8

the normal map stuff in pure used to work fine on my 8800 (which died a week ago)