Triplanar mapping

[TaaTT4]

This was a feature that I had in my engine for a while and, rather than letting it die with him, I preferred to make it available to the community (since someone also asked for it). As the title says, it adds support for enabling/disabling triplanar mapping on HLMS Terra detail maps.

Some remarks:

• It doesn't provide triplanar mapping to the non-detail maps (so only the diffuse map in terrain context) simply because it wasn't needed in my scenario; with some fairly simple modifications the triplanar mapping can easily be extended to non-detail maps as well.
• User can enable triplanar mapping per texture channels (e.g. for diffuse and normal detail maps only); this allows him to find a compromise between visual quality and performance and/or to save some calculations if he uses monochromatic textures in some channels (e.g. metalness). When triplanar mapping is enabled for some channels, but not for others, the textures of the latters are sampled using world coordinates from a top view projection.

Some screenshots (left w/o triplanar mapping; right w/ triplanar mapping):

[cryham]

Works in DX11, but there is a crash in OpenGL3+ and Vulkan when pressing T (for triplanar), pixel shader has an error.

[TaaTT4]

It seemed I had tested It both in OpenGL and Vulkan, but I could be wrong... What's the error?

[cryham]

So this is in Ogre.log when I press T on GL3+

11:33:26: Vertex Shader: 100000003VertexShader_vs
Fragment Shader: 100000003PixelShader_ps
 GLSL validation result : 
Validation Error: Samplers of different types point to the same texture unit
Validation Error: Samplers of different types point to the same texture unit
Validation Error: Samplers of different types point to the same texture unit
11:33:26: Vertex Shader: 700000006VertexShader_vs
Fragment Shader: 700000006PixelShader_ps
 GLSL validation result : 
Validation Error: Samplers of different types point to the same texture unit
Validation Error: Samplers of different types point to the same texture unit
Validation Error: Samplers of different types point to the same texture unit
Validation Error: Samplers of different types point to the same texture unit
Validation Error: Samplers of different types point to the same texture unit
11:33:26: Vertex Shader: 700000007VertexShader_vs
Fragment Shader: 700000007PixelShader_ps
 GLSL validation result : 
Validation Error: Samplers of different types point to the same texture unit
Validation Error: Samplers of different types point to the same texture unit
Validation Error: Samplers of different types point to the same texture unit
Validation Error: Samplers of different types point to the same texture unit
Validation Error: Samplers of different types point to the same texture unit
11:33:29: Shader Compiler:error(high) 0: SHADER_ID_COMPILE error has been generated. GLSL compile failed for shader 29, "700000008PixelShader_ps (./700000008PixelShader_ps.glsl)": ERROR: 0:1184: 'pow' : no matching overloaded function found (using implicit conversion) 
ERROR: 0:1184: 'pow' : function is not known 
ERROR: 0:1183: '=' :  cannot convert from 'const highp float' to '3-component vector of highp float'
ERROR: 0:1269: '=' :  cannot convert from '4-component vector of highp float' to 'highp float'
ERROR: 0:1277: '=' :  cannot convert from '4-component vector of highp float' to 'highp float'
ERROR: 0:1306: '=' :  cannot convert from '4-component vector of highp float' to 'highp float'
ERROR: 0:1314: '=' :  cannot convert from '4-component vector of highp float' to 'highp float'


11:33:29: GLSL compile log: 700000008PixelShader_ps
ERROR: 0:1184: 'pow' : no matching overloaded function found (using implicit conversion) 
ERROR: 0:1184: 'pow' : function is not known 
ERROR: 0:1183: '=' :  cannot convert from 'const highp float' to '3-component vector of highp float'
ERROR: 0:1269: '=' :  cannot convert from '4-component vector of highp float' to 'highp float'
ERROR: 0:1277: '=' :  cannot convert from '4-component vector of highp float' to 'highp float'
ERROR: 0:1306: '=' :  cannot convert from '4-component vector of highp float' to 'highp float'
ERROR: 0:1314: '=' :  cannot convert from '4-component vector of highp float' to 'highp float'
11:33:29: OGRE EXCEPTION(3:RenderingAPIException): Fragment Program 700000008PixelShader_ps failed to compile. See compile log above for details. in GLSLShader::compile at D:\_\sr\og3\Ogre\ogre-next\RenderSystems\GL3Plus\src\GLSL\OgreGLSLShader.cpp (line 316)
11:33:29: WARNING: GraphicsSystem::deinitialize() not called!!!

Honestly I can't even because the line numbers given in .log point to empty lines in .glsl.

But I suspected and changed the only pow from that triplanar line for blend factors midf3 triplanarBlend =
to some const, and I got other errors next. Which IDK where are (above reason).

[cryham]

Okay found a way, have to delete the block at start from #if 0 to #endif with all parameters dump, then the lines from .log match to right lines in .glsl.

I fixed the first issue with pow by replacing
midf3 triplanarBlend = pow( pixelData.geomNormal, _h( 4.0 ) );
to this
midf3 triplanarBlend = pow( pixelData.geomNormal, midf3_c( 4.0, 4.0, 4.0 ) );
And the second issue lands into all those

midf metalness0 = SampleMetalness0Triplanar( textureMaps2,
midf roughness0 = SampleRoughness0Triplanar( textureMaps2,

saying I can't put a float3 into float.
Fixed it by adding .x in DetailTriplanar_piece_ps.any for SampleMetalness* and SampleRoughness*, in 6 places total, replacing this

arrayIdx ) * triplanarBlend.x +													   \

to this

arrayIdx ).x * triplanarBlend.x +													   \

BTW this happened for me on Windows 10 in GL3+ with integrated Intel GPU and did also on Debian 11 with AMD GPU.

[TaaTT4]

Well spotted, congrats and thank you! :) As you correctly identified it, the problem is that OpenGL and Vulkan don't implicitly cast from float to float3. I'm quite busy in these days, but I'll try to fix it ASAP.

[darksylinc]

Merged manually!

I fixed the problems.

After fixing the pow3, another problem arise which boiled down to:

float4 a,b,c;

#define Foo()  a + b + c
float value = Foo().x;

// Which translates to:
float value = a + b + c.x; // The result is a float4!!! Cannot be casted down implicitly

// Solution:
#define Foo()  (a + b + c)
float value = Foo().x; // Everything's alright now

I was surprised that this works on GL3+ as well! (because GL uses a basic macro parser, and that multi-line macro looked scary)

Thanks for your contribution!

[cryham]

Hi @darksylinc in case you missed my post, I'm also adding link here
https://forums.ogre3d.org/viewtopic.php?p=554263#p554263
Could someone confirm if you have this too? Or is this just something wrong in my code, I merged manually.