Matrix-float multiplication cross-compilation fails

[nbentyNV]

float f;
mat4 a;
mat b = a*f; // This is legal in GLSL but fails compilation inside Slang

Looks like it's a Slang bug, since the GLSL compiler is never invoked.

We are using it for vertex-blending- ShaderCommon.slang::59 - getBlendedWorldMat

[tangent-vector-personal]

Can you copy-paste in the error message you see?

My guess is that I'm just missing some overloads of operator*. I hope it isn't something more serious.

[tangent-vector-personal]

One more note-to-self issue here is that I'm realizing there could be subtle problems if cross-compilation involves scalar-to-matrix conversion. float4x4(a) produces a matrix with 16 elements of a, while mat4(a) produces a diagonal matrix.

There is a similar issue where I need to make sure to translate HLSL operator* on matrices to the appropriate componentMul call.

[nbentyNV]

[tangent-vector]

I think I misunderstood this error. The problem is not that we are giving an error on a matrix-scalar multiply in GLSL code, but rather that we are giving the error on Slang code that tries to do matrix-times-scalar.

It seems like this error shouldn't be specific to cross-compilation, then, and should occur for any Slang code that tries to do this.

[tangent-vector]

Yep. I can repro this is a bare Slang file, so this is just a plain old screw-up on my part.

There are two reasonable ways to fix this:

• Implement implicit promotion of scalars to matrices, after which the ordinary component-wise multiply will be taken
• Implement explicit overloads of operator* (and probably others) for the matrix/scalar and scalar/matrix cases (and same for vector/scalar, ...)

It isn't clear to me which is the way that HLSL interprets this kind of things internally.