Disallow implicit LOD texture methods in Compute shaders. #1221
Description
Using an implicit LOD in a compute shader in SPIR-V causes shader validation errors as it requires extra capabilities.
[[stage(compute)]] fn main() -> void {
result0.value = textureSample(tex0, samp, vec2<f32>(0.5, 0.5)).x;
result1.value = textureLoad(tex1, vec2<i32>(0, 0), 0).x;
return;
}
error: line 61: OpEntryPoint Entry Point <id> '20[%main]'s callgraph contains function <id> 20[%main], which cannot be used with the current execution modes:
ImplicitLod instructions require DerivativeGroupQuadsNV or DerivativeGroupLinearNV execution mode for GLCompute execution model: ImageSampleImplicitLod
Given the extra capabilities required, I think we should disallow the use of the implicit LOD texture methods in compute shaders.