[clang] [HLSL][Matrix] Add support for Matrix element and trunc Casts (PR #168915)

[Chris B via cfe-commits]

================
@@ -12591,6 +12592,19 @@ void Sema::CheckImplicitConversion(Expr *E, QualType T, SourceLocation CC,
   if (auto VecTy = dyn_cast<VectorType>(Target))
     Target = VecTy->getElementType().getTypePtr();
 
+  if (isa<ConstantMatrixType>(Source)) {
+    if (!isa<ConstantMatrixType>(Target)) {
+      return DiagnoseImpCast(*this, E, T, CC, diag::warn_impcast_matrix_scalar);
+    } else if (getLangOpts().HLSL &&
----------------
llvm-beanz wrote:

Should this warning be emitted for "if source is a constant matrix and destination is not" or "if source is a constant matrix and destination is a scalar"? The latter seems more correct to me.

I'm also a bit unclear about the language mode requirements for this. I assume this can't actually occur in non-HLSL code. If it can, we should have some tests for the C/C++ exposure of this warning.

For the next block, I also suspect that we couldn't get this far with an AST that isn't from HLSL source, but I'm not sure the `getLangOpts().HLSL` check is really required there. I assume that if C/C++ ever had an implicit conversion for matrix truncation they would also want the compiler to emit a warning (@fhahn thoughts?).

Code style nit:
```suggestion
    if (!isa<ConstantMatrixType>(Target))
      return DiagnoseImpCast(*this, E, T, CC, diag::warn_impcast_matrix_scalar);
    if (getLangOpts().HLSL &&
```
see: https://llvm.org/docs/CodingStandards.html#don-t-use-else-after-a-return

https://github.com/llvm/llvm-project/pull/168915