[llvm] Fix mechanism propagating mangled names for TLI function mappings (PR #66656)

[via llvm-commits]

================
@@ -27,11 +27,40 @@ class Triple;
 /// Describes a possible vectorization of a function.
 /// Function 'VectorFnName' is equivalent to 'ScalarFnName' vectorized
 /// by a factor 'VectorizationFactor'.
-struct VecDesc {
+/// The MangledName string holds scalar-to-vector mapping:
+///    _ZGV<isa><mask><vlen><vparams>_<scalarname>(<vectorname>)
+///
+/// where:
+///
+/// <isa> = "_LLVM_"
+/// <mask> = "M" if masked, "N" if no mask.
+/// <vlen> = Number of concurrent lanes, stored in the `VectorizationFactor`
+///          field of the `VecDesc` struct. If the number of lanes is scalable
+///          then 'x' is printed instead.
+/// <vparams> = "v", as many as are the numArgs.
+/// <scalarname> = the name of the scalar function.
+/// <vectorname> = the name of the vector function.
+class VecDesc {
+private:
----------------
JolantaJensen wrote:

Fixed.

https://github.com/llvm/llvm-project/pull/66656