[clang] [flang] [clang][driver] When -fveclib=ArmPL flag is in use, always link against libamath (PR #116432)

[Paul Walker via cfe-commits]

================
@@ -490,6 +490,16 @@ void tools::AddLinkerInputs(const ToolChain &TC, const InputInfoList &Inputs,
     else
       A.renderAsInput(Args, CmdArgs);
   }
+  if (const Arg *A = Args.getLastArg(options::OPT_fveclib)) {
+    if (A->getNumValues() == 1) {
+      StringRef V = A->getValue();
+      if (V == "ArmPL") {
+        CmdArgs.push_back(Args.MakeArgString("-lamath"));
+        CmdArgs.push_back(Args.MakeArgString("-lm"));
----------------
paulwalker-arm wrote:

> Flang already has a hack for a different veclib implementation on darwin

The difference with the one instance of doing this is that Accelerate is a core part of that environment, much like libm is a core part of the linux environment.  libamath is not core to any environment.  I recommend seeking advice from the wider clang community before landing this work.

https://github.com/llvm/llvm-project/pull/116432