[llvm] [ARM] Emit an error when the hard-float ABI is enabled but can't be used (PR #111334)

[Peter Smith via llvm-commits]

================
@@ -1,4 +1,4 @@
-; RUN: llc -mtriple thumbv7em-apple-darwin -o - < %s | FileCheck %s
+; RUN: llc -mtriple thumbv7m-apple-darwin -o - < %s | FileCheck %s
----------------
smithp35 wrote:

This is just a comment for posterity, no action needed as I wanted to check the difference between `thumbv7em-apple-darwin` and `thumbv7m-apple-darwin` and why clang didn't recognise that triple (although it does canonicalise to it internally).

Seems like for Darwin there's a default CPU which for thumbv7em is Cortex-M4 which defaults to the floating point unit being enabled, and also has a special case that hard-codes hard-float. On a non Apple triple soft-float would have been used. 

https://github.com/llvm/llvm-project/pull/111334