[compiler-rt] [libunwind] [AArch64] Fix nofp regressions in compiler-rt and libunwind (PR #111235)

[Keith Packard via cfe-commits]

================
@@ -238,7 +236,8 @@ END_COMPILERRT_OUTLINE_FUNCTION(__arm_sc_memcpy)
 
 DEFINE_COMPILERRT_FUNCTION_ALIAS(__arm_sc_memmove, __arm_sc_memcpy)
 
-
+// This version uses FP registers. Use this only on targets with them
+#if defined(__aarch64__) && __ARM_FP != 0
----------------
keith-packard wrote:

I'm not sure; the previous version protected this with `#ifdef __aarch64__`, so I just kept that in place. Seemed odd to me as well.

https://github.com/llvm/llvm-project/pull/111235