[compiler-rt] [compiler-rt][AArch64] Provide basic implementations of SME memcpy/memmove in case of strictly aligned memory access (PR #138250)

[Victor Campos via llvm-commits]

================
@@ -22,3 +23,46 @@ const void *__arm_sc_memchr(const void *src, int c,
 
   return NULL;
 }
+
+#ifndef __ARM_FEATURE_UNALIGNED
----------------
vhscampos wrote:

The issue is it isn't either this or that. For example, for AArch64 soft-nofp with unaligned accesses, some of the functions will come from the optimised .S file, and others from the basic .c file.

In order to have it work without the #ifdefs and avoid multiple definition errors, I will have to refactor the files into smaller ones and select them accordingly in the CMakeLists.txt. I am having a go at this option.

https://github.com/llvm/llvm-project/pull/138250