[AArch64][SME] Rewrite __arm_sc_memset to remove invalid instruction (l…

…lvm#101522)

The implementation of __arm_sc_memset in compiler-rt contains
a Neon dup instruction which is not valid in streaming mode. This
patch rewrites the function, using an SVE mov instruction if available.

(cherry picked from commit d6649f2)

compiler-rt/lib/builtins/aarch64/sme-libc-mem-routines.S

@@ -252,7 +252,15 @@ DEFINE_COMPILERRT_FUNCTION_ALIAS(__arm_sc_memmove, __arm_sc_memcpy)
 #define zva_val  x5
 
 DEFINE_COMPILERRT_OUTLINE_FUNCTION_UNMANGLED(__arm_sc_memset)
-        dup     v0.16B, valw
+#ifdef __ARM_FEATURE_SVE
+        mov     z0.b, valw
+#else
+        bfi valw, valw, #8, #8
+        bfi valw, valw, #16, #16
+        bfi val, val, #32, #32
+        fmov d0, val
+        fmov v0.d[1], val
+#endif
         add     dstend2, dstin, count
 
         cmp     count, 96