[llvm] [M68k] always use movem for register spills (PR #106715)

[Janis Heims via llvm-commits]

================
@@ -0,0 +1,499 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=m68k -O0 %s -o - | FileCheck %s
+
+define zeroext i1 @get1() {
----------------
TechnoElf wrote:

The test requires these functions to generate values that can't be optimised away, thus forcing a register spill.
Is there a better way to create such values? I've replaced the function `define`s with `declare`s for now as the function body is unnecessary, like you said.

https://github.com/llvm/llvm-project/pull/106715