[llvm] Add out-of-line-atomics support to GlobalISel (PR #74588)

[Thomas Preud'homme via llvm-commits]

================
@@ -758,16 +758,39 @@ AArch64LegalizerInfo::AArch64LegalizerInfo(const AArch64Subtarget &ST)
           all(typeInSet(0, {s8, s16, s32, s64, s128}), typeIs(2, p0)));
 
   getActionDefinitionsBuilder(G_ATOMIC_CMPXCHG)
+      .libcallIf([&ST](const LegalityQuery &Query) {
+        return ST.outlineAtomics() && !ST.hasLSE();
+      })
       .customIf([](const LegalityQuery &Query) {
         return Query.Types[0].getSizeInBits() == 128;
       })
       .clampScalar(0, s32, s64)
       .legalIf(all(typeInSet(0, {s32, s64}), typeIs(1, p0)));
 
+  getActionDefinitionsBuilder({G_ATOMICRMW_XCHG, G_ATOMICRMW_ADD,
+                               G_ATOMICRMW_AND, G_ATOMICRMW_OR,
+                               G_ATOMICRMW_XOR})
+      .libcallIf([&ST](const LegalityQuery &Query) {
+        return ST.outlineAtomics() && !ST.hasLSE();
+      })
+      .clampScalar(0, s32, s64)
+      .legalIf(all(typeInSet(0, {s32, s64}), typeIs(1, p0)));
+
+  getActionDefinitionsBuilder(G_ATOMICRMW_SUB)
+      .lowerIf([&ST](const LegalityQuery &Query) {
----------------
RoboTux wrote:

It is lowerIf because it is lowered in terms of G_ATOMICRMW_ADD in lower(). Because it has a different action (lower instead of libcall) I put it separately.

I could put the code in libcall instead if you think it is a better approach.

https://github.com/llvm/llvm-project/pull/74588