[llvm] [AMDGPU] Refactor insertRelease into insertWriteback + insertWait (NFC) (PR #199486)

[Sameer Sahasrabuddhe via llvm-commits]

================
@@ -442,17 +442,27 @@ class SICacheControl {
                              SIAtomicAddrSpace AddrSpace,
                              Position Pos) const = 0;
 
-  /// Inserts any necessary instructions at position \p Pos relative to
-  /// instruction \p MI to ensure previous memory instructions by this thread
-  /// with address spaces \p AddrSpace have completed and can be observed by
-  /// subsequent memory instructions by any thread executing in memory scope \p
-  /// Scope. \p IsCrossAddrSpaceOrdering indicates if the memory ordering is
-  /// between address spaces. Returns true iff any instructions inserted.
-  virtual bool insertRelease(MachineBasicBlock::iterator &MI,
-                             SIAtomicScope Scope,
-                             SIAtomicAddrSpace AddrSpace,
-                             bool IsCrossAddrSpaceOrdering,
-                             Position Pos) const = 0;
+  /// Inserts any necessary writeback instructions at position \p Pos relative
+  /// to instruction \p MI to make previous memory operations by this thread
+  /// with address spaces \p AddrSpace available to other threads in memory
+  /// scope \p Scope. Does not insert waits; callers must call insertWait
+  /// separately. Returns true iff any instructions inserted.
+  virtual bool insertWriteback(MachineBasicBlock::iterator &MI,
+                               SIAtomicScope Scope, SIAtomicAddrSpace AddrSpace,
+                               Position Pos) const = 0;
+
+  /// Inserts writeback followed by an unconditional wait to implement a
+  /// release operation.
+  bool insertRelease(MachineBasicBlock::iterator &MI, SIAtomicScope Scope,
+                     SIAtomicAddrSpace AddrSpace, bool IsCrossAddrSpaceOrdering,
+                     Position Pos) const {
+    bool Changed = false;
+    Changed |= insertWriteback(MI, Scope, AddrSpace, Pos);
----------------
ssahasra wrote:

I'll go with @Pierre-vh's suggestion. It's preferable to have a patch complete in itself and let the next patch change things if/when it lands.

https://github.com/llvm/llvm-project/pull/199486