[llvm] [OpenMP] Make each atomic helper take an atomic scope argument (PR #122786)

[Joseph Huber via llvm-commits]

================
@@ -28,144 +28,145 @@ enum OrderingTy {
   seq_cst = __ATOMIC_SEQ_CST,
 };
 
-enum ScopeTy {
+enum MemScopeTy {
   system = __MEMORY_SCOPE_SYSTEM,
-  device_ = __MEMORY_SCOPE_DEVICE,
+  device = __MEMORY_SCOPE_DEVICE,
   workgroup = __MEMORY_SCOPE_WRKGRP,
   wavefront = __MEMORY_SCOPE_WVFRNT,
   single = __MEMORY_SCOPE_SINGLE,
 };
 
-enum MemScopeTy {
-  all,    // All threads on all devices
-  device, // All threads on the device
-  cgroup  // All threads in the contention group, e.g. the team
-};
-
 /// Atomically increment \p *Addr and wrap at \p V with \p Ordering semantics.
 uint32_t inc(uint32_t *Addr, uint32_t V, OrderingTy Ordering,
-             MemScopeTy MemScope = MemScopeTy::all);
+             MemScopeTy MemScope = MemScopeTy::device);
----------------
jhuber6 wrote:

This changes the default scope to device instead of system, which affects reductions. however, I believe it's fine since we don't need that part to be coherent with the CPU's memory.

https://github.com/llvm/llvm-project/pull/122786