[llvm-branch-commits] [llvm] [AMDGPU] wip: MIR pretty printing for S_WAITCNT_FENCE_soft (PR #150391)

[Matt Arsenault via llvm-branch-commits]

================
@@ -12,10 +12,135 @@
 //===----------------------------------------------------------------------===//
 
 #include "AMDGPUMIRFormatter.h"
+#include "SIDefines.h"
 #include "SIMachineFunctionInfo.h"
 
 using namespace llvm;
 
+bool parseAtomicOrdering(StringRef Src, unsigned &Order) {
+  Src.consume_front(".");
+  for (unsigned I = 0; I <= (unsigned)AtomicOrdering::LAST; ++I) {
+    if (Src == toIRString((AtomicOrdering)I)) {
----------------
arsenm wrote:

Can we make AtomicOrdering work with enum_seq? 

https://github.com/llvm/llvm-project/pull/150391