[llvm] [AMDGPU] Overload image atomic swap to allow float as well. (PR #107283)

[Jay Foad via llvm-commits]

================
@@ -1023,26 +1029,31 @@ defset list<AMDGPUImageDimIntrinsic> AMDGPUImageDimIntrinsics = {
 //////////////////////////////////////////////////////////////////////////
 defset list<AMDGPUImageDimIntrinsic> AMDGPUImageDimAtomicIntrinsics = {
   multiclass AMDGPUImageDimAtomicX<string opmod, list<AMDGPUArg> dataargs,
-                                   int isFloat = 0> {
+                                   int isFloat = 0, int isAny = 0> {
----------------
jayfoad wrote:

Instead of adding more `bool` arguments here, why not take a single `LLVMType` argument which callers can set to `llvm_anyint_ty`/`llvm_anyfloat_ty`/`llvm_any_ty`?

https://github.com/llvm/llvm-project/pull/107283