[llvm] [AMDGPU] Add intrinsic readanylane (PR #115696)

[Jay Foad via llvm-commits]

================
@@ -2152,6 +2152,13 @@ def int_amdgcn_readfirstlane :
   Intrinsic<[llvm_any_ty], [LLVMMatchType<0>],
             [IntrNoMem, IntrConvergent, IntrWillReturn, IntrNoCallback, IntrNoFree]>;
 
+// This is similar to readfirstlane, but marks value that is uniform, allowed sunk into
+// control flow. The result is undefined if the value is actual divergent.
+def int_amdgcn_readanylane :
+  Intrinsic<[llvm_any_ty], [LLVMMatchType<0>],
+            [IntrConvergent, IntrNoCallback, IntrNoFree, IntrNoMem, IntrSpeculatable,
----------------
jayfoad wrote:

Did you add IntrSpeculatable deliberately? If so, you should add a test showing what sort of transform it enables.

https://github.com/llvm/llvm-project/pull/115696