[Mlir-commits] [mlir] [mlir][NVVM] Add packed EX2 operations (PR #220501)

[Guray Ozen]

================
@@ -512,6 +512,46 @@ def NVVM_Ex2Op : NVVM_F32UnaryApproxOp<"ex2"> {
   }];
 }
 
+def NVVM_Ex2F16x2Op
+    : NVVM_Op<"ex2.f16x2", [Pure, SameOperandsAndResultType]> {
+  let summary = "Packed f16x2 base-2 exponential (fast approximation)";
+  let description = [{
+    Computes a fast approximation of 2 raised to the power of each f16 element
+    in the input vector. This operation lowers to the `ex2.approx.f16x2` PTX
----------------
grypp wrote:

let's not mention ptx

https://github.com/llvm/llvm-project/pull/220501