[llvm] [mlir] [mlir][AMDGPU] Implement AMDGPU DPP operation in MLIR. (PR #89233)

[Matt Arsenault via llvm-commits]

================
@@ -326,6 +326,68 @@ LogicalResult MFMAOp::verify() {
   return success();
 }
 
+//===----------------------------------------------------------------------===//
+// DPPOp
+//===----------------------------------------------------------------------===//
+LogicalResult DPPOp::verify() {
+  Type srcType = getSrc().getType();
+  if (srcType.getIntOrFloatBitWidth() > 32) {
+    return emitOpError("integer and floating point types larger than 32 bits "
+                       "are not supported");
----------------
arsenm wrote:

i64 has been supported for a while, and it's directly supported on some MI targets too. #91190 handled it for f64 

https://github.com/llvm/llvm-project/pull/89233