Add summary and description

Author: clementval

mlir/include/mlir/Dialect/LLVMIR/NVVMOps.td

@@ -826,6 +826,25 @@ def NVVM_VoteSyncOp
     : NVVM_Op<"vote.sync">,
       Results<(outs AnyTypeOf<[I32, I1]>:$res)>,
       Arguments<(ins I32:$mask, I1:$pred, VoteSyncKindAttr:$kind)> {
+  let summary = "Vote across thread group";
+  let description = [{
+    The `vote.sync` op will cause executing thread to wait until all non-exited
+    threads corresponding to membermask have executed `vote.sync` with the same
+    qualifiers and same membermask value before resuming execution.
+
+    The vote operation kinds are:
+    - `any`: True if source predicate is True for some thread in membermask.
+    - `all`: True if source predicate is True for all non-exited threads in
+      membermask. 
+    - `uni`: True if source predicate has the same value in all non-exited
+      threads in membermask.
+    - `ballot`: In the ballot form, the destination result is a 32 bit integer.
+      In this form, the predicate from each thread in membermask are copied into
+      the corresponding bit position of the result, where the bit position
+      corresponds to the thread’s lane id.
+
+    [For more information, see PTX ISA](https://docs.nvidia.com/cuda/parallel-thread-execution/#parallel-synchronization-and-communication-instructions-vote-sync)
+  }];
   string llvmBuilder = [{
     auto intId = getVoteSyncIntrinsicId($kind);
     $res = createIntrinsicCall(builder, intId, {$mask, $pred});