From e9ce47bb6b830631b68340c005575f74d4e14545 Mon Sep 17 00:00:00 2001
From: Alex Wu <itswu.alex@gmail.com>
Date: Tue, 2 Jun 2020 14:23:50 -0700
Subject: [PATCH] [iter] user-tip for better perf (#8712)

---
 doc/source/iter.rst | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/doc/source/iter.rst b/doc/source/iter.rst
index d46d1a096..728969770 100644
--- a/doc/source/iter.rst
+++ b/doc/source/iter.rst
@@ -198,6 +198,11 @@ distributed training:
     work = [train.remote(shard) for shard in it.shards()]
     ray.get(work)
 
+.. tip:: Using ParallelIterator built-in functions is typically most efficient.
+         For example, if you find yourself using list comprehensions like
+         ``[foo(x) for x in iter.gather_async()]``, consider using
+         ``iter.for_each(foo)`` instead!
+
 API Reference
 -------------