From e198c19b3cf4505a0e493dd7625230852b68b531 Mon Sep 17 00:00:00 2001 From: Takafumi Arakaki Date: Wed, 29 Sep 2021 01:40:07 -0400 Subject: [PATCH] Tweak README again --- README.md | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/README.md b/README.md index 2c477eb..2675ba8 100644 --- a/README.md +++ b/README.md @@ -18,7 +18,8 @@ the *throughput*) of your program, it is highly recommended to look for ways to **avoid** using concurrent collections first. In particular, consider applying the [data-parallel](https://juliafolds.github.io/data-parallelism/) pattern to dodge the difficulties in concurrent programming. For example, it is often a -better idea to use task-local copies of **non**-thread-safe `Dict` instead of -`ConcurrentDict` shared across tasks. For more information, see: [Efficient and -safe approaches to mutation in data +better idea to use task-local **non**-thread-safe `Dict`s instead of a +`ConcurrentDict` shared across tasks. One of the most important techniques in +data-parallel programming is how to merge such task-local states. For more +information, see, e.g., [Efficient and safe approaches to mutation in data parallelism](https://juliafolds.github.io/data-parallelism/tutorials/mutations/).